No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Analysis and evaluation of two short-term load forecasting techniques
Abstract
Short-term load forecasting (STLF) is very important for an efficient operation of the power system because the exact and stable load forecasting brings good results to the power system. This manuscript presents the application of two new models in STLF i.e. Cross multi-models and second decision mechanism and Residential load forecasting in smart grid using deep neural network models. In the cross multi-model and second decision mechanism method, the horizontal and longitudinal load characteristics are useful for the construction of the model with the calculation of the total load. The dataset for this model is considered from Maine in New England, Singapore, and New South Wales of Australia. While, In the residential load forecasting method, the Spatio-temporal correlation technique is used for the construction of the iterative ResBlock and deep neural network which helps to give the characteristics of residential load with the use of a publicly available Redd dataset. The performances of the proposed models are calculated by the Root Mean Square Error, Mean Absolute Error, and Mean Absolute Percentage Error. From the simulation results, it is concluded that the performance of cross multi-model and second decision mechanism is good as compare to the residential load forecasting.
... In recent years, increasing attention has been paid to research on energy conservation, consumption reduction, and refined management. Commonly used energy consumption prediction methods include parametric regression, time series analysis, and artificial neural networks [8][9][10][11][12][13]. For example, Chen et al. [14] proposed an energy consumption prediction model for air conditioning systems based on a deep learning gated recurrent unit (GRU) neural network, predicting energy consumption data for the air conditioning system of a tobacco factory's storage workshop. ...
... The continuous development of the global economy and increase in population, as well as the rapid growth in power demand pose unprecedented challenges to the stable operation and efficient dispatch of the power system [1]. Especially today, when distributed energy sources (such as solar energy, wind energy, and other renewable energy sources) are connected in large quantities, the load fluctuation and uncertainty of the power system have increased significantly [2]. Short-term load forecasting is an important component of power system dispatching, and its accuracy is directly related to the stability, economy, and security of the power supply [3]. ...
... Short-term power loads show strong randomness and volatility due to various factors such as climate, economic, and residential electricity consumption behavior. This increases the difficulty of load forecasting [2]. Currently, short-term load forecasting methods can be classified into three categories: mathematical statistics-based forecasting models, traditional machine learning-based forecasting models, and deep learning-based forecasting models. ...
Short-term load forecasting is an important prerequisite for smart grid controls. The current methods are mainly based on the convolution neural network (CNN) or long short-term memory (LSTM) model to realize load forecasting. For the multi-factor input sequence, the existing methods cannot obtain multi-scale features of the time series and the important parameters of the multi-factor, resulting in low accuracy and robustness. To address these problems, a multi-scale feature attention hybrid network is proposed, which uses LSTM to extract the time correlation of the sequence and multi-scale CNN to automatically extract the multi-scale feature of the load. This work realizes the integration of features by constructing a circular network. In the proposed model, a two-branch attention mechanism is further constructed to capture the important parameters of different influencing factors to improve the model’s robustness, which can make the network to obtain effective features at the curve changes. Comparative experiments on two open test sets show that the proposed multi-scale feature attention mixture network can achieve accurate short-term load forecasting and is superior to the existing methods.
... It is used for the planning and maintenance of power networks [17]. The factors which affect the STLF are considered for its work in [18]. ...
For solving the different optimization problems, the cuckoo search is one of the best nature's inspired algorithms. It is an effective technique compare to other optimization methods. For this manuscript, we are using a back propagation neural network for the Xintai power plant consist of short-term electrical load forecasting. The limitation of back propagation is overcome by the cuckoo search algorithm. The function is used for cuckoo search is Gamma probability distribution and its result is compared with other possible cuckoo search methods. The mean average percentage error of Gamma cuckoo search is 0.123%, cuckoo search with Pareto based is 0.127% and Levy based cuckoo search is 0.407%. Other results of the cuckoo search are also found by a linear decreasing switching parameter with a mean average error is 0.344% and 0.389% of mean average error with the use of an exponentially increasing switching parameter. This improved cuckoo search algorithm brings good results in the predicted load which is very important for the Xintai power plant using short-term load forecasting.
The aim of this paper is to examine the link between a family’s scholarly culture and the educational aspirations of parents for their children. Using data from the first and the second wave of the Czech Household Panel Survey (2015, 2016), the study reveals that the number of books in a household – the core aspect of scholarly culture – is significantly linked to the educational aspirations of parents for their children. However, other indicators of scholarly culture (frequency of reading, general interest in books) are not significantly associated with parents’ university aspirations. These results suggest that, in the Czech Republic, the presence of a large number of books in the home signals higher social status and refers to the economic power of parents rather than to a family’s literacy or reading culture.
The current conventional power system load forecasting method mainly outputs forecasting results by constructing a time model, which leads to poor forecasting results due to the lack of effective extraction of feature data of load signals. In this regard, an online compression and reconstruction-based load forecasting method for distribution network power systems is proposed. By introducing the concept of particle swarm ensemble, the discrete situation of power load signal data particles is characterized, and data normalization is carried out, and the load signal data is compressed and reconstructed. The maximum information coefficient is calculated and the load data features are extracted by combining the influencing factors, and finally a hybrid prediction model is constructed and the model is solved. In the experiments, the designed method is verified for its prediction effect. The experimental results show that the designed method has a good fit between the prediction results and the actual load curve, and has a good prediction performance.
Power load prediction is very important for power supply stability and power supply pressure control, and short-term power load prediction is the most important. Therefore, in order to accurately budget short-term power load, this paper will focus on the combined prediction method. Firstly, the concept and forecasting principle of short-term power load are discussed in this paper, and then the combined forecasting model is constructed by using various forecasting methods. At the same time, the short-term power load simulation test is carried out on the model. Through the research, the combined prediction model in the simulation test has achieved the ideal effect, the actual results are basically consistent with the preset results, so the accuracy of the method is guaranteed.KeywordsCombination prediction methodShort-term electrical loadAlgorithm of prediction
For planning and operation of an energy management system, load forecasting (LF) is essential. For smooth power system operation (PS), LF enhances the energy-efficient and reliable operation. LF also helps to calculate energy supplied by utilities to meet the load plus the energy lost in the PS. Every day, it is necessary to schedule the power generation for the next day. So, short-term load forecasting (STLF) is used to calculate the power dispatch for the next day. In unit commitment, economic allocation of generation and maintenance schedules, STLF is also used. So, to make the STLF more effective, fuzzy logic (FL) is used here. FL is essential for weather-sensitive and historical load data for forecasting the load. The fuzzy decision rule identifies the nonlinear relationship between the input and output data. The historical load and hourly data like temperature, humidity (relative humidity) and wind speed are used for input data. For the training and testing, the hourly based load data are collected from the state load dispatch and communication center of Rajasthan Vidyut Prasaran Nigam, Jaipur (JVN). The triangular membership function of the fuzzy logic model is used to predict the load. The performance of the work is determined by the mean absolute percentage error (MAPE) and the MAPE value for pre-holiday (Saturday), holiday (Sunday), post-holiday, and working day is 0.37%, 0.24%, 0.09%, and 0.09%, respectively.
The prediction of the load from a day ahead or a week ahead is called short-term load forecasting. STLF using ANN gives better results in the power grid because the construction of the model is precise, implementation is easy and the performances are good. The weight consisted neural model is a good whose optimal value was found by using various optimization techniques. This paper explains the effect of different machine learning techniques like genetic algorithm, particle swarm optimization, autoregressive integrated moving average, empirical mode decomposition-particle swarm optimization-adaptive network-based fuzzy inference system in STLF and fuzzy logy for the optimization of renewable energy sources, i.e. solar and wind which is also used for the training of the artificial neural network with the silent effect of backpropagation. The study of different machine learning techniques presented their ability to work to produce the results and their extended application in STLF. From the simulation results, we got an empirical mode decomposition-particle swarm optimization-adaptive network-based fuzzy interference system that provides minor error, which is very much permissible compared to other techniques.
The power systems are important by using short term load forecasting (STLF) because it predicts the load in 24 hours ahead or a week ahead. The artificial neural network (ANN) using short term load forecasting brings good result in the predicted load because of its accurateness, easiness in the processing of data, construction of the model as well as excellent performances. The optimization value of ANN is found by different methods which consist of some weights. This manuscript explains the work of ANN with back propagation (BP), genetic algorithm (GA) as well as particle swarm optimization (PSO) for the STLF. The detailed work of the GA and PSO based BP is presenting in this paper which helps for its utilization in the STLF and also able to find the good result in the predicted load. Finally, the result of GA and PSO are compared by simulation and after that, it concluded, the PSO-BP is a good method for STLF using ANN.
This paper proposes an effective computing framework for Short-Term Load Forecasting (STLF). The proposed technique copes with the stochastic variations of the load demand using a stacked generalization approach. This approach combines three models, namely, Light Gradient Boosting Machine (LGBM), eXtreme Gradient Boosting machine (XGB), and Multi-Layer Perceptron (MLP). The inner mechanism of Stacked XGB-LGBM-MLP model consists of generating a meta-data from XGB and LGBM models to compute the final predictions using MLP network. The performance of the proposed Stacked XGB-LGBM-MLP model is validated using two datasets from different locations: Malaysia and New England. The main contributions of this paper are: 1) A novel stacking ensemble-based algorithm is proposed; 2) An effective STLF technique is introduced; 3) A critical multi-study analysis for hyperparameter optimization with five techniques is comprehensively performed; 4) A performance comparative study using two datasets and reference models is conducted. Several case studies have been carried out to prove the performance superiority of the proposed model compared to both existing benchmark techniques and hybrid models.
In this paper, we propose a new ensemble residual network model for short-term load forecasting (STLF). This model improves the accuracy of short-term load forecasting (24 hours in advance). The model has a two-stage network structure. First, the different fully-connected layers are combined, and the combined structure is similar to a recurrent neural network (RNN). Features obtained from historical load data are input to the first stage of the model to get preliminary prediction results. The second stage of the model is a modified residual network, and the final predictions are output from here. We use the ensemble snapshot model with learning rate decay to improve the generalization capability of the model. The model proposed in this paper was trained and tested on two public datasets. Numerical testing shows that the proposed model can get better forecasting results in comparison with other methods, and the ensemble method adopted effectively improves the generalization ability of the model.
Deep neural networks of deep learning algorithms can be applied into regressions and classifications. While the regression performances and classification performances of the deep neural networks are depending on the hyper-parameters of the deep neural networks. To mitigate the adverse effect of the hyper-parameters for the deep learning algorithms, this paper proposes deep forest regression for the short-term load forecasting of power systems. Deep forest regression includes two procedures, i.e., multi-grained scanning procedure and cascade forest procedure. These two procedures can be effectively trained by two completely random forests and two random forests with the default configuration. Then, the deep forest regression is applied into the short-term load forecasting of power systems. The forecasting performances of deep forest regression are compared with that of numerous intelligent algorithms and conventional regression algorithms under the model with the data of previous 7-day, 21-day, and 40-day. Besides, the forecasting performances of deep forest regression with different parameters are compared. The numerical results show that the deep forest regression with default configured parameters can increase the accuracy of the short-term forecasting and mitigate the influences of the experiences for the configuration of the hyper-parameters of deep learning model.
The accurate prediction approach of urban buildings’ electricity consumption is an important foundation for smart urban energy management. It provides the decision basis for reasonable electricity deployments upon different scenarios. Usually, a single model cannot solve linear and nonlinear problems that may occur in electricity consumption prediction effectively, and may produce predictions with unsatisfactory accuracy and stability. Moreover, some prediction models are also poorly interpretable and generalized, which makes them difficult to be applied in practice. To overcome these problems, this paper proposes an ensemble prediction model called gravity gated recurrent unit electricity consumption model which integrates the gated recurrent unit model and the proposed logarithmic electricity consumption gravity model. The weights are derived from average mutual information and weighted entropy. We use 2-years (17520 hours) electricity consumption of a five-star hotel building in Shanghai, China as the study case to illustrate our approach, and apply 9 common prediction models as the benchmarks to conduct the computational experiments and comparisons. Furthermore, we also employ the electricity consumption data of another type of building (office building) to evaluate the generalization capability of the proposed ensemble model. Our approach outperforms all benchmarks in terms of accuracy, stability and generalization.
Accurate forecasting of electric loads has a great impact on actual power generation, power distribution, and tariff pricing. Therefore, in recent years, scholars all over the world have been proposing more forecasting models aimed at improving forecasting performance; however, many of them are conventional forecasting models which do not take the limitations of individual predicting models or data preprocessing into account, leading to poor forecasting accuracy. In this study, to overcome these drawbacks, a novel model combining a data preprocessing technique, forecasting algorithms and an advanced optimization algorithm is developed. Thirty-minute electrical load data from power stations in New South Wales and Queensland, Australia, are used as the testing data to estimate our proposed model’s effectiveness. From experimental results, our proposed combined model shows absolute superiority in both forecasting accuracy and forecasting stability compared with other conventional forecasting models.
Forecasting energy or power usage is an important part of providing a stable supply of power to all customers on a power grid. We present a novel method that aims to forecast the power consumption of a single house, or a set of houses, based on non-intrusive load monitoring (NILM) and graph spectral clustering. In the proposed method, the aggregate power signal is decomposed into individual appliance signals and each appliance's power is forecasted separately. Then the total power forecast is formed by aggregating forecasted power levels of individual appliances. We use four publicly available datasets (REDD, RAE, AMPds2, tracebase) to test our forecasting method and report its accuracy. The results show that our method is more accurate compared to popular existing approaches such as autoregressive integrated moving average (ARIMA), similar profile load forecast (SPLE), artificial neural network (ANN), and recent NILM-based forecasting.
Electric load forecasting (ELF) is vitally beneficial for electrical power planning and economical running in smart grid. However, the medium-term load forecasting (MTLF) has been rarely studied. In addition, existing ELF models mainly consider the impact of limited external factors, which are usually difficult to forecast accurately. In this paper, the characteristics of the electric loads are analyzed and used as guideline for the design of the proposed methods. To fully exploit the quasi-periodicity with different time ranges, i.e., year, quarter, month and week, two deep learning (DL) methods, time-dependency convolutional neural network (TD-CNN) and cycle-based long short term memory (C-LSTM) network, are proposed to improve the forecasting performance of short-term load forecasting (STLF) and MTLF with a little payload of computational complexity. Both of them only utilize the historical electric load, and can mine the underlying load patterns by extracting the long-term global integrated features and short-term local similar features. By representing the loads as pixels and rearranging them into a two-dimensional picture, TD-CNN transforms the temporal correlation of load series into the spatial correlation and keeps the longterm memory. Specifically, the convolutional kernel with special size targeted to load data is designed to extract the local pattern with similar characteristic while the pooling layer is removed in order to keep the finer features. Moreover, in order to extract the temporal correlation between the long-term sequences with lower complexity, the proposed C-LSTM method generates a new short series from the original long load series without information loss. The LSTM is then applied to model the dynamical relationship of the load series with shorter time steps. Experimental results show that the proposed methods outperform the existing method with greatly reduced computation complexity, whose training time is about two to five times shorter than the existing method.
Given the significant fluctuation of errors for single forecasting model and limitation of linear combined forecasting models, A nonlinear multi-model ensemble method for short-term power load forecasting is proposed. Firstly, the power load big data is pre-processed, and multi-dimensional input feature variables are constructed and selected. On this basis, three kinds of single prediction models of random forest, support vector machine and Xgboost are modelled, and three different prediction results are obtained. Then, each individual prediction result and actual load are taken as a new training data set, and secondary learning is performed to obtain a final prediction result. The numerical experiments show that the proposed ensemble method combines the advantages of the single model, and has strong generalization ability and higher stability and accuracy, and has a high practical value.
Accurate day-ahead load prediction plays a significant role to electric companies because decisions on power system generations depend on future behavior of loads. This paper presents a strategy for short-term load forecasting that utilizes support vector regression machines. Proper data preparation, model implementation and model validation methods were introduced in this study. The SVRM model being implemented is composed of specific features, parameters, data architecture and kernel to achieve accurate pattern discovery. The developed model was implemented into an electric load forecasting system using the java open source library called LibSVM. To confirm the effectiveness of the proposed model, the performance of the developed model is evaluated through the validation set of the study and compared to other published models. The created SVRM model produced the lowest Mean Average Percentage Error (MAPE) of 1.48% and was found to be a viable forecasting technique for a day-ahead electric load forecasting system.
With the prevalence of smart meters, fine-grained sub profiles reveal more information about the aggregated load and further help improve the forecasting accuracy. Ensemble is an effective approach for load forecasting. It either generates multiple training datasets or applies multiple forecasting models to produce multiple forecasts. In this letter, a novel ensemble method is proposed to forecast the aggregated load with sub profiles where the multiple forecasts are produced by different groupings of sub profiles. Specifically, the sub profiles are first clustered into different groups and forecasting is conducted on the grouped load profiles individually. Thus, these forecasts can be summed to form the aggregated load forecast. In this way, different aggregated load forecasts can be obtained by varying the number of clusters. Finally, an optimal weighted ensemble approach is employed to combine these forecasts and provide the final forecasting result. Case studies are conducted on two open datasets and verify the effectiveness and superiority of the proposed method.
As the power system is facing a transition towards a more intelligent, flexible and interactive system with higher penetration of renewable energy generation, load forecasting, especially short-term load forecasting for individual electric customers plays an increasingly essential role in the future grid planning and operation. Other than aggregated residential load in a large scale, forecasting an electric load of a single energy user is fairly challenging due to the high volatility and uncertainty involved. In this paper, we propose a long short-term memory (LSTM) recurrent neural network (RNN) based framework, which is the latest and one of the most popular techniques of deep learning, to tackle this tricky issue. The proposed framework is tested on a publicly available set of real residential smart meter data, of which the performance is comprehensively compared to various benchmarks including the state-of-the-arts in the field of load forecasting. As a result, the proposed LSTM approach outperforms the other listed rival algorithms in the task of short-term load forecasting for individual residential households.
Deep neural networks (DNNs) are currently widely used for many artificial intelligence (AI) applications including computer vision, speech recognition, and robotics. While DNNs deliver state-of-the-art accuracy on many AI tasks, it comes at the cost of high computational complexity. Accordingly, techniques that enable efficient processing of deep neural network to improve energy-efficiency and throughput without sacrificing performance accuracy or increasing hardware cost are critical to enabling the wide deployment of DNNs in AI systems. This article aims to provide a comprehensive tutorial and survey about the recent advances towards the goal of enabling efficient processing of DNNs. Specifically, it will provide an overview of DNNs, discuss various platforms and architectures that support DNNs, and highlight key trends in recent efficient processing techniques that reduce the computation cost of DNNs either solely via hardware design changes or via joint hardware design and network algorithm changes. It will also summarize various development resources that can enable researchers and practitioners to quickly get started on DNN design, and highlight important benchmarking metrics and design considerations that should be used for evaluating the rapidly growing number of DNN hardware designs, optionally including algorithmic co-design, being proposed in academia and industry. The reader will take away the following concepts from this article: understand the key design considerations for DNNs; be able to evaluate different DNN hardware implementations with benchmarks and comparison metrics; understand trade-offs between various architectures and platforms; be able to evaluate the utility of various DNN design techniques for efficient processing; and understand of recent implementation trends and opportunities.
Machine learning plays a vital role in several modern economic and industrial fields, and selecting an optimized machine learning method to improve time series’ forecasting accuracy is challenging. Advanced machine learning methods, e.g., the support vector regression (SVR) model, are widely employed in forecasting fields, but the individual SVR pays no attention to the significance of data selection, signal processing and optimization, which cannot always satisfy the requirements of time series forecasting. By preprocessing and analyzing the original time series, in this paper, a hybrid SVR model is developed, considering periodicity, trend and randomness, and combined with data selection, signal processing and an optimization algorithm for short-term load forecasting. Case studies of electricity power data from New South Wales and Singapore are regarded as exemplifications to estimate the performance of the developed novel model. The experimental results demonstrate that the proposed hybrid method is not only robust but also capable of achieving significant improvement compared with the traditional single models and can be an effective and efficient tool for power load forecasting.
Demand response programs are currently being proposed as a solution to deal with issues related to peak demand and to improve the operation of the electric power system. In the demand response paradigm, electric utilities provide incentives and benefits to private consumers as a compensation for their flexibility in the timing of their electricity consumption. In this paper, a dynamic energy management framework, based on highly resolved energy consumption models, is used to simulate automated residential demand response. The models estimate the residential demand using a novel bottom-up approach that quantifies consumer energy use behavior, thus providing an accurate estimation of the actual amount of controllable resources. The optimal schedule of all of the controllable appliances, including plug-in electric vehicles, is found by minimizing consumer electricity-related expenditures. Recently, time-varying electricity rate plans have been proposed by electric utilities as an incentive to their customers with the objective of re-shaping the aggregate demand. Large-scale simulations are performed to analyze and quantitatively assess the impact of demand response programs using different electricity price structures. Results show that simple time-varying electricity price structures, coupled with large-scale adoption of automated energy management systems, might create pronounced rebound peaks in the aggregate residential demand. To cope with the rebound peaks created by the synchronization of the individual residential demands, innovative electricity price structures—called Multi-TOU and Multi-CPP—are proposed.
Extreme learning machines (ELMs) have proven to be efficient and effective learning mechanisms for pattern classification and regression. However, ELMs are primarily applied to supervised learning problems. Only a few existing research papers have used ELMs to explore unlabeled data. In this paper, we extend ELMs for both semi-supervised and unsupervised tasks based on the manifold regularization, thus greatly expanding the applicability of ELMs. The key advantages of the proposed algorithms are as follows: 1) both the semi-supervised ELM (SS-ELM) and the unsupervised ELM (US-ELM) exhibit learning capability and computational efficiency of ELMs; 2) both algorithms naturally handle multiclass classification or multicluster clustering; and 3) both algorithms are inductive and can handle unseen data at test time directly. Moreover, it is shown in this paper that all the supervised, semi-supervised, and unsupervised ELMs can actually be put into a unified framework. This provides new perspectives for understanding the mechanism of random feature mapping, which is the key concept in ELM theory. Empirical study on a wide range of data sets demonstrates that the proposed algorithms are competitive with the state-of-the-art semi-supervised or unsupervised learning algorithms in terms of accuracy and efficiency.
In future smart grids, consumers of electricity will be enabled to react to electricity prices. The aggregate reaction of consumers can potentially shift the demand curve in the market, resulting in prices that may differ from the initial forecasts. In this paper, a hybrid forecasting framework is proposed that takes such dynamics into account when forecasting electricity price and demand. The proposed framework combines a multi-input multi-output (MIMO) forecasting engine for joint price and demand prediction with data association mining (DAM) algorithms. In this framework, a DAM-based rule extraction mechanism is used to determine and extract the patterns in consumers' reaction to price forecasts. The extracted rules are then employed to fine-tune the initially generated demand and price forecasts of a MIMO engine. Simulation results are presented using Australia's and New England's electricity market data.
Demand response programs are useful options in reducing electricity price, congestion relief, load shifting, peak clipping, valley filling and resource adequacy from the system operator’s viewpoint. For this purpose, many models of these programs have been developed. However, the availability of these resources has not been properly modeled in demand response models making them not practical for long-term studies such as in the resource adequacy problem where considering the providers’ responding uncertainties is necessary for long-term studies.
In this paper, a model considering providers’ unavailability for unforced demand response programs has been developed. Temperature changes, equipment failures, simultaneous implementation of demand side management resources, popular TV programs and family visits are the main reasons that may affect the availability of the demand response providers to fulfill their commitments. The effectiveness of the proposed model has been demonstrated by numerical simulation.
Microgrid is an effective means of integrating multiple energy sources of distributed energy to improve the economy, stability and security of the energy systems. A typical microgrid consists of Renewable Energy Source (RES), Controllable Thermal Units (CTU), Energy Storage System (ESS), interruptible and uninterruptible loads. From the perspective of the generation, the microgrid should be operated at the minimum operating cost, whereas from the perspective of demand, the energy cost imposed on the consumer should be minimum. The main key in controlling the relationship of microgrid with the utility grid is managing the demand. An Energy Management System (EMS) is required to have real time control over the demand and the Distributed Energy Resources (DER). Demand Side Management (DSM) assesses the actual demand in the microgrid to integrate different energy resources distributed within the grid. With these motivations towards the operation of a microgrid and also to achieve the objective of minimizing the total expected operating cost, the DER schedules are optimized for meeting the loads. Demand Response (DR) a part of DSM is integrated with MG islanded mode operation by using Time of Use (TOU) and Real Time Pricing (RTP) procedures. Both TOU and RTP are used for shifting the controllable loads. RES is used for generator side cost reduction and load shifting using DR performs the load side control by reducing the peak to average ratio. Four different cases with and without the PV, wind uncertainties and ESS are analyzed with Demand Response and Unitcommittment (DRUC) strategy. The Strawberry (SBY) algorithm is used for obtaining the minimum operating cost and to achieve better energy management of the Microgrid.
Transformer may enter the ageing cycle sooner if it is loaded more than the rated value for longer periods of time in its life cycle. This paper exploits demand response as a way to improve transformer life by reducing the hottest spot temperature (HST) which is caused due to better load profile. The aim of the paper is to investigate the impact of various types of price-based and incentive-based demand response programs (DRPs) on the transformer life and other attributes like energy consumption, peak to average ratio, etc. Entropy method is used to determine the weights of multi-attributes in a multi-attribute decision-making (MADM) model formed by the various attributes and the multifarious demand response programs. Using these weights, the various DRPs are ranked using Program Ranking Index to assist the utility in deciding which DRP is to be employed. IEEE transformer model is used to calculate the transformer ageing for two cases with and without demand response programs. The simulation results validate the effectiveness of demand response in mitigating transformer loss of life. Furthermore, the economic and technical benefits of employing demand response are quantified.
An efficient and economic scheduling of power plants relies on an accurate demand forecast especially for the short-term due to its tight relation to power markets and trading operations in interconnected power systems. A slight deviation of load prediction from real demand could engender the start-up of a conventional power station which could be either time-consuming or requiring expensive combustible, a deviation that could interfere as well with renewables intermittency and demand response strategies. Hence, load forecasting still a challenging subject because of the various transformations that the energy sector undergoes and that directly impact the demand profile shape. Therefore, conceiving dynamic load demand forecast approaches will permit utilities save money in different vertical structures and regulation schemes. In this paper, we propose a novel approach for short-term demand prediction valid for normal and special days to address the impact of climate changes along with events occurrence on forecast accuracy. This approach is based on the prediction of hourly loads, established on the daily peak load prediction using backpropagation combined to chi-squared method for weighting historical data to enhance the training process. Obtained results from extensive testing on the Moroccan’s power system confirm the strength of the developed approach, that improved the forecast accuracy by a range of 1.1–4% compared to the existing methods.
The electrical load forecasting is a fundamental technique for consumer load prediction for utilities. The accurate load forecasting is crucial to design Demand Response (DR) programs in the paradigm of smart grids. Artificial Neural Network (ANN) based techniques have been widely used in recent years and applied to predict the electric load with high accuracy to participate in DR programs for commercial, industrial and residential consumers. This research work is focused on the use and comparison of two ANN-based load forecasting techniques, i.e. Feed-Forward Neural Network (FFNN) and Echo State Network (ESN), on a dataset related to commercial buildings, in view of a possible DR program application. The results of both models are compared based on the load forecasting accuracy through experimental measurements and suitably defined metrics.
Nowadays, the sustainable energy management of industrial environments is of great importance because of their heavy loads and behaviors. In this paper, the Virtual Power Plant (VPP) idea is commented as a collected generation to be an appropriate approach for these networks handling. Here, Technical Industrial VPP (TIVPP) is characterized as a dispatching unit contains demands and generations situated in an industrial network. A complete structure is proposed here for possible conditions for different VPPs cooperation in the power market. This structure carries out a day-ahead and intra-day generation planning by choosing the best Demand Response (DR) programs considering wind power and market prices as the uncertain parameters. A risk management study is likewise taken into account in the proposed stages for contingency conditions. So, some component changes, like, regular demand changes and single-line outage are prepared in the framework to authorize the suggested concept in the contingency situation. To determine the adequacy and productivity of the proposed strategy, the IEEE-RTS modified framework is examined to test the technique and to evaluate some reassuring perspectives too. By the proposed methodology, the delectability of DR projects is uncovered in industrial networks and the improvement level of load shedding and the lower cost will be achieved.
Due to the influence of random factors such as travel behavior of car owners and traffic condition, the electric vehicle (EV) charging station load has strong randomness. Establishing an appropriate probability model to describe the stochastic of EV charging station load is of great significance to the safe operation analysis of distribution network. Therefore, this paper proposes the probability modeling and the scenario generation method for EV charging station load based on historical data. Firstly, the load of each period is regarded as a random variable, and the probability distribution model of each random variable is obtained by fitting historical data. Secondly, according to the analysis of load correlation, 96 periods of a day are divided into several sets of adjacent periods. Considering the correlation between different period loads in each set of adjacent periods, the Pair-copula method and the D-Vine structure is used to obtain the joint distribution model in each set of adjacent period loads is obtained. Thirdly, according to the joint distributionmodel, all sets of adjacent periods of a day and the corresponding load scenarios are obtained, and the daily load curve considering time correlation are generated. Finally, referring to the actual historical load curve data, the effectiveness of the proposed method in this paper is verified by comparing with the daily load curve, which is generated based on the independent distribution model of each period load without considering time correlation.
During the past two decades, providing solutions to enhance the efficiency of power systems, like optimal consumption management has been attracting a good deal of attention. Demand Response (DR) programs, have always been among the appropriate ways to persuade consumers to alter consumption patterns. In the main, the implementation of DR programs is carried out by price-based and incentive-based strategies. In this paper, first, a brief overview of the smart grid principles on retail electricity pricing is presented. Then, a hybrid price-based demand response (HPDR) is proposed, which is more adaptable to pricing principles compared to other existing strategies. This strategy is implemented in day-ahead scheduling of a residential microgrid. Moreover, to increase the accuracy of the proposed model, the uncertainty regarding decision variables and parameters including the generation units, load dispatch in the Micro-grid is considered. Finally, the results of numerical studies show the effectiveness of the proposed retail pricing strategy, and demonstrate a decrease in Peak-to-Valley (PtV) index and Coefficient of Variation Percentage (CVP) by almost 12% and 25% as well as an increase in social welfare indicator, power sale at peak times, respectively, by approximately 18%, 24%, and 25% in comparison with other methods.
Short-term electric load forecasting is important for evaluating the power utility performance in terms of price and income elasticities, energy transfer scheduling, unit commitment and load dispatch. Support vector regression (SVR) approach applies a simple linear regression in the high-dimensional feature space (Hilbert space) by using kernel functions and has many attractive features and profound empirical performances for small sample, nonlinearity and high dimensional dataset. However, the SVR modeling processing has computation complexity of order O(K×N³) (where N is the size of the training dataset, and K is the evaluation number of the parameter selection process). To forecast short-term power load accurately, quickly and efficiently, a sequential grid approach based support vector regression (SGA-SVR) is proposed in this work. Specifically, for a given data set, parameter regression surface is conducted in SVR modeling processing with its forecasting performance as dependent variable and the three parameters (ε,C,γ) as independent variables. Then, a novel grid algorithm is presented to provide a new way for fitting the parameter regression surface. The statistical inference is also given by introducing the asymptotic normality of a fixed grid point of parameters. The numerical experiments using SGA-SVR model demonstrate the superiority over the standard SVR model and accuracy of forecast is greatly improved especially for short-term forecasts.
The electric load forecasting is extremely important for energy demand management, stability and security of power systems. A sufficiently accurate, robust and fast short-term load forecasting (STLF) model is necessary for the day-to-day reliable operation of the grid. The characteristics of load series such as non-stationarity, non-linearity, and multiple-seasonality make such prediction a troublesome task. This difficulty is conventionally tackled with model-driven methodologies that demand domain-specific knowledge. However, the ideal choice is a data-driven methodology that extracts relevant and meaningful information from available data even when the physical model of the system is unknown. The present work is focused on developing a data-driven strategy for short-term load forecasting (STLF) that employs dynamic mode decomposition (DMD). The dynamic mode decomposition is a matrix decomposition methodology that captures the spatio-temporal dynamics of the underlying system. The proposed data-driven model efficiently identifies the characteristics of load data that are affected by multiple exogenous factors including time, day, weather, seasons, social activities, and economic aspects. The effectiveness of the proposed DMD based strategy is confirmed by conducting experiments on energy market data from different smart grid regions. The performance advantage is verified using output quality measures such as RMSE, MAPE, MAE, and running time. The forecasting results are observed to be competing with the benchmark methods. The satisfactory performance suggests that the proposed data-driven model can be used as an effective tool for the real-time STLF task.
This work presents a collaborative scheme for the end-users in a smart building with multiple housing units. This approach determines a day-ahead operational plan that provides demand-response services by taking into account the amount of energy consumed per household, the use of shared storage and solar panels, and the amount of shifted load. We use a biobjective optimization model to trade off total user satisfaction versus total cost of energy consumption. The optimization works in combination with a price structure based on time and level of use that encourages load shifting and benefits the participants. Computational experiments and an extensive sensitivity analysis validate the performance of the proposed approach and help to clarify its strengths, its limits, and the requirements for ensuring the desired outcome.
Short term probabilistic load forecasting is essential for any power generating utility. This paper discusses an application of partially linear additive quantile regression models for predicting short term electricity demand during the peak demand hours (i.e. from 18:00 to 20:00) using South African data for January 2009 to June 2012. Additionally the bounded variable mixed integer linear programming technique is used on the forecasts obtained in order to find an optimal number of units to commit (switch on or off. Variable selection is done using the least absolute shrinkage and selection operator. Results from the unit commitment problem show that it is very costly to use gas fired generating units. These were not selected as part of the optimal solution. It is shown that the optimal solutions based on median forecasts (Q0.5 quantile forecasts) are the same as those from the 99th quantile forecasts except for generating unit g8c, which is a coal fired unit. This shows that for any increase in demand above the median quantile forecasts it will be economical to increase the generation of electricity from generating unit g8c. The main contribution of this study is in the use of nonlinear trend variables and the combining of forecasting with the unit commitment problem. The study should be useful to system operators in power utility companies in the unit commitment scheduling and dispatching of electricity at a minimal cost
particularly during the peak period when the grid is constrained due to increased demand for electricity.
Electrical power system (EPS) forecasting plays a significant role in economic and social development but it remains an extremely challenging task. Because of its significance, relevant studies on EPS are especially needed. More specifically, only a few of the previous studies in this area conducted in-depth investigations of the entire EPS forecasting and merely focused on modeling individual signals centered on wind speed or electrical load. Moreover, most of these past studies concentrated on accuracy improvements and usually ignore the significance of forecasting stability. Therefore, to simultaneously achieve high accuracy and dependable stability, a hybrid forecasting framework based on the multi-objective dragonfly algorithm (MODA) was successfully developed in this study. The framework consists of four modules—data preprocessing, optimization, forecasting, and evaluation modules. In this framework, MODA is employed to optimize the Elman neural network (ENN) model as a part of the optimization module to overcome the drawbacks of single-objective optimization algorithms. In addition, data preprocessing and evaluation modules are incorporated to improve forecasting performance and conduct a comprehensive evaluation for this framework, respectively. Empirical results reveal that the developed forecasting framework can be an effective tool for the planning and management of power grids.
The treatment of trend components in electricity demand is critical for long-term peak load forecasting. When forecasting high frequency variables, like daily or hourly loads, a typical problem is how to make long-term scenarios - regarding demographics, GDP growth, etc. - compatible with short-term projections. Traditional procedures that apply de-trending methods are unable to simulate forecasts under alternative long-term scenarios. On the other hand, existing models that allow for changes in long-term trends tend to be characterized by end-of-year discontinuities. In this paper a novel forecasting procedure is presented that improves upon these approaches and is able to combine long and short-term features by employing temporal disaggregation techniques. This method is applied to forecast electricity load for Spain and its performance is compared to that of a nonlinear autoregressive neural network with exogenous inputs. Our proposed procedure is flexible enough to be applied to different scenarios based on alternative assumptions regarding both long-term trends as well as short-term projections.
In this paper, we newly propose a holographic ensemble forecasting method (HEFM). First, we use the MI and statistical method to select feature variables, which is an ensemble of information about the cross-border multi-source data at the dataset level. Then, we generate multiple training sets by performing diversity sampling with bootstrap, which is an ensemble of information about multiple sample sets at the sampling space level. Next, we construct a multi-model using different artificial intelligence and machine-learning algorithms, which is an ensemble of information about multiple nonlinear heterogeneous models at the forecasting model level. Finally, we use the original features, the forecasting load which is output of the multiple heterogeneous models trained in the first learning, and the actual load of the recent period before each forecasted time to generate a new training set, which is used for the online second learning and final forecasting. This is an ensemble of information about online second learning at the decision level. The ensemble of multi-category multi-state information for four levels (dataset, sampling space, forecasting model, and decision) constitutes the framework of HEFM, whose essence is a forecasting method with comprehensive information integration for the whole life cycle of the forecasting process. We study the load in Guangzhou, China and New England, USA. Compared to the state-of-the-art forecasting methods, the MAPE of HEFM is reduced by 7.69%-65.77%. The results demonstrate that the forecasting performance may not be improved with the number of algorithms, and that there is a need to understand the positive and negative fusion effect between different algorithms and data characteristics.
The key challenge for household load forecasting lies in the high volatility and uncertainty of load profiles. Traditional methods tend to avoid such uncertainty by load aggregation (to offset uncertainties), customer classification (to cluster uncertainties) and spectral analysis (to filter out uncertainties). This paper, for the first time, aims to directly learn the uncertainty by applying a new breed of machine learning algorithms – deep learning. However simply adding layers in neural networks will cap the forecasting performance due to the occurrence of overfitting. A novel pooling-based deep recurrent neural network (PDRNN) is proposed in this paper which batches a group of customers’ load profiles into a pool of inputs. Essentially the model could address the over-fitting issue by increasing data diversity and volume. This work reports the first attempts to develop a bespoke deep learning application for household load forecasting and achieved preliminary success. The developed method is implemented on Tensorflow deep learning platform and tested on 920 smart metered customers from Ireland. Compared with the state-of-art techniques in household load forecasting, the proposed method outperforms ARIMA by 19.5%, SVR by 13.1% and classical deep RNN by 6.5% in terms of RMSE.
Consumer Demand Response (DR) is an important research and industry problem, which seeks to categorize, predict and modify consumer's energy consumption. Unfortunately, traditional clustering methods have resulted in many hundreds of clusters, with a given consumer often associated with several clusters, making it difficult to classify consumers into stable representative groups and to predict individual energy consumption patterns. In this paper, we present a shape-based approach that better classifies and predicts consumer energy consumption behavior at the household level. The method is based on Dynamic Time Warping. DTW seeks an optimal alignment between energy consumption patterns reflecting the effect of hidden patterns of regular consumer behavior. Using real consumer 24-hour load curves from Opower Corporation, our method results in a 50% reduction in the number of representative groups and an improvement in prediction accuracy measured under DTW distance. We extend the approach to estimate which electrical devices will be used and in which hours.
Short-term load forecasting (STLF) plays an irreplaceable role in the efficient management of electrical systems but remains an extremely challenging task. To achieve the goal of load forecasting with both accuracy and stability, a combined model based on a multi-objective optimization algorithm, the multi-objective flower pollination algorithm (MOFPA), is developed in this study. In this combined model, MOPFA is used to optimize the weights of single models to simultaneously obtain high accuracy and great stability, which are two mostly independent objectives and are equally important to the model effectiveness. Data preprocessing techniques, such as the fast ensemble empirical mode decomposition and multiple seasonal patterns, are also incorporated in this model. Case studies of half-hourly electrical load data from the State of Victoria, the State of Queensland, and New South Wales, Australia, are considered as illustrative examples to evaluate the effectiveness and efficiency of the developed combined model. The experimental results clearly show that both the accuracy and stability of the combined model are superior to those of the single models.
Large-scale utilization of electric vehicles (EVs) affects the total electricity demand considerably. Demand forecast is usually designed for the seasonally changing load patterns. However, with the high penetration of EVs, daily charging demand makes traditional forecasting methods less accurate. This paper presents an autoregressive integrated moving average (ARIMA) method for demand forecasting of conventional electrical load (CEL) and charging demand of EV (CDE) parking lots simultaneously. Our EV charging demand prediction model takes daily driving patterns and distances as an input to determine the expected charging load profiles. The parameters of the ARIMA model are tuned so that the mean square error (MSE) of the forecaster is minimized. We improve the accuracy of ARIMA forecaster by optimizing the integrated and auto-regressive order parameters. Furthermore, due to the different seasonal and daily pattern of CEL and CDE, the proposed decoupled demand forecasting method provides significant improvement in terms of error reduction. The impact of EV charging demand on the accuracy of the proposed load forecaster is also analyzed in two approaches: (1) integrated forecaster for CEL + CDE, and (2) decoupled forecaster that targets CEL and CDE independently. The forecaster outputs are used to formulate a chance-constrained day-ahead scheduling problem. The numerical results show the effectiveness of the proposed forecaster and its influence on the stochastic power system operation.
Although combining forecasts is well-known to be an effective approach to improving forecast accuracy, the literature and case studies on combining electric load forecasts are relatively limited. In this paper, we investigate the performance of combining so-called sister load forecasts, i.e. predictions generated from a family of models which share similar model structure but are built based on different variable selection processes. We consider 11 combination algorithms (three variants of arithmetic averaging, four regression based, one performance based method and three forecasting techniques used in the machine learning literature) and two selection schemes. Through comprehensive analysis of two case studies developed from public data (Global Energy Forecasting Competition 2014 and ISO New England), we demonstrate that combing sister forecasts outperforms the benchmark methods significantly in terms of forecasting accuracy measured by Mean Absolute Percentage Error. With the power to improve accuracy of individual forecasts and the advantage of easy generation, combining sister load forecasts has a high academic and practical value for researchers and practitioners alike.
Renewable electricity producers must trade in day-ahead electricity markets in the same manner as conventional producers. However, their power production may be highly unpredictable and nondispatchable. This is the case, for example, of wind and solar power producers, which thus need to use the balancing market to mend eventual deviations with respect to their day-ahead schedule. This chapter presents close formulae to determine the optimal offering strategy of stochastic producers in the day-ahead market. The analytical solution to these formulae is available under certain assumptions on the probabilistic structure characterizing power production and market prices. Stochastic programming is then introduced as a powerful mathematical framework to rid the solution to the trading problem for stochastic producers of these simplifying assumptions.
With the deployment of advanced metering infrastructure (AMI), an avalanche of new energy-use information became available. Better understanding of the actual power consumption patterns of customers is critical for improving load forecasting and efficient deployment of smart grid technologies to enhance operation, energy management, and planning of electric power systems. Unlike traditional aggregated system-level load forecasting, the AMI data introduces a fresh perspective to the way load forecasting is performed, ranging from very short-term load forecasting to long-term load forecasting at the system level, regional level, feeder level, or even down to the consumer level. This paper addresses the efforts involved in improving the system level intraday load forecasting by applying clustering to identify groups of customers with similar load consumption patterns from smart meters prior to performing load forecasting.
Little is known about variations in electricity use at finely-resolved timescales, or the drivers for those variations. Using measured electricity use data from 103 homes in Austin, TX, this analysis sought to (1) determine the shape of seasonally-resolved residential demand profiles, (2) determine the optimal number of normalized representative residential electricity use profiles within each season, and (3) draw correlations to the different profiles based on survey data from the occupants of the 103 homes. Within each season, homes with similar hourly electricity use patterns were clustered into groups using the k-means clustering algorithm. Then probit regression was performed to determine if homeowner survey responses could serve as predictors for the clustering results. This analysis found that Austin homes fall into one of two seasonal groups with some homes using more expensive electricity (from a wholesale electricity market perspective) than others. Regression results indicate that variables such as if someone works from home, hours of television watched per week, and education levels have significant correlations with average profile shape, but might vary across seasons. The results herein also indicate that policies such as time-of-use or real-time electricity structures might be more likely to affect lower income households during some high electricity use parts of the year.
In this comprehensive empirical study we critically evaluate the use of forecast averaging in the context of electricity prices. We apply seven averaging and one selection scheme and perform a backtesting analysis on day-ahead electricity prices in three major European and US markets. Our findings support the additional benefit of combining forecasts of individual methods for deriving more accurate predictions, however, the performance is not uniform across the considered markets and periods. In particular, equally weighted pooling of forecasts emerges as a simple, yet powerful technique compared with other schemes that rely on estimated combination weights, but only when there is no individual predictor that consistently outperforms its competitors. Constrained least squares regression (CLS) offers a balance between robustness against such well performing individual methods and relatively accurate forecasts, on average better than those of the individual predictors. Finally, some popular forecast averaging schemes – like ordinary least squares regression (OLS) and Bayesian Model Averaging (BMA) – turn out to be unsuitable for predicting day-ahead electricity prices.
This addition to the ISOR series addresses the analytics of the operations of electric energy systems with increasing penetration of stochastic renewable production facilities, such as wind- and solar-based generation units.
As stochastic renewable production units become ubiquitous throughout electric energy systems, an increasing level of flexible backup provided by non-stochastic units and other system agents is needed if supply security and quality are to be maintained.
Within the context above, this book provides up-to-date analytical tools to address challenging operational problems such as:
• The modeling and forecasting of stochastic renewable power production.
• The characterization of the impact of renewable production on market outcomes.
• The clearing of electricity markets with high penetration of stochastic renewable units.
• The development of mechanisms to counteract the variability and unpredictability of stochastic renewable units so that supply security is not at risk.
• The trading of the electric energy produced by stochastic renewable producers.
• The association of a number of electricity production facilities, stochastic and others, to increase their competitive edge in the electricity market.
• The development of procedures to enable demand response and to facilitate the integration of stochastic renewable units.
This book is written in a modular and tutorial manner and includes many illustrative examples to facilitate its comprehension. It is intended for advanced undergraduate and graduate students in the fields of electric energy systems, applied mathematics and economics. Practitioners in the electric energy sector will benefit as well from the concepts and techniques explained in this book.
This study addresses the problem of modeling the electricity demand loads in Greece. The provided actual load data is deseasonilized and an AutoRegressive Moving Average (ARMA) model is fitted on the data off-line, using the Akaike Corrected Information Criterion (AICC). The developed model fits the data in a successful manner. Difficulties occur when the provided data includes noise or errors and also when an on-line/adaptive modeling is required. In both cases and under the assumption that the provided data can be represented by an ARMA model, simultaneous order and parameter estimation of ARMA models under the presence of noise are performed. The produced results indicate that the proposed method, which is based on the multi-model partitioning theory, tackles successfully the studied problem. For validation purposes the produced results are compared with three other established order selection criteria, namely AICC, Akaike's Information Criterion (AIC) and Schwarz's Bayesian Information Criterion (BIC). The developed model could be useful in the studies that concern electricity consumption and electricity prices forecasts.
Short-term electricity load is effected by various factors, It has the certain difficulty to make prediction accurate, but we can improve prediction precise by continuously optimizing forecasting methods. This paper carried out the combination of ARIMA several methods based on the idea of time sequence, to avoid deficiencies in various aspects, perfect forecasting methods, make ARIMA model can conduct electricity short-term load forecasting better.
This paper presents a summary of Demand Response (DR) in deregulated electricity markets. The definition and the classification of DR as well as potential benefits and associated cost components are presented. In addition, the most common indices used for DR measurement and evaluation are highlighted, and some utilities’ experiences with different demand response programs are discussed. Finally, the effect of demand response in electricity prices is highlighted using a simulated case study.
Forecasters are generally concerned about the properties of model-based predictions in the presence of structural change. In this paper, it is argued that forecast errors can under those conditions be greatly reduced through systematic combination of forecasts. We propose various extensions of the standard regression-based theory of forecast combination. Rolling weighted least squares and time-varying parameter techniques are shown to be useful generalizations of the basic framework. Numerical examples, based on various types of structural change in the constituent forecasts, indicate that the potential reduction in forecast error variance through these methods is very significant. The adaptive nature of these updating procedures greatly enhances the effect of risk-spreading embodied in standard combination techniques.
Load forecasting is usually made by constructing models on relative information, such as climate and previous load demand data. In 2001, EUNITE network organized a competition aiming at mid-term load forecasting (predicting daily maximum load of the next 31 days). During the competition we proposed a support vector machine (SVM) model, which was the winning entry, to solve the problem. In this paper, we discuss in detail how SVM, a new learning technique, is successfully applied to load forecasting. In addition, motivated by the competition results and the approaches by other participants, more experiments and deeper analyses are conducted and presented here. Some important conclusions from the results are that temperature (or other types of climate information) might not be useful in such a mid-term load forecasting problem and that the introduction of time-series concept may improve the forecasting.