No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Machine learning based very short term load forecasting of machine tools
Abstract
With the ongoing integration of renewable energies into the electrical power grid, industrial energy flexibility gains importance. To enable demand response applications, knowledge about the future energy demand is necessary. This paper presents a machine learning process to forecast the very short term load of two machine tools, which can be utilized as a decision support basis for control schemes and measures to increase energy flexibility and decrease energy cost in manufacturing. The presented process is developed and evaluated on production machines in a research factory. The results indicate that the developed machine learning process is feasible and creates an accurate very short term load forecasting model for different production machines. It can be used as a blueprint to develop load forecasting models for other production machines using the historic load profile and various machine and process data. A combination of time series features and an Artificial Neural Network proves to be the most robust model regarding the presented machine tools with achieved coefficients of determination between 0.57 and 0.64 for a 100 step forecast. Improvements are still needed regarding the forecasting accuracy, especially of load peaks, for which different measures are proposed.
... Historical data is the fundamental ingredient of the ML-based prediction models. Although most construction organizations and institutions have accident databases, they are hardly used regarding advanced analytics [11], since comprehensive and detailed preprocessing should be applied with a list of issues related to the quality of the data [12,13] such as missing data problem, outliers, different types and scaling of features, and imbalanced class distribution [7]. Therefore, the requirement of multistep comprehensive preprocessing is suggested in the literature [14] to minimize erroneous and misleading results. ...
... It is also well acknowledged that the majority of the time required for the entire process in ML applications was devoted to preprocessing [13]. These facts highlight that building a foundation for ML implementations with a robust data preprocessing phase could become as significant as developing a robust ML model [12]. Despite these, the effects of different preprocessing techniques on the performance of the proposed models with a scenario-basis investigation have not been explored comprehensively in the literature. ...
... Such a process takes the users out of the tedious loop of searching for the best preprocessing scenario [46]. In this vein, one of the preliminary studies that adopted a comprehensive preprocessing approach was performed by Dietrich et al. [12]. The researchers implemented an iterative preprocessing procedure including MDH, OUTH, DSCL, and DENC to predict the short-term electric load of production machines. ...
Occupational accidents are common in the construction industry, therefore developing prediction models to detect high severe accidents would be useful. However, existing studies are limited and usually focus on selecting the most appropriate machine learning method rather than identifying the most effective preprocessing pipeline before the prediction. In this study, a scenario-basis automated preprocessing model that identifies the best scenario is developed to predict the severity of construction accidents. The results show that the scenario combination of not removing missing data, not applying data binning, considering outliers, applying Min-Max-Scaler and one-hot encoding, and data resampling with random oversampling yielded the highest prediction performance with 0.6092 of F1-score. Permutation importance of XGBoost analysis indicates that year, cause material, age, past accidents, experience, and salary are the most influential attributes. This study contributes to society/practice through a model preventing high-severe accidents and theory/technology with novel preprocessing model to perform more reliable predictions.
... The research of load forecasting can be divided into two broad categories, traditional or classical approaches and data-driven approaches [6][7][8][9]. The traditional approaches are time-series analysis and regression analysis. ...
... The machine learning algorithms often used in load forecasting are artificial neural network (ANN), support vector regression (SVR), random forest (RF), long short-term memory (LSTM), knearest neighbors (KNN). Among these algorithms, the artificial neural network is the most widely used algorithm [5,7]. For instance, Dagdougui et al. [14] applied ANN to forecast very short-term load and short-term load in district buildings. ...
Electricity is important in our modern life and also essential to the development of the country. Since the electricity load consumption in Thailand increases almost every year, the power systems capacity expansion is unavoidable. The purpose of this paper was to apply the machine learning algorithms to forecast the long-term electricity load consumption in Thailand. Three algorithms employed were artificial neural network (ANN), support vector regression (SVR), and k-nearest neighbors (KNN). The performances were evaluated in terms of mean absolute percentage error (MAPE). The results showed that the ANN outperformed other algorithms with a MAPE of 2.586%.
... Short-term load forecasting is divided into traditional methods and artificial intelligence methods. In the traditional methods, the regression analysis algorithm represented by Auto-Regression Integrated Moving Average model (ARIMA) [4] and the machine learning models represented by Support Vector Machine (SVM) [5,6], Random Forest [7,8], XGBoost [9], and Sparse Algorithm [10] are used. With the increasing complexity of energy systems and the diversification of the energy demand, it is difficult to obtain satisfactory prediction targets using the regression analysis method. ...
... Figure 8 is the comparison histogram of the model prediction errors for different combinations of CNN and LSTM hyper-parameters. The number of convolutional kernel hyper-parameters of CNN and the number of neuron hyper-parameters of LSTM are [7,14,21] and [8,16,24,32], which are combined to compare the prediction errors of the electricity, heating, and cooling loads. The CNN filters equal to 7 and LSTM units equal to 16 had the least prediction error in terms of the electricity load prediction performance, while the combination with more convolutional kernels did not perform better. ...
Integrated Energy Systems (IES) are an important way to improve the efficiency of energy, promote closer connections between various energy systems, and reduce carbon emissions. The transformation between electricity, heating, and cooling loads into each other makes the dynamic characteristics of multiple loads more complex and brings challenges to the accurate forecasting of multi-energy loads.Improving the accuracy of multi-energy load forecasting for integrated energy systems (IES) helps to optimize the dispatch of power systems to make better use of renewable energy. In order to further improve the accuracy of IES short-term load forecasting, we propose [d=b]the Convolutional Neural Network, the Long Short-Term Memory Network, and Auto-Regression (CLSTM-AR) combined with the multi-dimensional feature fusion (MFFCLA)CLSTM-AR multi-energy load forecasting model based on multi-dimensional feature fusion (MFFCLA). In detail, [d=b]CLSTM canthe Convolutional Neural Network and the Long Short-Term Memory Network (CLSTM) extract the coupling and periodic characteristics implied in IES load data from multiple time dimensions. [d=b]ARAuto-Regression (AR) takes load data as the input to extract features of sequential auto-correlation over adjacent time periods. Then, the [d=b]diverse and effectivemultiple effective features extracted by CLSTM, LSTM, and AR can be fused using the multi-dimensional feature fusion technique. Ultimately, the model achieves the accurate prediction of multiple loads. In conclusion, compared with other forecasting models, the case study results show that MFFCLA has higher forecasting precision compared with the comparable model in the short-term multi-energy load forecasting performance of electricity, heating, and cooling. The accuracy of MFFCLA can help to optimize and dispatch IES to make better use of renewable energy.
... However, these approaches imply that the forecasting approach requires a huge and complex amount of data, incorporating the building's composition, the physical characteristics, the weather conditions, and societal behavioral information, which is usually hard to find due to the time and effort required for acquiring such kind of data. They also exhibit several limitations in that they 1) are incapable to handle huge volumes of data (Ibrahim et al., 2020); 2) lack toughness against ambiguity (Dietrich et al., 2020); 3) lack efficient optimization in the existence of controls (Nam et al., 2020); 4) lack adjustive and self-sufficient functionality (Luo et al., 2020); 5) lack smart and real-time decision making (Antonopoulos et al., 2020); 6) fail to model complex features and high non-linearity incurred by consumer's energy-exploitation activities. As a result, these approaches are not commonly employed for precise load forecasting . ...
... Among them, the XGBoost was reported to be more robust load predicted. Dietrich et al. (2020) adopted the ANN to predict the very short-term loads of two machine tools in order to act as a decision maker to improve energy management and reduce corresponding manufactural costs. Dai and Zhao (2020) address the limitation of SVM and introduced some enhancements to perform effective prediction based on informative features extracted by the minimal redundancy maximal relevance. ...
Developing an appropriate model for accurate prediction of energy consumption is very essential for developing an effective energy management system for residential buildings. In view of this, the Short-term Load Forecasting (STLF) of household appliances has been performing an important role in supervising and managing energy in the residential community. In the domain of big data analytics, data-driven load forecasting approaches have realized an amazing performance in the recognition of patterns of residential electric loads and forecasting energy consumption. Nevertheless, current research emphasizes the use of powerful feature-engineering methods, which are ineffective and result in low generalization performance. Further, considering the differences in the consumption behavior of various home appliances, it is unfeasible to discover energy consumption characteristics physically in the power system. Thus, this study addresses the problems of STLF using a novel two-stream deep learning (DL) model called STLF-Net. The first stream is designed with Gated Recurrent Units (GRUs) to learn and capture the long-term temporal representations of the energy utilization data. Simultaneously, in the second stream, the short-term information and positional representations are modeled using a stack of temporal convolutional (TC) modules. The TC module is designated using dilated causal convolutions and residual connection to enable efficient feature extraction while alleviating the gradient vanishing issues. The learned representations from the two streams are fused and subsequently passed to several dense layers to generate the final hour-ahead load forecasts. Experimental assessments on two public energy consumption predictions datasets (IHEPC and AEP) demonstrated the superior performance of the STLF-Net over the recent cutting-edge data-driven approaches.
... Among the commonly used ML-based methods, application of multilayer perceptron (MLP)-based neural network [6]- [10], support vector machine (SVM) [11], random forest (RF) [12], linear regression (LR) [13], etc., were found. As feature engineering, in [7], to deal with the lagged input variables time-series decomposition was proposed. ...
... In [8], a decision-tree-based approach was performed for feature engineering using the bootstrap aggregated RF ensemble method. Recursive feature elimination was applied in [10], where features are removed until the specified number of features is reached. The method is subjected to perform optimally upon the choice of the number of specified features. ...
To meet up fluctuations of the real-time electric load demands, many electricity markets have gone for the real-time market-based operation. To do so, online forecasting of the real-time load demand is necessary. Due to changes in the relation between impacting variables and output over time, a continuous learning-based approach is highly desired. A fixed dataset-based training may perform accurately for a certain amount of time, but as the load pattern and impact of different external variables change, the performance of such a model may decrease. Thus, to overcome such inherent problem of fixed-sized databased forecasting model development, in this work, novel simultaneous online learning and feature engineering based appropriate time-delay neural network has been proposed. The data for developing a forecasting model is collected through different sensors available in the Internet of Things (IoT) based networks. To develop an optimal-cost effective IoT network and a parsimonious model for load forecasting, variable type-dependent correlation considering the multi-collinearity has been performed in online training. The proper choice of the model has been also proved using numerical analysis with the help of time delay embedding theory. Interestingly it is found that, with the proper choice of inputs and their lagged variables, the proposed model performs better over general feedforward, general regression neural networks, several deep learning and advanced models including recurrent neural networks, fully connected deep neural networks, and Dendritic Neuron Model.
... Thus, the Machine Learning (ML) approach is introduced that can study the information, detect patterns and make decisions with least human intervention. So, ML method and optimization technique have drawn the attention of researchers to improve the precision of LF (Walther et al. 2019;Dietrich et al. 2020). Artificial Neural Networks (ANNs) are designed to imitate the functions as well as structure of the human nervous system to resolve complicated issues. ...
The electric load demand is thriving faster day-to-day which is highly nonlinear, complex and noisy in nature because of its partial dependency on weather condition (temperature, humidity, pressure and etc.). This makes the load prediction extremely difficult. In this work, the performance of Empirical Mode Decomposition (EMD) based optimized Extreme Learning Machine (OELM) is demonstrated for day ahead load forecasting of Chhattisgarh state of India. EMD method is used to disintegrate the nonlinear load data into some simple and stationary dataset by which the prediction competency of machine learning algorithm can be enhanced. Each decomposed dataset is predicted individually by dedicated OELM. OELM learning algorithm is established by optimizing the parameters of ELM (weights and biases) using Crow Search Algorithm (CSA). Further, the competency of OELM is improved by proposing Craziness based CSA (CCSA) with upgraded potential to equipoise between investigation and exploitation. The consolidation of EMD method and proposed CCSA is applied to achieve better efficacy in load demand prediction. The predicted demand of proposed predictive model is demonstrated by using performance measures and hypotheses tests. The superiority of proposed predictive model over recently published work is substantiated for load demand forecasting of three different cities of Australia. The simulation results contribute the indication that the proposed EMD CCSA OELM model can be taken as a better tool for load forecasting.
... Currently, there are few studies on predicting demand-side carbon emission factors. Load is the initial factor that affects the carbon emission factors of each node in power grids, and load prediction methods for power grids are relatively well-established and developed [11], including a statistical method based on historical data [12], artificial intelligence algorithm based on machine learning and deep learning [13][14][15], a combination of long-term and short-term memory [16][17][18] and other methods. So, a neural network model with load as input can predict the corresponding carbon emission factors in the future. ...
Advanced carbon emission factors of a power grid can provide users with effective carbon reduction advice, which is of immense importance in mobilizing the entire society to reduce carbon emissions. The method of calculating node carbon emission factors based on the carbon emissions flow theory requires real-time parameters of a power grid. Therefore, it cannot provide carbon factor information beforehand. To address this issue, a prediction model based on the graph attention network is proposed. The model uses a graph structure that is suitable for the topology of the power grid and designs a supervised network using the loads of the grid nodes and the corresponding carbon factor data. The network extracts features and transmits information more suitable for the power system and can flexibly adjust the equivalent topology, thereby increasing the diversity of the structure. Its input and output data are simple, without the power grid parameters. We demonstrated its effect by testing IEEE-39 bus and IEEE-118 bus systems with average error rates of 2.46% and 2.51%.
... In addition, Li [21] experimented with an ANN model to forecast the PC for smart grid management. Dietrich et al. [22] created an ANN to forecast how much electricity is used by machine tools in a research facility. Their research emphasized that power management and cost reduction in industrial settings play a critical part in machine tool PC prediction. ...
The rise in power consumption (PC) is caused by several factors such as the growing global population, urbanization, technological advances, economic development, and growth of businesses and commercial sectors. In these days, intermittent renewable energy sources (RESs) are widely utilized in electric grids to meet the need for power. Data-driven techniques are essential to assuring the steady operation of the electric grid and accurate power consumption and generation forecasting. Conversely, the available datasets for time series electric power forecasting in the energy industry are not as large as those for other domains such as in computer vision. Thus, a deep learning (DL) framework for predicting PC in residential and commercial buildings as well as the power generation (PG) from RESs is introduced. The raw power data obtained from buildings and RESs-based power plants are conceded by the purging process where absent values are filled in and noise and outliers are eliminated. Next, the proposed generative adversarial network (GAN) uses a portion of the cleaned data to generate synthetic parallel data, which is combined with the actual data to make a hybrid dataset. Subsequently, the stacked gated recurrent unit (GRU) model, which is optimized for power forecasting, is trained using the hybrid dataset. Six existent power data are used to train and test sixteen linear and nonlinear models for energy forecasting. The best-performing network is selected as the proposed method for forecasting tasks. For Korea Yeongam solar power (KYSP), individual household electric power consumption (IHEPC), and advanced institute of convergence technology (AICT) datasets, the proposed model obtains mean absolute error (MAE) values of 0.0716, 0.0819, and 0.0877, respectively. Similarly, its MAE values are 0.1215, 0.5093, and 0.5751, for Australia Alice Springs solar power (AASSP), Korea south east wind power (KSEWP), and, Korea south east solar power (KSESP) datasets, respectively.
... The analysis is based on consideration of both local and international variables with consideration of unemployment rate, global oil prices, and several foreign visitors. In, [20] developed a methodology integrated with a genetic algorithm for sales revenue forecasting for the travel agency to achieve desired results. In, [21] concentrated on the booking revenue management for cancellation rates forecasting through effective data mining techniques. ...
Forecasting is involved in the estimation of statements about particular events concerned those are uncertain events or computation of future. The ultimate purpose ofthe forecasting model is to acquire knowledge about uncertain events. Medical tourism is an emerging economy in several countries with incipient of apatient from one country to diverse countries for medical needs. However, due to the pandemic of COVID-19 patients were subjected to peregrinate restrictions that affect the movement of people. In this scenario, room booking is a substantial concern for the management of resources to withstand demand for those who are engaged in medical tourism. This paper presented a Forecast Demand Artificial Intelligence (FdAI) model for demand forecasting. The proposed FdAI model incorporates a Q-learning-based reinforcement learning model for the automatic computation of demand estimation. The proposed FdAI model remains processed with ARIMA, SARIMA, and Prophet model for accurate of demand forecast related to medical tourism. Proposed FdAI–ARIMA provides the standard error value variance of 0.0001527, the statistics variance value of 8.5467 witha p-value of 1.2671e− 17. The simulation analysis expressed that the proposed FdAI model accurately estimates the demand higher month with minimal Mean square error. Also, the proposed FdAI computed that demand forecasting becomes significantly increased in the upward trend.
... Such algorithms deal with the construction of machines that move automatically by gaining experience, the formation of these algorithms with low computational costs, the design of new algorithms and the usability of big data have made progress in recent years. Since artificial intelligence has a very wide usage area, studies on artificial intelligence can be found in nearly every subject when reviewing the literature [16][17][18][19][20][21][22][23][24][25][26][27][28]. Due to artificial intelligence, computers can be programmed to perform specific tasks or process [29], desired classifications can be made [30][31][32][33], models can be designed, and these models can make predictions about the future [21,34], based on previous experiences or dataset presented as examples [35]. ...
Human activities are linked to atmospheric pollution and are affected by economic development. Ground-level ozone has become an important and harmful pollutant for many countries, adversely affecting public health. As there is a limited number of on-site measurements, alternative methods are required to accurately estimate ozone concentrations. In this study, a database containing annual average concentrations of CO2, N2O, CO, NOx, SOx, and O3, covering the years 2008-2018 for ten countries in Europe, was created. Ten different artificial intelligence regression methods were developed to predict the O3 concentration using these variables. The predictive performance of the developed artificial intelligence models was compared using the coefficient of determination, mean absolute error, root mean square error, and relative absolute error criteria. Experimental results show that the Bagging-MLP method has a better predictive performance than other models in ozone concentration estimation, with an R 2 value of 0.9994, mean absolute error of 24.67, root mean square error of 33.85, and relative absolute error of 2.9%. This study shows that the O3 concentration can be estimated very close to the actual value by using the Bagging-MLP method, one of the artificial intelligence methods. Bagging-MLP Yöntemiyle Troposferik Ozon Konsantrasyonunun Tahmini ÖZ İnsan faaliyetleri atmosfer kirliliği ile bağlantılıdır ve ekonomik gelişmelerden etkilenir. Yer seviyesindeki ozon birçok ülke için önemli ve zararlı bir kirletici haline gelmiş olup halk sağlığını olumsuz etkiler. Yerinde yapılan ölçümlerin sınırlı sayıda olmasından dolayı, ozon konsantrasyonlarını doğru bir şekilde tahmin etmek için alternatif yöntemlere ihtiyaç vardır. Bu çalışmada, Avrupa'da on ülkede 2008-2018 yıllarını kapsayan CO2, N2O, CO, NOx, SOx, ve O3 yıllık ortalama konsantrasyonlarını içeren bir veritabanı oluşturuldu. Bu değişkenleri kullanarak O3 konsantrasyonunu tahmin etmek için on farklı yapay zeka regresyon yöntemi geliştirildi. Geliştirilen yapay zeka modellerinin tahmin performansı, determinasyon katsayısı, ortalama mutlak hata, kök ortalama karesel hata ve göreceli mutlak hata ölçütleri kullanılarak karşılaştırıldı. Deneysel sonuçlar, Bagging-MLP yönteminin diğer modellere göre ozon konsantrasyonu tahmininde daha iyi bir performansa sahip olduğunu, R 2 değeri 0.9994, ortalama mutlak hata 24.67, kök ortalama karesel hata 33.85 ve göreceli mutlak hata ise %2.9 olarak ortaya koydu. Bu çalışma, yapay zeka yöntemlerinden olan Bagging-MLP yöntemi kullanılarak O3 konsantrasyonunun gerçek değere oldukça yakın bir şekilde tahmin edilebileceğini göstermektedir.
... Liang et al. (2018) developed an Artificial Neural Network (ANN) energy model to accurately predict the power profile of the cutting workpiece for energy efficiency improvement and machining strategy optimization. To forecast the short-term power demand of the milling machine tool, an ANN model was developed considering different types of workpieces, and the trained ANN model also was applied to other machine tools by transfer learning, but the accuracy was only 70% (Dietrich et al., 2020). The performance of ANN model and SVM model in predicting the energy consumption of the CNC micro-machining center with the process parameters such as spindle speed, feed rate, cut depth and width was compared in reference (Kant & Sangwan, 2015). ...
As the most promising and advanced technology, ultra-precision machining (UPM) has dramatically increased its production volume for wide-range applications in various high-tech fields such as chips, optics, microcircuits, biotechnology, etc. The concomitantly negative environmental impact resulting from huge-volume UPM has attracted unprecedented attention from both academia and industry. Accurate energy prediction of ultra-precision machine tools (UPMTs) can provide significant insight into energy planning, machining strategy, and energy conservation. Data-driven models for predicting energy have become increasingly popular due to their high accuracy and low modeling difficulty. However, existing data-driven models only focus on ordinary precision machine tools, and their applications on UPMTs are hardly studied. To fill the gap, this paper proposed a data-driven model constructed with 1DCNN-LSTM-Attention layers for predicting the instantaneous power profile of a five-axes UPMT. In the data-preparation phase, an advanced G-code interpreter was developed to generate the working status dataset from the G-code command and accurately match them with the power data collected. Random hyperparameters searching method was adopted to tune the 1DCNN-LSTM-Attention structure for better accuracy in the model creation phase. Finally, the sensitivity of these hyperparameters on the model performance was analyzed. Results demonstrate that the learning rate, 1DCNN, LSTM and dense layer numbers are identified as critical parameters affecting the model performance. The optimized 1DCNN-LSTM-Attention model outperforms other models, achieving an R² value of 0.93. This work first validate the feasibility of utilizing advanced machine learning techniques for predicting energy consumption in UPM field, which can further promoting energy-efficient and sustainable UPM practices by digitalizing the energy consumption process.
... Load forecasting is the way of anticipating future electric power based on previous data and the weather conditions [9]. There are several load forecasting horizons employed by the power system companies for different applications in the industries [10]. These applications include planning [11], control [12], future load scheduling [13], staff hiring [14] and equipment expansion [15]. ...
This paper investigates the effect of Support Vector Regression hyperparameters optimization on electrical load prediction. Accurate and robust load prediction helps policy makers in the energy sector to make inform decision and reduce losses. To achieve this, Bayesian optimization technique was employed for the hyperparameters optimization which are then used for the load prediction. The hyperparameters are the regularization parameters and the epsilon. In addition, the effects of sliding window during the load prediction were also evaluated. The sliding window values were varied from 1 to 5. The results showed that the sliding window of 1 had the optimized hyperparameters with the best performing evaluation metrics of 0.01912 and 0.09493 for MSE and MAE respectively.
... This may pose serious issues, including missing data, outliers, different types and scaling of features, and imbalanced class distribution (Koc et al. 2021). Therefore, the captured data must be comprehensively preprocessed to eliminate erroneous and misleading conclusions (Dietrich et al. 2020). ...
Nonconformance (NCR) has long been a subject of research interest for its potential to extrapolate information leading to a more productive environment in construction projects. Despite a variety of traditional attempts, a systematic understanding of how machine learning (ML) approaches can contribute to proactively detecting the severity of NCRs remains limited. This study aims to develop a data-driven ML framework to predict the cost impacts of NCRs (high severity versus low severity) in construction projects. To accomplish this aim, the random forest (RF) algorithm reinforced with a metaheuristic hyperparameter-tuning strategy, namely the gravitational search algorithm (GSA), is adopted for the binary classification problem. Furthermore, this study incorporates the Shapley additive explanations (SHAP) ensuring transparent interpretations into the GSA-RF predictive framework to tackle the inherent black-box nature of the ML rationale. The results reveal that the proposed model detects the severity of NCRs in terms of their cost impact with an overall AUROC value of 0.776 for the preseparated and blinded testing set. This indicates that the proposed model can be used confidently for newly introduced datasets from real-life cases. In addition, the SHAP analysis results emphasized the role of season, inadequate application procedure, and NCR type in detecting the severity of NCRs. Overall, this research not only makes an important contribution through its novel data-driven approaches but also provides insights for project managers concerning productivity improvements in the sector.
... Different from statistical models, the input-output mapping in machine learning models does not need to be defined in advance. Instead, it is learned during the training process [13]. The support vector regression (SVR), which maps historical load data to a higher dimensional space through a nonlinear mapping and then performs linear regression on the mapped elements, is one of the most commonly used machine learning models for load forecasting [14]. ...
Load forecasting is critical to the task of energy management in power systems, for example, balancing supply and demand and minimizing energy transaction costs, etc. There are many approaches used for load forecasting such as the support vector regression, the autoregressive integrated moving average, and neural networks, but most of these methods focus on single-step load forecasting, whereas, multi-step load forecasting can provide better insights for optimizing the energy resource allocation and assisting the decision-making process. In this work, a novel sequence-to-sequence based deep learning model
based on a time series decomposition strategy for multi-step load forecasting is proposed. The model consists of a series of basic blocks, each of which includes one encoder and two decoders; and all basic blocks are connected by residuals. In the inner of
each basic block, the encoder is realized by temporal convolution network for its benefit of parallel computing, and the decoder is implemented by long short-term memory neural network to predict and estimate time series. During the forecasting process,
each basic block is forecasted individually. The final forecasted result is the aggregation of the predicted results in all basic blocks. Several cases within multiple real-world datasets are conducted to evaluate the performance of the proposed model. The results demonstrate that the proposed model achieves the
best accuracy compared with several benchmark models.
... In addition, Li [26] conducted experiments for PC forecasting in smart grid management where the used model was ANN. Further, Dietrich et al. [27] developed ANN for the PC of machine tools in a factory. Their study found that machine tools PC prediction can play an important role in power management and cost reduction of power in industries. ...
Energy management systems for residential and commercial buildings must use an appropriate and efficient model to predict energy consumption accurately. To deal with the challenges in power management, the short-term Power Consumption (PC) prediction for household appliances plays a vital role in improving domestic and commercial energy efficiency. Big data applications and analytics have shown that data-driven load forecasting approaches can forecast PC in commercial and residential sectors and recognize patterns of electric usage in complex conditions. However, traditional Machine Learning (ML) algorithms and their features engineering procedure emphasize the practice of inefficient and ineffective techniques resulting in poor generalization. Additionally, different appliances in a home behave contrarily under distinct circumstances, making PC forecasting more challenging. To address these challenges, in this paper a hybrid architecture using an unsupervised learning strategy is investigated. The architecture integrates a one-dimensional Convolutional Neural Network (CNN) based Autoencoder (AE) and online sequential Extreme Learning Machine (ELM) for commercial and residential short-term PC forecasting. First, the load data of different buildings are collected and cleaned from various abnormalities. A subsequent step involves AE for learning a compressed representation of spatial features and sending them to the online sequential ELM to learn nonlinear relations and forecast the final load. Finally, the proposed network is demonstrated to achieve State-of-the-Art (SOTA) error metrics based on two benchmark PC datasets for residential and commercial buildings. The Mean Square Error (MSE) values obtained by the proposed method are 0.0147 and 0.0121 for residential and commercial buildings datasets, respectively. The obtained results prove that our model is suitable for the PC prediction of different types of buildings.
... For this purpose, four MLP architectures were tested on hourly load and temperature data for North American and Slovakian electric utilities. In [8], a machine learning process based on artificial neural networks (ANN) is used to forecast the very short-term load of two machine tools. In terms of robustness, the work points toward the use of a combination of time series features and ANNs. ...
In the contemporary context, both production and consumption of energy, being concepts intertwined through a condition of synchronicity, are pivotal for the orderly functioning of society, with their management being a building block in maintaining regularity. Hence, the pursuit to develop reliable computational tools for modeling such serial and time-dependent phenomena becomes similarly crucial. This paper investigates the use of ensemble learners for medium-term forecasting of the Greek energy system load using additional information from injected energy production from various sources. Through an extensive experimental process, over 435 regression schemes and 64 different modifications of the feature inputs were tested over five different prediction time frames, creating comparative rankings regarding two case studies: one related to methods and the other to feature setups. Evaluations according to six widely used metrics indicate an aggregate but clear dominance of a specific efficient and low-cost ensemble layout. In particular, an ensemble method that incorporates the orthogonal matching pursuit together with the Huber regressor according to an averaged combinatorial scheme is proposed. Moreover, it is shown that the use of multivariate setups improves the derived predictions.
... With the development of data mining technology, many prediction technologies based on machine learning have been developed. Machine learning can adaptively learn nonlinear characteristics, , such as Support Vector Machine (SVM), Extreme Learning Machine (ELM), eXtreme Gradient Boosting (XGBoost) [1], and they can better capture the nonlinear characteristics of time series [2]. Deep learning models such as Recurrent neural network(RNN) and Long Short Term Memory (LSTM) have been successfully applied to time series prediction [3]. ...
Accuracy and rapidity are the primary objectives of load forecasting, and also the necessary conditions for ensuring power supply and production schedule. However, in actual production, due to the variability of operating modes and the interference of production environment, the difficulties such as non-stable and high fluctuation are in the power load prediction. In light of this, we propose an adaptive hybrid prediction model based on Discrete Wavelet decomposition(DWT). It is well known that DWT can show local features such as mutation and fluctuation in the sequence, and has excellent multi-scale analysis ability. Adopt energy entropy evaluate the aggregation of wavelet coefficients in order to obtain the subsequence with the highest degree of preserving the main frequency of the original signal. Random Forest (RF) and Long Short-Term Memory (LSTM) were used to track low-frequency profile and high-frequency detail fluctuations respectively. Further, parameters of the heterogeneou models are optimized using the Particle Swarm Optimization(PSO) to improve the adaptability of hybrid models to different frequency components. Compared with other excellent metallurgical load forecasting techniques, the effectiveness and superiority of the proposed model are verified by experiments on the actual industrial data set of an electrical smelting furnaces for magnesia.
... These three methods represent different modeling strategies typically used in energy modeling [31]. ...
The following paper examines the practicality of a methodical approach for energy-flexible and energy-optimal operation in the field of metal-cutting production. The analysis is based on the example of a grinding machine and its central cooling-supply system. In the first step, an energy-flexibility data model is built for each subsystem, which describes energy flexibility potentials generically. This is then extended to enable combined energy cost-optimal production planning. As a basis for the links between the data model representations, the cold flows between the subsystems are modeled using parameter-estimation methods, which have a mean absolute error of only 2.3 percent, making the subsequent installation of heat meters unnecessary. Based on the presented approach, the results successfully validate the possibility of energy-flexible cost-optimal and sensor-reduced production planning by reducing energy costs by 6.6 percent overall and 1.9 percent per workpiece produced.
... A VSTLF with a combination of single output using RF and multi-output model proposed [32] using machine learning technique. The author combines time-series features with ANN to enhance the load prediction accuracy. ...
Electricity demand is increasing at a rapid rate. Sustainability related challenges are posing an immediate cause of concern for the planet. Smart Grid provides an efficient way to manage the complex scenario. The challenge of enhancing energy efficiency and integration of renewables effectively is addressed by utilizing smart grid technology. Accurate demand forecasting and demand side management are needed to balance the demand and supply gaps over time of day and season related variations. This survey paper aims to provide a comprehensive overview of the various techniques and approaches used for demand forecasting in the residential, industrial, and agricultural sectors. It also examines the role of demand response in managing peak electricity demand and maintaining grid stability through the shifting of usage to off-peak hours. The aim is to understand the nuances and prevailing practices in these sectors to identify potential areas for improvement.
... The prediction of various parameters such as short-term load profile, peak load, contingency during transients, etc. using the U/VSTLF approach. 10 2 ...
The rapid growth of hybrid renewable Distributed Energy Resources (DERs) generation possess various challenges with inaccurate forecast models in stochastic power systems. The prime objective of this research is to maximum utilization of scheduled power from hybrid renewable based DERs to maintain the load-demand profile with reduce distributed grid burden. The proposed mixed input-based cascaded artificial neural network [Formula: see text] is realized for the prediction of a short-term based hourly solar irradiance and wind speed. The testing approach is performed through a historical hourly dataset of the proposed site. Further, the normalized data sets are divided into hourly-based samples for validating the load demand power with respect to the variation in metrological data. In this paper, Adaptive Neuro-Fuzzy Inference System (ANFIS) model is simulated for short-term power demand prediction. This adaptive methodology is an effective approach for load-demand management which is based on cross-entropy. It also confirmed that during testing, the forecasting mean error and cross-entropy are less than 5% under a specific time slap of an individual day. The regression analysis is performed through the time series fitting simulation tool at different time horizons. The performance evaluation of the designed model is compared with the multi-layer perceptron model. Simulation results display the proposed mixed input-based cascaded system has enhanced accuracy and optimal performance than the multi-output correlated perceptron model.
... However, with the changing policies and continuous economic development, the traditional forecasting methods are no longer able to accurately predict the trend of energy consumption. Machine learning has gradually made its mark in the field of energy prediction [9][10][11]. Wang et al. predicted the building cooling and heating loads based on an improved BP neural network, and the results were shown [12]. ...
In order to accurately predict China’s future total energy consumption, this article constructs a random forest (RF)–sparrow search algorithm (SSA)–support vector regression machine (SVR)–kernel density estimation (KDE) model to forecast China’s future energy consumption in 2022–2030. It is explored whether China can reach the relevant target in 2030. This article begins by using a random forest model to screen for influences to be used as the input set for the model. Then, the sparrow search algorithm is applied to optimize the SVR to overcome the drawback of difficult parameter setting of SVR. Finally, the model SSA-SVR is applied to forecast the future total energy consumption in China. Then, interval forecasting was performed using kernel density estimation, which enhanced the predictive significance of the model. By comparing the prediction results and error values with those of RF-PSO-SVR, RF-SVR and RF-BP, it is demonstrated that the combined model proposed in the paper is more accurate. This will have even better accuracy for future predictions.
... Flexibility potential for DR in the industry is evaluated in Lee et al. (2020). Load forecasting of industrial machine tools is discussed in Dietrich et al. (2020). General estimation of flexibility potential is explored based on long-term historical data (D'hulst et al., 2015;Dyson et al., 2014;Bustos-Turu et al., 2015), or by surveys on customer readiness to participate in DR programs (Yamaguchi et al., 2020;Vellei et al., 2020). ...
Existing grids have been designed with traditional large centralized generation in mind; however, with the ever-increasing utilization of renewable distributed energy resources, the challenges of proper grid management have intensified. Demand-side energy flexibility is seen as one potential way to alleviate these challenges. Presently, residential demand-side energy flexibility has remained a largely untapped resource since individual prosumers are too small to provide enough capacity, thus necessitating the need for an aggregator. In view of the aforementioned, this paper conducts a literature review on the aggregated residential demand-side energy flexibility. The paper gives an overview of characterization methods of energy flexibility. The sources of residential energy flexibility are identified and categorized based on their flexibility characteristics. In addition, the quantification methods and parameters of energy flexibility are analyzed. Moreover, the forecasting methods of energy flexibility in the context of different flexibility sources are outlined. Additionally, an overview of existing markets and potential new emerging flexibility markets is given. The challenges and barriers faced by the aggregators attempting to enter flexibility markets are examined. Finally, the paper is concluded by providing a discussion of the key findings that summarize the current research directions and highlight the gaps for future development of aggregated energy flexibility.
... where x N is the standardized variable, x is the raw data, min(x) is the minimum value of a variable of the same type, max(x) is the maximum value of a variable of the same type. while moving average smoothens the curve and therefore reduces noise in the features [11]. ...
Accurate load forecasting can efficiently reduce the day-ahead dispatch stress of power system or microgrid. The overview of load forecasting based on artificial intelligence models are comprehensively summarized in this paper. As the steps of load forecasting based on artificial intelligence model mainly include data processing, setting up forecasting strategy and model forecasting, the paper firstly reviewed the data processing studies. According to the kinds of data obtained, the data can be classified into two categories: multivariate time series and single variate time series. Secondly the forecasting methodologies including one-step forecasting and rolling forecasting are summarized and compared. In addition, according to the form of the prediction results, point prediction, interval prediction and probability prediction are summarized. Thirdly, the paper reviews the artificial intelligence models used in load forecasting. In light of the application scenarios, it can be classified into single model and combination model. Finally, we also discussed the future trend for the research, such as fuzzy reasoning, intelligent optimization in forecasting, novel machine learning and transfer learning technologies, etc.
... Such data-driven methods, including ML techniques, can be implemented for long-term energy and performance predictions (Amasyali and El-Gohary 2018). In the literature, ML methods have been implemented for the prediction of different building energy conservation indices such as energy demand in existing buildings (Amasyali and El-Gohary 2018), setpoint management (Brandi et al. 2020), HVAC system optimization or fault diagnosis (Han et al. 2020), and peak load prediction (Dietrich et al. 2020). ...
Recent studies have focused on data-driven methods for building energy efficiency, by using simulated or empirical data, for energy-based design assessment rather than the common physics-based techniques, which are mostly time-consuming. In this paper, the feasibility of using seven different Machine Learning models, including three single models and four ensemble ones, is studied to predict annual energy demand and thermal comfort of the model. For this purpose, 3024 synthetic samples of a single zone model with seven input features are simulated through the EnergyPlus engine for training in addition to 360 unseen samples as testing data for accuracy reporting. Heating and cooling demands, in addition to five annual thermal comfort indices, are calculated for each data point and used as target indices. Results show Extremely Randomized Trees and Random Forest models had the highest R2 of 0.99 and 0.85 for cooling and heating demands respectively. Also, the R2 of these models for predicting annual comfort was between 0.71 and 0.95. Results are then used to develop a prediction framework of thermal comfort and energy demand performance in the early stages of building design, where most of the information about building characteristics is not yet known.
... [41] Energy simulations are needed for some of the use cases already mentioned above, but due to its importance are mentioned as separate use case as well. Simulation models are used to analyse production environments regarding their energy efficiency, predict energy consumption [42] and also to evaluate identified improvement potentials through fast scenario comparisons as exemplarily described in [43]. To enable a holistic view with regard to the product life cycle and to understand the interaction between the individual product states, the last use case addresses an external learning factory. ...
... Currently, the power system load forecasting can be divided into four types: ultra-short-term [7], short-term [8], medium-term [9], and long-term [10]. Short-term load forecasting has been a research hotspot, and multi-feature load forecasting methods are commonly used. ...
Affected by the new coronavirus (COVID-19) pandemic, global energy production and consumption have changed a lot. It is unknown whether conventional short-term load forecasting methods based on single-task, single-region, and conventional indicators can accurately capture the load pattern during the COVID-19 and should be carefully studied. In this paper, we make the following contributions: 1) A mobility-optimized load forecasting method based on multi-task learning and long short-term memory network is innovatively proposed to alleviate the impact of the COVID-19 on short-term load forecasting. The incorporation of mobility data and data sharing layers potentially reduces the difficulty of capturing the load patterns and improves the generalization of the load forecasting models. 2) The real public data collected from multiple agencies and companies in the United States and European countries are used to conduct horizontal and vertical tests. These tests prove the failure of the conventional models and methods in the COVID-19 and demonstrate the high accuracy (error mostly less than 1%) and robustness of the proposed model. 3) The Shapley additive explanations technology based on game theory is innovatively introduced to improve the objectivity of the models. It visualizes that mobility indicators are of great help to the accurate load forecasting. Besides, the non-synchronous relationships between the indicators’ correlations and contributions to the load have been proved.
... We remark that further improvements might be obtained by including more specific conditioning variable within the models. However, as the aim of this work is to compare PLF techniques under consistent conditions, we leave such investigation to future extensions, e.g., by exploiting dedicated feature selection techniques within the forecasting framework (see e.g., [81,84]). Considering the experimental setup adopted in [32], the available samples have been split into training, validation and test subsets by a 70%/15%/15% decomposition. ...
This work presents a novel approach to address a challenging and still unsolved problem of neural network based load forecasting systems, that despite the significant results reached in terms of prediction error reduction, still lack suitable indications regarding sample-wise trustworthiness of their predictions. The present approach is framed on Bayesian Mixture Density Networks, enhancing the mapping capabilities of neural networks by integrated predictive distributions, and encompassing both aleatoric and epistemic uncertainty sources. An end-to-end training method is developed, aimed to discover the latent functional relation to conditioning variables, characterize the inherent load stochasticity, and convey parameters uncertainty in a unique framework. To achieve reliable and computationally scalable estimators, both Mean Field variational inference and deep ensembles are integrated. Experiments have been performed on short-term load forecasting tasks at both regional and fine-grained household scale, to investigate heterogeneous operating conditions. Different architectural configurations are compared, showing by Continuous Ranked Probability Score based tests that significant performance improvements are achieved by integrating flexible aleatoric uncertainty patterns and multi-modalities in the parameters posterior space.
... Fungsi ini digunakan untuk menghitung hubungan antara sebuah seri dan pergeseran waktunya. Fungsi ini juga digunakan untuk mengidentifikasi musiman dan tren [13]. Informasi ini selanjutnya akan digunakan untuk mengidentifikasi jeda waktu dan membatasi ukuran waktu yang digunakan sebagai input pada model prediksi inflasi. ...
Bank Indonesia mendefinisikan inflasi merupakan meningkatkan harga-harga secara umum dan terus-menerus. Kenaikan harga barang dan jasa dapat disebut inflasi apabila kenaikan tersebut meluas atau mempengaruhi kenaikan harga lainnya. Naiknya harga barang dan jasa tersebut dapat menyebabkan turunnya nilai uang. Dengan ini, inflasi dapat menurunkan nilai uang terhadap nilai barang dan jasa secara umum. Jika inflasi yang terjadi dapat dikendalikan dengan baik, tingkat inflasi tersebut dapat memberikan dampak positif terhadap pertumbuhan ekonomi. Tujuan dari penelitian ini yaitu dapat memprediksi tingkat inflasi agar inflasi dapat dikontrol tiap bulannya dan dapat meberikan dampak yang positif. Penelitian ini menggunakan metode jaringan syaraf tiruan yang sesuai digunakan pada data time series dengan data training. Data yang digunakan adalah data inflasi bulanan kelompok pengeluaran dari bulan Desember 2011 sampai Desember Januari 2020 diambil dari Badan Pusat Statistik. Penelitian ini diharapkan dapat membantu untuk memutuskan tindakan yang tepat berdasarkan hasil prediksi. Pengujian menggunakan beberapa model diperoleh hasil terbaik dari model dengan konfigurasi 7-15-1 dengan learning rate 0,01 yang menghasilkan MSE sebesar 0,026. Hasil ini menunjukkan bahwa jaringan syaraf tiruan dapat digunakan untuk prediksi inflasi dengan akurasi yang tinggi.
The national holiday policy has a significant impact on the holiday load, and the curve's shape differs from the regular daily load, making it challenging to directly anticipate the overall load. The categorization predicting approach presented in this research is based on the division of regular days and holidays. First, the load is divided depending on the date variable after an analysis of the characteristics of an ordinary day and a holiday load. The combined model based on similar day selection and generalized regression network is then utilized to forecast in accordance with the usual daily load. In order to predict the holiday load, a fusion model based on LightGBM and XGBoost is used. The experimental results demonstrate, using the data set provided by a power supply bureau in southern China as a practical example, that the classification forecasting method put out in this study increases the precision of global load forecasting.
This work introduces a Machine Learning (ML) model designed to predict solar radiation in diverse cities representing Colombia's climatic variability. It is crucial to assert that the amount of solar energy received in a specific region is directly related to solar radiation and its availability, which is influenced by each area's particular climatic and geographic conditions. Due to the high variability and resulting uncertainty, various approaches have been explored, including the use of numerical models to estimate solar radiation. The primary objective of this study was to develop and validate an ML model that accurately predicts solar radiation in cities. The methodology employed was specific to data treatment and ML model development. It was structured into three fundamental stages: clustering, estimation, and response, considering that the model is based on historical data. The obtained results were assessed using appropriate statistical definitions, not only determining the model's efficiency in terms of prediction but also considering interactions between data for the approximation and prediction of solar radiation. In this context, it is crucial to emphasize that the research contributes to understanding solar radiation in Colombia. This study underscores the importance of developing ML models to predict solar radiation, emphasizing the need to consider the country's climatic diversity. The results obtained, following the model's application, provide valuable information for comprehending and anticipating the availability of this primary resource.
Short-term load forecasting remains pivotal in managing sustainable energy grids, with accuracy directly influencing operational decisions. Conventional forecasting methodologies often falter in adapting to the dynamic complexities inherent in modern energy systems. This paper introduces a predictive intelligence technique rooted in machine learning aimed at enhancing short-term load forecasting accuracy within sustainable energy grids. Leveraging historical data, weather patterns, grid operations, and consumer behavior insights, our study develops a robust predictive model. The model's adaptability to evolving patterns and real-time data integration offers a promising solution to the limitations of existing forecasting methods. Through a comparative analysis and validation against established benchmarks, the proposed technique showcases superior performance, demonstrating its potential for more efficient resource allocation and improved grid management. This research contributes to advancing sustainable energy practices by offering a reliable and adaptive solution for short-term load forecasting, fostering more resilient and responsive energy grid operations.
Short-term utility grid planning and operation are based on projected future energy consumption and transmission/generation capacity. The industrial sector being a major influence on the total electrical power consumption. Despite its importance, industrial load forecasting (LF) is not a widely discussed topic in literature, despite the fact that it is influenced by a number of factors, such as planned operations and work shifts, that are unusual or unnecessary in conventional load forecasting models. Unlike time-series analysis, machine learning (ML) techniques are well-suited to the nature of the loads since they can simulate complex nonlinear relationships through a process that uses previous data patterns. In this paper, four data-driven machine learning strategies including Support Vector Machine (SVM), Decision Tree (DT), Artificial Neural Network (ANN), and Gaussian Process (GP), are compared to forecast an industrial load. These strategies can help in selecting the optimal model for load prediction. The results show that GP outperformed other ML techniques, thus yielding more accurate and reliable readings with lower estimation errors for electrical LF, but at the expense of higher computational time.
This paper proposes a boosted multi-task learning framework for inter-district collaborative load forecasting. The proposed framework involves two subsequent stages: in the first stage, districts would collaborate under a seamlessly-integrated federated learning scheme to capture the global load pattern; in the second stage, districts would withdraw and perform local training to capture the local load patterns. The probabilistic Gradient-Boosted Regression Tree (GBRT) is applied as the bottom-level machine learning algorithm, which would allow for an easy and intuitive embodiment of the generalized multi-task learning framework. We further propose two candidate district withdrawal mechanisms to connect the two stages: the simultaneous withdrawal, which prioritizes prediction accuracy, and the dynamic withdrawal, which prioritizes training efficiency and district incentivization. The follow-up performance analyses and the case study on 11 districts of the Zhuhai city confirm the superiority of the proposed framework and district withdrawal mechanisms.
This special issue on “digital-based production” gives an overview about the current research on the integration of digital technologies into production processes and their fields of application. It covers topics from Industry 4.0, artificial intelligence and data analytics to the Industrial Internet of Things and Cyber-Physical Production Systems. This issue offers valuable insights for those interested in improving production efficiency, quality, and sustainability through digital technologies. In this foreword, we describe promising application fields of digital-based production and classify the submitted articles accordingly.
Machine learning (ML) can be a valuable tool for discovering opportunities to save energy and resources in manufacturing systems. However, the hype around ML in the context of Industry 4.0 in the past few years has led to blind usage of the approach, occasionally resulting in usage when another analysis approach would be better suited. The research presented here uses a novel matrix approach to address this lack of differentiation of when to best use ML for improving energy and resource efficiency in manufacturing, by systematically identifying situations in which ML is well suited. Seventeen generic levers for improving manufacturing energy and resource efficiency are defined. Next, a generic list of six manufacturing data scenarios for when ML is a good method of choice for analysis is created. This results in a comprehensive matrix in which each lever is evaluated along each ML scenario and given a point, providing a quantitative ML suitability score for each lever. The evaluation is conducted by drawing on past studies demonstrating whether ML is appropriate. Specifically, operation parameter and input material optimization, as well as intelligent maintenance, are the levers that score the highest and are thus identified to be most suitable for machine learning. The majority of the remaining levers is deemed to have low suitability for machine learning. This simple yet informative matrix can be used as a guideline in data-driven manufacturing energy and resource efficiency projects to provide an appraisal on the applicability of ML as the initial analysis tool of choice.
With users’ increasing knowledge and intellectualization, users’ abnormal electricity consumption behaviors (AECB) are becoming more prevalent. Since access to renewable energy sources leads to a volatile and intermittent electricity load, the existing artificial intelligence methods are challenging to detect the AECB of users. To quickly and accurately identify the AECB of a massive number of users, this work proposes the GoogLeResNet3 network module, which contains fully connected layers, the Inception module, and a residual module. The GoogLeResNet3 network is compared with the GoogLeNet module, ResNet-50, ResNet-101, and 11 other neural networks. The results of the comparison experiments indicate that: the GoogLeResNet3 network with the highest accuracy is 335 s quicker than the second-fast network, and the accuracy is 10.57 % higher than the second-best network at least.
Accurate load forecasting is conducive to stable power supply. It is difficult to forecast electrometallurgical load. It tends to fluctuate rapidly and randomly, especially for the ultra-short-term ones. In this work, we decompose the load into components and abstracting them with heterogeneous learners for the different characteristics. Complete ensemble empirical model decomposition adaptive noise (CEEMDAN) is introduced to data pre-treatment and signal decomposition, and different frequency sub-series are obtained. Then, we explore the complexity of sub-series by sample entropy (SE). For the low-frequency components, a low-complexity Random Forest (RF) model is designed for drawing the outline of the load fluctuating. For the high-frequency ones, a long short-term memory (LSTM) network is adopted to get the details of the fluctuating. Combined forecasting is implemented by reconstructing the predictions of the above two learners. The novel model is compared with several common models for electrometallurgical load forecasting by using industrial data of an electro-fused magnesium enterprise. Experimental results show that the proposal significantly outperforms the others.
In recent years, digitalisation has rendered machine learning a key tool for improving processes in several sectors, as in the case of electrical power systems. Machine learning algorithms are data-driven models based on statistical learning theory and employed as a tool to exploit the data generated by the power system and its users. Energy communities are emerging as novel organisations for consumers and prosumers in the distribution grid. These communities may operate differently depending on their objectives and the potential service the community wants to offer to the distribution system operator. This paper presents the conceptualisation of a local energy community on the basis of a review of 25 energy community projects. Furthermore, an extensive literature review of machine learning algorithms for local energy community applications was conducted, and these algorithms were categorised according to forecasting, storage optimisation, energy management systems, power stability and quality, security, and energy transactions. The main algorithms reported in the literature were analysed and classified as supervised, unsupervised, and reinforcement learning algorithms. The findings demonstrate the manner in which supervised learning can provide accurate models for forecasting tasks. Similarly, reinforcement learning presents interesting capabilities in terms of control-related applications.
With the rapid growth in the volume of relevant and available data, feature engineering is emerging as a popular research subject in data-driven building energy prediction owing to its essential role in improving data quality. Many studies have examined the feasibility of applying feature engineering methods to data-driven building energy prediction. However, a systematic review of this area's research status, characteristics, and limitations is lacking. Therefore, this study analyzes the current status of research and directions of future work in feature engineering for building energy prediction. In this article, we first discuss the concept of feature engineering and its main methods, including the construction, selection, and extraction of features. We, then, summarize the status and characteristics of feature engineering research in the building energy domain using a comprehensive study of 172 relevant articles. We also discuss critical issues in feature engineering in data-driven building energy prediction, including why feature engineering has recently received increasing attention, whether it is useful in this domain, and effective ways to apply it. Finally, we identify promising research directions in the area based on its current state and limitations. The results here provide researchers and the industry with a better understanding of the state of the art and future research trends in feature engineering for data-driven building energy prediction.
The benefits of energy, economics, and greenhouse gas emission reduction have interested consumers towards photovoltaic system-based energy generation. The fluctuating nature of solar energy is the biggest challenge for energy predictions. This chapter is an attempt to provide a forecasting tool for energy planners and energy economists. This chapter focuses on three forecasting variables, solar irradiance, solar energy, and load. A 1 MW solar plant with corresponding load was considered in this analysis, specifically temperature, wind speed, humidity, solar flux, energy generation per hour, and its history were the inputs for the energy prediction model and historical data of hourly load consumption were considered for load prediction. The Recurrent Artificial Neural Network-based algorithm was used for reliable short-term forecasting. The output of this model is reported in root mean square error and mean absolute percentage error form. The results show better agreement between actual and forecasted samples. The prediction of solar energy and load helps to fulfill economic benefits.
Microgrids (MGs) are considered an important part of the evolving power grid in terms of reliability, cost, and environmental impact of distributed generation systems. The main production resource in MGs is renewable energy (RE) sources which have grown even more recently. The energy management system in MGs has many features such as decreasing the power losses in transmission lines and flexibility of monitoring the distributed energy resources (DERs). However, the power output from the DERs may get badly affected due to the intermittent nature of RE connected with the MGs system. Therefore accurate short-term solar generation forecasting is an important issue in MG to predict the required amount of power to be dispatched by DERs, which can lead to economic advantages for end-users. In this chapter, a comparative study of different machine learning (ML) approaches has been applied for forecasting solar radiation and temperature. In the proposed method, the most applied models in ML such as linear regression, random forest (RF) regression, K-nearest neighbors, support vector machines (SVMs) models are compared. We assessed the performance of ML proposed by evaluation of the root mean square error, and then investigated the influences of the parameter’s techniques. The simulation result shows that RF and SVM perform well in the short-term forecasting of solar irradiation and temperature.
Accurate electric power load forecasting plays an important role in supporting the power system reliability, promoting the distributed renewable energy integration, and developing effective Demand Response strategies. In this chapter, a deep learning model for the load forecasting with a one-hour resolution of residential buildings is presented. Both model complexity and variability are considered. Hourly-measured residential load data in Austin, Texas, USA are used to demonstrate the effectiveness of the presented model, and the forecasting error was quantitatively evaluated using several metrics. The results showed that the presented model forecasts the aggregated and disaggregated load of residential buildings with higher accuracy compared to conventional methods. Furthermore, the presented deep learning model is also an effective method for filling missing data through learning from history data. It reveals that the presented model has a good learning ability that can accommodate time dependencies to achieve high forecasting accuracy with limited input variables.
The use of artificial neural network models to enrich the analytical and predictive capabilities of decision support systems in manufacturing has increased. The growing complexity and uncertainty in the manufacturing sector demand improved decision-making to ensure low operations costs, high productivity, and sustainable use of resources. Artificial neural networks have the inherent capacity to analyze the most uncertain and complex patterns in unstructured decision problems. This review aims to synthesize and provide a comprehensive summary of recent studies on artificial neural network-based decision support systems as applied in manufacturing processes. First, the specific processes in manufacturing where artificial neural network-based decision support systems are used are analyzed. A total of 99 multi-disciplinary publications on artificial neural network-based decision support systems published between 2011 and 2021 are retrieved and processed following a rigorous execution of the designated acceptance criteria and quality assessment. A review of the selected studies indicates a growing interest in applying artificial neural networks in decision support systems. Product and process design, performance evaluation, and predictive maintenance are the main application areas identified. A growing tendency to combine artificial neural network models with other intelligent tools, notably fuzzy logic, and genetic algorithm, is noted to overcome drawbacks such as slow convergence when training the algorithms. Further research should extend to other tools for enriching the performance of artificial neural networks in manufacturing processes.
Due to the high cost of electricity in commercial and industrial sectors, demand forecast models have gained increasing attention. However, there are two unresolved issues: (1) Models are not adaptable when exposed to previously unknown data (2) The value of regression methods vs. state-of-the-art machine learning models has not been made apparent before. This study’s goal is to develop probabilistic demand estimation models. We propose a probabilistic Bayesian regression framework that can not only estimate future demands with high accuracy but also be updated once new information is available. By applying the proposed algorithm to two real-world case studies (commercial and manufacturing), we show a 40.3% and 30.8% improvement in terms of mean absolute error for the two cases. Moreover, the proposed technique outperforms powerful machine learning approaches, including support vector machine by 10.39%, random forest by 6.17%, and multilayer perceptron by 9.14% in terms of mean absolute percentage error.
The introduction of renewable resources into the distributed energy system has challenged the operation optimization of the distributed energy system. Integration of new technologies and diversified characteristics on the demand side has exerted a great influence on the distributed energy system. In this paper, by way of literature review, first, the topological structure and the mathematical expression of the distributed energy system were summarized, and the trend of enrichment and diversification and the new characteristics of the system were evaluated. Then, the load forecasting technology was reviewed and analyzed from two aspects, fundamental research and application research. Research methods of the distributed energy system under the new trend of energies were discussed, and the boundaries of the broadened distributed energy technology were explored.
The present work proposes a holonic-based mechanism for self-learning factories based on a hybrid learning approach. The self-learning factory is a manufacturing system that gains predictive capability by machine self-learning, and thus automatically anticipates the performance results during the process planning phase through learning from past experience. The system mechanism, including a modeling method, architecture, and operational procedure, is structured to agentize machines and manufacturing objects under the paradigm of Holonic Manufacturing Systems. This mechanism allows machines and manufacturing objects to acquire their data and model interconnection and to perform model-driven autonomous and collaborative behaviors. The hybrid learning approach is designed to obtain predictive modeling ability in both data-existent and even data-absent environments via accommodating machine learning (which extracts knowledge from data) and transfer learning (which extracts knowledge from existing knowledge). The present work also implements a prototype system to demonstrate automatic predictive modeling and autonomous process planning for energy reduction in milling processes. The prototype generates energy-predictive models via hybrid learning and seeks the minimum energy-using machine tool through the contract net protocol combined with energy prediction. As a result, the prototype could achieve a reduction of 9.70% with respect to energy consumption as compared with the maximum energy-using machine tool.
In the context of energy transition in Germany, precise load forecasting enables reducing the impact of increased volatility in power generation induced by renewable energies. This paper presents a machine learning approach to generate a 15 minutes forecasting model of the electric load for the ETA research factory at TU Darmstadt on a factory level. In the first iteration, a feature selection process was conducted to select significant features for machine learning datasets. The raw data contained 1,554 features from machine tools, technical building equipment, the building itself and external factors like the weather. The second iteration examined the forecasting capabilities of six hyperparameter tuned algorithms on the feature selected datasets. In the third iteration, feature engineering and hyperparameter tuning led to an optimized Gradient Boosting Regression Trees (GBRT) algorithm. The results indicate that the utilized machine learning approach is feasible and creates a precise very short term load forecasting model, depending on the use case of the load forecast.
A thermohydraulic linkage of machines, technical building equipment and the building itself is proven to improve energy efficiency by recovering waste heat and transforming it into useful energy. This approach increases the potential for savings, but inevitably leads to an increased complexity of the interacting sub-systems and automation solutions to control the energy production, storage and consumption. This paper presents a framework for an energy efficient and flexible operating strategy of supply systems within a thermally-linked factory. Additionally, an approach is presented to implement an interface for external control optimizers while ensuring the safe operation of supply grids and facilities.
The metering and analysis of energy demands is widely applied in today's manufacturing industry in order to reduce energy costs and environmental impacts alike. However, especially on machine level the interpretation of energy data is still challenging due to huge amounts of metering data and a lack of methodological knowledge. The presented approach focuses on the evaluation of electric load profiles on machine and workgroup level in dependency on available complementing data such as product, scheduling or machine data. The approach aims at extracting a maximum degree of information from the available data in order to improve machine operation modes, production scheduling, energy cost allocation and factory planning processes. The approach is exemplified by use cases from the automotive industry.
Energy prediction of machine tools can deliver many advantages to a manufacturing enterprise, ranging from energy-efficient process planning to machine tool monitoring. Physics-based energy prediction models have been proposed in the past to understand the energy usage pattern of a machine tool. However, uncertainties in both the machine and the operating environment make it difficult to predict the energy consumption of the target machine reliably. Taking advantage of the opportunity to collect extensive, contextual, energy-consumption data, we discuss a data-driven approach to develop an energy prediction model of a machine tool in this paper. First, we present a methodology that can efficiently and effectively collect and process data extracted from a machine tool and its sensors. We then present a data-driven model that can be used to predict the energy consumption of the machine tool for machining a generic part. Specifically, we use Gaussian process (GP) regression, a nonparametric machine-learning technique, to develop the prediction model. The energy prediction model is then generalized over multiple process parameters and operations. Finally, we apply this generalized model with a method to assess uncertainty intervals to predict the energy consumed by any part of the machine using a Mori Seiki NVD1500 machine tool. Furthermore, the same model can be used during process planning to optimize the energy-efficiency of a machining process.
The energy efficiency is important evaluation criterion for new investment in machinery and equipment in addition to the classical parameters accuracy, performance, cost and reliability. Even the users in the automotive industry demand new acquisitions of energy consumed by a machine tool during machining. Large interrelated parameters that influence the energy consumption of a machine tool make the development of an appropriate predictive model a very difficult task. In this paper, a real machining experiment is referred to investigate the capability of artificial neural network model for predicting the value of energy consumption. Results indicate that the model proposed in the research is capable of predicting the energy consumption. The present scenario demands such type of models so that the acceptability of prediction models can be raised and can be applied in sustainable process planning during the manufacturing phase of life cycle of a machine tool.
The objective of this work is to highlight the modeling capabilities of artificial intelligence techniques for predicting the power requirements in machining process. The present scenario demands such types of models so that the acceptability of power prediction models can be raised and can be applied in sustainable process planning. This paper presents two artificial intelligence modeling techniques - artificial neural network and support vector regression - used for predicting the power consumed in machining process. In order to investigate the capability of these techniques for predicting the value of power, a real machining experiment is performed. Experiments are designed using Taguchi method so that effect of all the parameters could be studied with minimum possible number of experiments. A L16 (43) 4-level 3-factor Taguchi design is used to elaborate the plan of experiments. The power predicted by both techniques are compared and evaluated against each other and it has been found that ANN slightly performs better as compare to SVR. To check the goodness of models, some representative hypothesis tests t-test to test the means, f-test and Leven's test to test variance are conducted. Results indicate that the models proposed in the research are suitable for predicting the power.
This study presented an empirical study to model the cost of the energy for high speed hard turning. A set of experimental machining data to cut hard AISI 4340 steel was obtained with a different range of cutting speed, feed rate and depth of cut with negative rake angle. Regression models were developed by using Box-Behnken Design (BBD) as one of Respond Surface Methodology (RSM) collections. Neural network technique was deployed using MATLAB to predict the energy as a part of the artificial intelligent methods. The data collected was statistically analyzed using Analysis of Variance (ANOVA) technique. Second order energy prediction models were developed by using (RSM) then the measured data were used to train the neural network models. A comparison of neural network models with regression models is also carried out. Predictive Box-Behnken models are found to be capable of better predictions for energy within the range of the design boundary.
Specifically for companies which involve high energy consuming production processes and need defined production environments as well as process-related auxiliary media optimizing energy efficiency gets more and more important nowadays. Thereby it is crucial to have a holistic view on the whole system including the strong interdependencies of production equipment and technical building services to derive global optima. Against this background this paper presents an integrated approach which basically enfolds the coupling of four different simulation tools for technical building services, building climate, production machines / material flow and production management. Thereby the approach supports the energy efficient design and management of production facilities and technical building services.
Soaring energy prices, signs of a human-made climate change and legislative pressure have brought the energy consumption of machines to the attention of machine tool builders and their customers. To enable manufacturers and operators of machines to include energy consumption into their considerations in an objective way, we introduce an efficient method to model the energy consumption behaviour of machines. Based on that, we present methods to forecast the actual power drain profile and to optimise machines for minimal energy consumption under any given application scenario 1. Motivation For decades, economies worldwide and most of all in Europe ha-ve been able to grow fast based on the almost unlimited availabi-lity of cheap energy and resources. First signs of the dependency on energy for prosperity have become apparent in the 1970s, with a number of economic setbacks resulting from energy shortages. In that decade, the acclaimed book 'The Limits to Growth' [Meadows 1972] highlighted the global consequen ces of limited resources for the first time. Since then, many new and large economies have en-tered the com petition for resources with force, while new sources of energy have not been made available to the same extent. As a result, the energy prices have experienced a sharp increase and brought the cost for energy onto the agenda. Moreover, as most means to produce energy result in the emission of carbon dioxide to the atmosphere, human induced climate change is directly lin-ked to the energy consumption worldwide and is expected to lead to additional cost in the future. In addition to raising the consciousness of people worldwide, this has led to legislative pressure. A recent Directive of the European Parliament on Energy using Products [Directive 2005/32/EC] aims to establish a framework for the setting of ecodesign requirements for energy-using products. The final report [Working Plan of the EcoDesign Directive 2008] of a study group initiated by the Europe-an Commission cites tool machines as the top three priority for inclu-sion into the product categories to be regulated in this framework. Although the results of this study have been at the heart of some de-bate [Welcker 2008], it is highly probable that tool machines will be subjected to energy efficiency regulation and classification in the years to come. Motivated by the mentioned financial and legislative pressure, en-terprises that buy and operate tool ma chines are currently investigat-ing methods to limit the life cycle cost of their production lines. The automotive industry will require energy consumption data to be in-cluded into bids in the upcoming years. Tool machine manufactur-ers will therefore have to be able to provide accurate data on the energy their products require to manufacture a product. Unfortu-nately, today neither the tool machine manufacturers nor their cus-tomers have a clear picture of the energy use of machines and pro-duction lines. The ability to make informed decisions about design aspects that have an influence on both the energy consumption and the invest-ment will be an important aspect of competitiveness in the future. While some measurements have been made, what has been missing is a practical method to predict the actual energy consumption. This paper therefore introduces a lean modelling technique that helps to overcome these limitations.
In this work we present a large scale comparison study for the major machine learning models for time series forecasting. Specifically, we apply the models on the monthly M3 time series competition data (around a thousand time series). There have been very few, if any, large scale comparison studies for machine learning models for the regression or the time series forecasting problems, so we hope this study would fill this gap. The models considered are multilayer perceptron, Bayesian neural networks, radial basis functions, generalized regression neural networks (also called kernel regression), K-nearest neighbor regression, CART regression trees, support vector regression, and Gaussian processes. The study reveals significant differences between the different methods. The best two methods turned out to be the multilayer perceptron and the Gaussian process regression. In addition to model comparisons, we have tested different preprocessing methods and have shown that they have different impacts on the performance.
The Variance Inflation Factor (VIF) and tolerance are both widely used measures of the degree of multi-collinearity of the
ith independent variable with the other independent variables in a regression model. Unfortunately, several rules of thumb
– most commonly the rule of 10 – associated with VIF are regarded by many practitioners as a sign of severe or serious multi-collinearity
(this rule appears in both scholarly articles and advanced statistical textbooks). When VIF reaches these threshold values
researchers often attempt to reduce the collinearity by eliminating one or more variables from their analysis; using Ridge
Regression to analyze their data; or combining two or more independent variables into a single index. These techniques for
curing problems associated with multi-collinearity can create problems more serious than those they solve. Because of this,
we examine these rules of thumb and find that threshold values of the VIF (and tolerance) need to be evaluated in the context
of several other factors that influence the variance of regression coefficients. Values of the VIF of 10, 20, 40, or even
higher do not, by themselves, discount the results of regression analyses, call for the elimination of one or more independent
variables from the analysis, suggest the use of ridge regression, or require combining of independent variable into a single
index.
Scikit-learn is a Python module integrating a wide range of state-of-the-art
machine learning algorithms for medium-scale supervised and unsupervised
problems. This package focuses on bringing machine learning to non-specialists
using a general-purpose high-level language. Emphasis is put on ease of use,
performance, documentation, and API consistency. It has minimal dependencies
and is distributed under the simplified BSD license, encouraging its use in
both academic and commercial settings. Source code, binaries, and documentation
can be downloaded from http://scikit-learn.sourceforge.net.
Variable renewable energies (VRE), in particular wind and solar PV, constitute a key option to reduce global greenhouse gas emissions. Future policy scenarios therefore propose a dominant role for VRE. However, relying almost entirely on the stochastic weather-determined output of VRE will require a transformation of the way power systems are planned and operated: a growing amount of flexibility will be needed to match variable demand with increasingly variable supply.
Due to the complexity of power systems as well as their long investment cycles, it is crucial to prepare the strategic development of flexibility now. The key question for the transition to energy systems based on variable renewables becomes: “How can we ensure that future power systems have the flexibility needed to match demand and variable supply?” Power system operators and regulators need to assess the current flexibility level in their system, analyze all possible flexibility options, and clearly prioritize the needed actions.
This paper presents the Flexibility Tracker, an assessment methodology developed to monitor and compare the readiness of power systems for high VRE shares. The Flexibility Tracker builds 14 flexibility assessment domains, by screening systems across the possible flexibility sources (supply, demand, energy storage) and enablers (grid, markets), via 80 standardised Key Performance Indicators (KPIs) scanning the potential, deployment, research activities, policies and barriers regarding flexibility.
The methodology allows monitoring the progress made in individual power systems with respect to their potential for integrating VRE, comparing and ranking of different systems, and identifying best practices, common challenges and needed actions to enable and advance flexibility. It ensures that the complex flexibility question has a clear reference which looks at all relevant flexibility options, without being restricted to a single technology scope. As such it provides a useful instrument for market actors operating in multiple countries, as well as policy makers. As case study, the paper presents a comparative assessment of key European systems using this methodology. The results show that the although flexibility deployment depends on the specifics of each system, a coordinated approach would be beneficial as there are clear no-regret options that face barriers in some systems.
In this paper, a novel mechanistic model is proposed and validated for the consumption of energy in milling processes. The milling machine is considered as a thermodynamic system. Mechanisms of the significant energy conversion processes within the system are used to construct an explicit expression for the power consumption of the machine as a function of the cutting parameters. This model has been validated experimentally and is shown to be significantly more accurate than popular existing models. A simplified form of the model is also proposed that provides a balance between complexity and accuracy. The novelty of the model is that it maps the flow of energy within a machine tool, based solely on the active mechanisms of energy conversion. As a result, only limited assumptions are made in the model, resulting in an error of less than one per cent, verified by experiments. This accurate model can be used to substantially reduce energy consumption in milling processes at machine and factory levels leading to massive cost savings and reduction of environmental impact of numerous industries. The generality of the modelling method makes it applicable to other types of machine tools with minimal adjustments.
Electricity load forecasting is an important tool which can be utilized to enable effective control of commercial building electricity loads. Accurate forecasts of commercial building electricity loads can bring significant environmental and economic benefits by reducing electricity use and peak demand and the corresponding GHG emissions. This paper presents a review of different electricity load forecasting models with a particular focus on regression models, discussing different applications, most commonly used regression variables and methods to improve the performance and accuracy of the models. A comparison between the models is then presented for forecasting day ahead hourly electricity loads using real building and Campus data obtained from the Kensington Campus and Tyree Energy Technologies Building (TETB) at the University of New South Wales (UNSW). The results reveal that Artificial Neural Networks with Bayesian Regulation Backpropagation have the best overall root mean squared and mean absolute percentage error performance and almost all the models performed better predicting the overall Campus load than the single building load. The models were also tested on forecasting daily peak electricity demand. For each model, the obtained error for daily peak demand forecasts was higher than the average day ahead hourly forecasts. The regression models which were the main focus of the study performed fairly well in comparison to other more advanced machine learning models.
In this paper, energy mapping and optimization in rough machining of impellers was investigated. Experiments were first designed based on the response surface methodology (RSM) to minimize operation specific energy consumption in machining through selection of machining parameters (spindle speed, cutting depth, and feed rate) in the Siemens NX computer aided manufacturing (CAM) simulation. With the simulated machining solution and G-code, experiments were conducted on the CNC lathe and mill to cut Al 6061 impellers. The machine energy consumption was measured using a power meter. The operation specific energy was analyzed in analysis of variance (ANOVA), regression models, and desirability functions. The minimum specific energy in the rough and semi-finish turning process is 0.16 J/mm³ and 0.23 J/mm³ respectively. The minimum specific energy in the blades milling process with 6mm ball mill or 3mm ball mill is 0.08 J/mm³ and 0.42 J/mm³ respectively. In the experiment settings, it identified that cutting depth is the most critical factor to affect the specific energy consumption in impeller machining. The empirical equations between the specific energy and material removal rate (MRR) concluded that specific energy is proportional to the inverse of MRR. From the study, it would suggest that in order to minimize the specific energy in machining of impellers, it should selected the MRR as large as possible.
Copyright © 2016 by ASME Country-Specific Mortality and Growth Failure in Infancy and Yound Children and Association With Material Stature
Use interactive graphics and maps to view and sort country-specific infant and early dhildhood mortality and growth failure data and their association with maternal
Increasing energy costs, new environmental legislation, and concerns over energy security are driving efforts to increase industrial energy efficiency across the European Union and the world. Manufacturers are keen to identify the most cost-effective techniques to increase energy efficiency in their factories. To achieve the desired efficiency improvements, energy use should be measured in more detail and in realtime, to derive an awareness of the energy use patterns of every part of the manufacturing system. In this paper, we propose a framework for energy monitoring and management in the factory. This will allow decision support systems and enterprise services to take into consideration the energy used by each individual productive asset and related energy using processes, to facilitate both global and local energy optimization. The proposed framework incorporates standards for energy data exchange, on-line energy data analysis, performance measurement and display of energy usage.
MACHINE LEARNING SYSTEMS automatically learn programs from data. This is often a very attractive alternative to manually constructing them, and in the last decade the use of machine learning has spread rapidly throughout computer science and beyond. Machine learning is used in Web search, spam filters, recommender systems, ad placement, credit scoring, fraud detection, stock trading, drug design, and many other applications. A recent report from the McKinsey Global Institute asserts that machine learning (a.k.a. data mining or predictive analytics) will be the driver of the next big wave of innovation. 15 Several fine textbooks are available to interested practitioners and researchers (for example, Mitchell 16 and Witten et al. 24). However, much of the "folk knowledge" that is needed to successfully develop machine learning applications is not readily available in them. As a result, many machine learning projects take much longer than necessary or wind up producing less-than-ideal results. Yet much of this folk knowledge is fairly easy to communicate. This is the purpose of this article.
A survey revealed that researchers still seem to encounter difficulties to cope with outliers. Detecting outliers by determining an interval spanning over the mean plus/minus three standard deviations remains a common practice. However, since both the mean and the standard deviation are particularly sensitive to outliers, this method is problematic. We highlight the disadvantages of this method and present the median absolute deviation, an alternative and more robust measure of dispersion that is easy to implement. We also explain the procedures for calculating this indicator in SPSS and R software.
Energy consumption reduction is critical in various industrial environments. Machine tool manufacturers could contribute to this matter by developing advanced functions for machines. Power consumption of machining center was measured in various conditions. The conclusion was that modifying cutting conditions reduces energy consumption. This applies for either regular drilling, face/end milling or deep hole machining. Also, a new acceleration control method is developed to reduce energy consumption by synchronizing spindle acceleration with feed system. Experiments were performed to verify these methods and promising results were achieved.
Planning and operating energy-efficient production systems require detailed knowledge on the energy consumption behaviour of their components, energy consumption of production processes, and methods to evaluate design variants. In this paper, the EnergyBlocks methodology for accurate energy consumption prediction is introduced. The methodology is based on the representation of production operations as segments of specific energy consumption for each operating state of the production equipment. Modelling any process chain is possible by arranging the segments according to the production programme. The application of the methodology is demonstrated on the manufacturing of a swash plate expander.
In 1995 CIRP STC “Cutting” started a working group “Modelling of Machining Operations” with the aim of stimulating the development of models capable of predicting quantitatively the performance of metal cutting operations which will be better adapted to the needs of the metal cutting industry in the future. This paper has the character of a progress report. It presents the aims of the working group and the results obtained up to now. The aim is not to review extensively what has been done in the past. It is basically a critical assessment of the present state-of-the-art of the wide and complex field of modelling and simulation of metal cutting operations based on information obtained from the members of the working group, from consultation in industry, study of relevant literature and discussions at meetings of the working group whit the aim to stimulate and pilot future developments. For this purpose much attention is given to a discussion of desirable and possible future developments and planned new activities.
The aim of the work reported in this paper was to develop a new model and methodology for optimising the energy footprint for a machined product. The total energy of machining a component by the turning process was modelled and optimised to derive an economic tool-life that satisfies the minimum energy footprint requirement. The work clearly identifies critical parameters in minimising energy use and hence reducing the energy cost and environmental footprint. Additionally, the paper explores and discusses the conflict and synergy between economical and environmental considerations as well as the effect of system boundaries in determining optimum machining conditions.
The CRISP-DM (CRoss Industry Standard Process for Data Mining) project proposed a comprehensive process model for carrying out data mining projects. The process model is independent of both the industry sector and the technology used. In this paper we argue in favor of a standard process model for data mining and report some experiences with the CRISP-DM process model in practice.
Power consumption is a factor of increasing interest in manufacturing due to its obvious impact on production costs and the environment. The aim of this work is to analyze the influence of process parameters on power consumption in high-speed ball-end milling operations carried out on AISI H13 steel. A total of 300 experiments were carried out in a 3-axis vertical milling center, the Deckel-Maho 105 V linear. The power consumed by the spindle and by the X, Y, and Z machine tool axes was measured using four ammeters located in the respective power cables. The data collected was used to develop an artificial neural network (ANN) which was used to predict power consumption during operations. The results obtained from the ANN are very accurate. Power consumption predictions can help operators to determine the most effective cutting parameters for saving energy and money while bringing the milling process closer to the goal of environmentally sensitive manufacturing which has become a topic of general importance.
This paper collapses the specific electrical energy requirements for a wide range of manufacturing processes into a single plot. The analysis is cast in an exergy framework. The results show: 1) the specific energy requirements for manufacturing processes are not constant as many life cycle analysis tools assume, 2) the most important variable for estimating this energy requirement is the process rate, and 3) the trend in manufacturing process development is toward more and more energy intensive processes. The analysis presented here also provides insight into how equipment can be redesigned in order to be more energy efficient. 1 INTRODUCTION Manufacturing processes include a wide variety of operations, from subtractive processes such as machining and grinding, to net shape processes such as injection molding, to additive processes such as chemical vapor deposition (CVD) and sputtering. All of these manufacturing processes take material inputs, including working materials and auxiliary materials, and transform them into products and wastes. Similarly, the energy inputs into these processes (primarily from electricity) are transformed into useful work, some of which is embodied into the form and composition of the products and wastes, and waste heat. In addition, the energy inputs usually require fuels and produce emissions. For electrical energy inputs, this occurs at the power station. A manufacturing process, along with material and energy flows to and from the process, is diagrammed in Figure 1.
This study focuses on developing a good empirical relationship between the cutting force in an end milling operation and the cutting parameters such as speed, feed and depth-of-cut, by using both multiple regression and neural network modeling processes. A regression model was first fitted to experimentally collected data and any abnormal data points indicated by this analysis were filtered out. By repeating this process several times, a final set of filtered data was obtained and analyzed using neural networks to yield a good, final model. This study shows that analyzing milling force data using conventional regression can lead to a more accurate neural networks model for force prediction.
In this paper we demonstrate that finite linear combinations of compositions of a fixed, univariate function and a set of
affine functionals can uniformly approximate any continuous function ofn real variables with support in the unit hypercube; only mild conditions are imposed on the univariate function. Our results
settle an open question about representability in the class of single hidden layer neural networks. In particular, we show
that arbitrary decision regions can be arbitrarily well approximated by continuous feedforward neural networks with only a
single internal, hidden layer and any continuous sigmoidal nonlinearity. The paper discusses approximation properties of other
possible types of nonlinearities that might be implemented by artificial neural networks.
During the past decade there has been an explosion in computation and information technology. With it have come vast amounts of data in a variety of fields such as medicine, biology, finance, and marketing. The challenge of understanding these data has led to the development of new tools in the field of statistics, and spawned new areas such as data mining, machine learning, and bioinformatics. Many of these tools have common underpinnings but are often expressed with different terminology. This book describes the important ideas in these areas in a common conceptual framework. While the approach is statistical, the emphasis is on concepts rather than mathematics. Many examples are given, with a liberal use of color graphics. It is a valuable resource for statisticians and anyone interested in data mining in science or industry. The book's coverage is broad, from supervised learning (prediction) to unsupervised learning. The many topics include neural networks, support vector machines, classification trees and boosting---the first comprehensive treatment of this topic in any book.
This major new edition features many topics not covered in the original, including graphical models, random forests, ensemble methods, least angle regression and path algorithms for the lasso, non-negative matrix factorization, and spectral clustering. There is also a chapter on methods for ``wide'' data (p bigger than n), including multiple testing and false discovery rates.
Trevor Hastie, Robert Tibshirani, and Jerome Friedman are professors of statistics at Stanford University. They are prominent researchers in this area: Hastie and Tibshirani developed generalized additive models and wrote a popular book of that title. Hastie co-developed much of the statistical modeling software and environment in R/S-PLUS and invented principal curves and surfaces. Tibshirani proposed the lasso and is co-author of the very successful An Introduction to the Bootstrap. Friedman is the co-inventor of many data-mining tools including CART, MARS, projection pursuit and gradient boosting.
The application of time series analysis methods to load forecasting is reviewed. It is shown than Box and Jenkins time series models, in particular, are well suited to this application. The logical and organized procedures for model development using the autocorrelation function and the partial autocorrelation function make these models particularly attractive. One of the drawbacks of these models is the inability to accurately represent the nonlinear relationship between load and temperature. A simple procedure for overcoming this difficulty is introduced, and several Box and Jenkins models are compared with a forecasting procedure currently used by a utility company.
Umweltnutzung und wirtschaft-teil 2: Energie
- Statistisches Bundesamt
Ausgangsbedingungen für die vermarktung von nachfrageflexibilität: Status-quo-analyse und metastudie. 2 fassung
- H U Buhl
- G Fridgen
- M F Körner
- A Michaelis
- V Rägo
- M Schöpf
URL: , licensed under CC BY-SA
- K Jensen
Einsatz maschineller Lernverfahren zur lebenszyklusbasierten Energieprognose für
- Doreth
Systematische Energiedatenerfassung in der Produktion, Ph.D. thesis
- Liebl
Streaming analytics platform tools, solutions & software: Apama
- A G Software