No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A review on time series forecastingtechniques for building energy consumption
Abstract
Energy consumption forecasting for buildings has immense value in energy efficiency and sustainability research. Accurate energy forecasting models have numerous implications in planning and energy optimization of buildings and campuses. For new buildings, where past recorded data is unavailable, computer simulation methods are used for energy analysis and forecasting future scenarios. However, for existing buildings with historically recorded time series energy data, statistical and machine learning techniques have proved to be more accurate and quick. This study presents a comprehensive review of the existing machine learning techniques for forecasting time series energy consumption. Although the emphasis is given to a single time series data analysis, the review is not just limited to it since energy data is often co-analyzed with other time series variables like outdoor weather and indoor environmental conditions. The nine most popular forecasting techniques that are based on the machine learning platform are analyzed. An in-depth review and analysis of the ‘hybrid model’, that combines two or more forecasting techniques is also presented. The various combinations of the hybrid model are found to be the most effective in time series energy forecasting for building.
... In general, time series comprises four components: (i) trend-which refers to the upward or downward movement of the series during the period of observation; (ii) seasonality-which is the periodic fluctuation of the variable submitted to the analysis; (iii) cycles-which concerns cyclical movements that occur in unknown periods and that are completed in a period of more than 1 year; and (iv) residuals-which refers to random movements, also known as noises, that represent the remaining, almost inexplicable part of the time series (Makridakis et al. 1998;Deb et al. 2017). Moreover, time series analysis can be further divided into two categories: univariate (used to describe time series that have a single observation sequenced over time) and multivariate analyses (when a group of variables and their involved interactions are considered in the time series) (Deb et al. 2017). ...
... In general, time series comprises four components: (i) trend-which refers to the upward or downward movement of the series during the period of observation; (ii) seasonality-which is the periodic fluctuation of the variable submitted to the analysis; (iii) cycles-which concerns cyclical movements that occur in unknown periods and that are completed in a period of more than 1 year; and (iv) residuals-which refers to random movements, also known as noises, that represent the remaining, almost inexplicable part of the time series (Makridakis et al. 1998;Deb et al. 2017). Moreover, time series analysis can be further divided into two categories: univariate (used to describe time series that have a single observation sequenced over time) and multivariate analyses (when a group of variables and their involved interactions are considered in the time series) (Deb et al. 2017). ...
... The forecasting techniques used by ARIMA models are based on the idea of transforming time series into stationary by the differentiation process (Deb et al. 2017). Stationarity is a property that indicates whether the statistical attributes (mean, variance, and autocorrelation function) remain constant over time. ...
This study aims to predict the potential for secondary lead recovery from motorcycle batteries in Brazil, since this is considered the second largest category of automobiles in the country. To achieve this objective, a forecasting model based on the ARIMA methodology was applied, with input data taken from Brazilian sectorial platforms. Furthermore, an analysis of the data, of the residuals, autocorrelation tests, as well as Kolmogorov-Smirnov and Dickey-Fuller tests, were performed. The SARIMA model (3,1,0) (2,0,0)12 presented a better adaptation to the behavior of the series. The results showed that the amount of secondary lead obtained based on the forecast model will be 89,972,842.08 million tons between 2021 and 2030 (14 million tons of lead originated only from motorcycle LABs in 2021). These results show a possible insufficiency of the installed capacity to supply the amount of lead to be processed in the country, not to mention the LABs from other vehicles (light and heavy) and other emerging battery technologies from electric vehicles. In addition, an analysis was conducted on the importance of secondary lead for the economy and the dangers of illegal recycling in Brazil. In general, this study contributes to the understanding of the importance of secondary production of lead in Brazil, an important asset for a country that does not have sufficient primary production for its domestic demand. The findings may assist in several alternatives for the proper planning and management of the collection, disposal and recycling of lead, providing the Brazilian government with directions for the development of new policies related to lead recycling.
... The growth and development rate of countries around the world is and has been, annually and inevitably, increasing significantly [1]. The extraordinary increase in the global population, related to economic advancement, industrialization, social advances, and expectations of prosperity, has had a significant impact on energy and environmental issues [2]. Associated with the growth of the human population in the demand for housing and well-being, the development of countries and societies will also continue to increase. ...
... Associated with the growth of the human population in the demand for housing and well-being, the development of countries and societies will also continue to increase. To ensure these growths and developments, an increase in energy generation will be needed to stimulate global demand and, at the same time, the environment should be kept safe [1,2]. The increase in population and consumption patterns have promoted the increase in energy consumption that, unavoidably, has been growing at an annual high rate. ...
... To meet the objectives of the Paris Agreement and reduce Greenhouse Gas (GHG) emissions, it is essential to move towards a low-carbon energy system [4]. The International Energy Agency (IEA) has identified energy efficiency as one of the measures to ensure the long-term decarbonization of the energy sector [2]. One of the main solutions to reduce emissions is to reduce the intensity of primary energy through energy efficiency [5] and electrification-based on renewable energy sources (RES)-a solution increasingly adopted for the industrial, commercial, residential, and transport sectors. ...
In this paper, a systematic literature review is presented, through a survey of the main digital databases, regarding modelling methods for Short-Term Load Forecasting (STLF) for hourly electricity demand for residential electricity and to realize the performance evolution and impact of Artificial Intelligence (AI) in STLF. With these specific objectives, a conceptual framework on the subject was developed, along with a systematic review of the literature based on scientific publications with high impact and a bibliometric study directed towards the scientific production of AI and STLF. The review of research articles over a 10-year period, which took place between 2012 and 2022, used the Preferred Reporting Items for Systematic and Meta-Analyses (PRISMA) method. This research resulted in more than 300 articles, available in four databases: Web of Science, IEEE Xplore, Scopus, and Science Direct. The research was organized around three central themes, which were defined through the following keywords: STLF, Electricity, and Residential, along with their corresponding synonyms. In total, 334 research articles were analyzed, and the year of publication, journal, author, geography by continent and country, and the area of application were identified. Of the 335 documents found in the initial research and after applying the inclusion/exclusion criteria, which allowed delimiting the subject addressed in the topics of interest for analysis, 38 (thirty-eight) documents were in English (26 journal articles and 12 conference papers). The results point to a diversity of modelling techniques and associated algorithms. The corresponding performance was measured with different metrics and, therefore, cannot be compared directly. Hence, it is desirable to have a unified dataset, together with a set of benchmarks with well-defined metrics for a clear comparison of all the modelling techniques and the corresponding algorithms.
... The STLF techniques can be categorized roughly into Time series and Artificial Intelligence (AI) methods [4,5]. Time series methods rely on the historical load data and use statistical techniques to model the pattern of the load demand over time. ...
... This network architecture has multiple input nodes that can take as many inputs as there are features in the dataset. Individual input sequences of size 96 are converted to multidimensional input of shape (4,24), corresponding to 4 days with 24-hour entries. These multidimensional inputs are processed independently through separate network branches, each specialized for modeling a specific feature. ...
Short-term Load Forecasting (STLF) is a challenging task for an Energy Management System (EMS) that depends on highly unpredictable and volatile factors, making it difficult to predict the electricity load demand accurately. Despite the challenges, it is an essential component, as it helps to ensure energy demand-supply equilibrium, prevents blackouts, reduces the need for expensive peak power generation, and improves the efficiency and reliability of the EMS. Motivated by these factors, we have proposed a novel STLF framework using a multi-input parallel ConvLSTM model. The effectiveness of the proposed model is verified using two publicly available load-series datasets. On the Malaysia dataset, the proposed model yields 998.12, 2.59%, 1590.34, and 0.987 for MAE, MAPE, RMSE, and R ² , respectively. Similarly, on the Tetouan dataset, this model yields 1737.32, 2.49%, 2254.91, and 0.976 for MAE, MAPE, RMSE, and R ² , respectively. These outperforming results found in the comparative experiments are further statistically verified using Friedman's test. The presenting framework of STLF can help EMS to make informed decisions about resource allocation and system operations.
... There are statistical and AI methods employed in the forecast of carbon prices; while statistical time prediction models, such as GARCH and ARIMA, have difficulties modeling the nonlinear characteristics of time series and are unable to effectively handle carbon price sequences, AI methods have been employed with more success [17]. AI models that predict carbon prices can be divided into two categories: traditional machine learning models and emerging deep learning models [14]. The application of extensive machine learning models has proven to be nonlinearly advantageous in predicting carbon prices [19,20,43]. ...
Under the strict carbon emission quota policy in China, the urban carbon price directly affects the operation of enterprises, as well as forest carbon sequestration. As a result, accurately forecasting carbon prices has been a popular research topic in forest science. Similar to stock prices, urban carbon prices are difficult to forecast using simple models with only historical prices. Fortunately, urban remote sensing images containing rich human economic activity information reflect the changing trend of carbon prices. However, properly integrating remote sensing data into carbon price forecasting has not yet been investigated. In this study, by introducing the powerful transformer paradigm, we propose a novel carbon price forecasting method, called MFTSformer, to uncover information from urban remote sensing and historical price data through the encoder–decoder framework. Moreover, a self-attention mechanism is used to capture the intrinsic characteristics of long-term price data. We conduct comparison experiments with four baselines, ablation experiments, and case studies in Guangzhou. The results show that MFTSformer reduces errors by up to 52.24%. Moreover, it outperforms the baselines in long-term accurate carbon price prediction (averaging 15.3%) with fewer training resources (it converges rapidly within 20 epochs). These findings suggest that the effective MFTSformer can offer new insights regarding AI to urban forest research.
... Several studies (e.g., [4][5][6][7]) utilized ordinary regression and built a predictive model from time-series variables. Running a linear regression model with time series data results in an incorrect estimate of the parameters of interest, together with an inflated standard error term [8][9][10]. ...
When ordinary regression analysis is performed using time-series variables, it is common for the errors (residuals) to have a time-series structure. This violates the usual assumption of independent errors in ordinary least squares (OLS) regressions. Consequently, the estimates of the coefficients and their standard errors are incorrect if the time-series structure of the errors is ignored. In this study, an investigation of a regression model with time-series variables, particularly a simple case, was conducted using the conventional method. The ‘AirPassengers Dataset’ was downloaded from the R repository used for the analysis. Ordinary least squares and Cochrane-Orcutt procedures were used as methodologies. The results show that the adjusted regression model with autoregressive errors outperformed the ordinary regression model.
... The findings of these studies may be explained by the fact that econometric forecasting models perform poorly under the application of multivariable and heteroskedasticity problems (Khan et al., 2020). Other forecasting approaches have used simulation techniques to forecast several types of time series (Deb, Zhang, Yang, Lee, & Shah, 2017;Mawson & Hughes, 2020;Park & Kim, 2023;Zheng, Yu, Wang, & Tao, 2019). For instance, Khalil and Fatmi (2022) investigated the impact of the COVID-19 pandemic on residential energy consumption by adopting a hybrid approach consisting of agent-based simulation, machine learning and energy simulation techniques. ...
... A similar approach based on LSTM network is proposed by Somu et al. [6], who propose a hybrid model using an improved sine cosine optimization algorithm for accurate and robust building energy consumption forecasting. A general overview of existing machine learning techniques which use multiple time-series variables (not only energy consumption, but also for example weather and indoor conditions) is provided in [7]. A structured description of methods for estimating the customer baseline load is presented in [8], with a focus on existing standardisation efforts and lessons from use cases, while Segovia et al. [9] analyze in their work the most common forecasting algorithms used in the electrical grids, providing a comprehensive comparison. ...
... Implementing such models is widely applicable across various sectors, including energy forecasting, where they can aid in establishing mitigation policies to reduce operational costs and address challenges related to climate management. Additionally, they can be beneficial for modeling building maintenance and predictive control for HVAC systems [16,25], bringing operational benefits. ...
Forecasting energy consumption models allow for improvements in building performance and reduce energy consumption. Energy efficiency has become a pressing concern in recent years due to the increasing energy demand and concerns over climate change. This paper addresses the energy consumption forecast as a crucial ingredient in the technology to optimize building system operations and identifies energy efficiency upgrades. The work proposes a modified multi-head transformer model focused on multi-variable time series through a learnable weighting feature attention matrix to combine all input variables and forecast building energy consumption properly. The proposed multivariate transformer-based model is compared with two other recurrent neural network models, showing a robust performance while exhibiting a lower mean absolute percentage error. Overall, this paper highlights the superior performance of the modified transformer-based model for the energy consumption forecast in a multivariate step, allowing it to be incorporated in future forecasting tasks, allowing for the tracing of future energy consumption scenarios according to the current building usage, playing a significant role in creating a more sustainable and energy-efficient building usage.
... Many review works have been done on load forecasting techniques [4][5][6], but few have a focus on DL techniques. Moreover, previous review work does not cover many of the papers. ...
Short-term load forecasting (STLF) is a part of the smart grid (SG) system used in maintenance and management operations. Traditional machine learning (ML) techniques entail complicating and time-consuming processes of feature extraction and selection. Deep learning (DL) techniques of artificial neural network (ANN) have shown great potential in STLF. The modernization of SG and the availability of huge load data offer an opportunity for these DL techniques in STLF. Different techniques based on DL models have been proposed for STLF in the past few years. In this paper, a survey of DL model for STLF is presented. This literature survey includes papers published from 2016 to 2019. Common DL architectures such as stack auto-encoder (SAE), recurrent neural network (RNN), convolution neural network (CNN), and deep belief network (DBN) are frequently applied in combination with clustering methods. These DL architectures are briefly explained with a diagram before presenting a review of related papers. The strengths and limitations of the reviewed methods are discussed. Based on this review, the gaps in the existing research work on DL-based STLF are identified, and future directions are described. This paper is expected to serve as an initial guide for new researchers who are interested in the application of deep learning in STLF.
... Artificial Neural Networks (ANNs) play a significant role in Deep Learning as highly effective models for solving complex and nonlinear prediction problems (Zhao & Magoules, 2012). They excel at extracting meaningful insights from intricate and uncertain data (Deb et al., 2017). ANNs demonstrate their expertise by learning specific patterns within a given dataset. ...
High-performance concrete finds extensive application in a diverse range of civil engineering structures, such as towering buildings, swift roadways, sea-spanning bridges, dams, and marine constructions. Anticipating the compressive strength of concrete poses significant challenges due to its inherent complexity. However, this study successfully utilized machine learning approaches to accurately predict the compressive strength of high-strength concrete. Through experimental programs, the researchers gathered 100 data points, enabling a comprehensive evaluation of the model's predictive capabilities across various strength parameters. The constructed models exhibited precise predictions of the compressive strength of geopolymer concrete, as evidenced by the high R² value and low root-mean-square error value.
... Machine learning techniques are increasingly used to predict building energy consumption, 11−13 indoor environmental quality, 14−17 occupant sensation, 18−23 and airborne exposure. 24−26 So far, the majority of relevant studies have focused on forecasting using time series analysis 27 or neural network models. 28,29 Further, the dynamics of indoor environmental parameters can be linked with building physics and occupant activities, 30,31 yet understanding this complex link requires advanced data analysis tools. ...
Low-cost air quality monitors are increasingly being deployed in various indoor environments. However, data of high temporal resolution from those sensors are often summarized into a single mean value, with information about pollutant dynamics discarded. Further, low-cost sensors often suffer from limitations such as a lack of absolute accuracy and drift over time. There is a growing interest in utilizing data science and machine learning techniques to overcome those limitations and take full advantage of low-cost sensors. In this study, we developed an unsupervised machine learning model for automatically recognizing decay periods from concentration time series data and estimating pollutant loss rates. The model uses k-means and DBSCAN clustering to extract decays and then mass balance equations to estimate loss rates. Applications on data collected from various environments suggest that the CO2 loss rate was consistently lower than the PM2.5 loss rate in the same environment, while both varied spatially and temporally. Further, detailed protocols were established to select optimal model hyperparameters and filter out results with high uncertainty. Overall, this model provides a novel solution to monitoring pollutant removal rates with potentially wide applications such as evaluating filtration and ventilation and characterizing indoor emission sources.
... Natural gas consumption prediction was evaluated on the basis of several models [4,5]. The predicting models for natural gas consumption are distributed into three groups: (a) conventional statistical models [6], (b) artificial intelligence (AI)-based models [7], and (c) hybrid models [8]. ...
T he aims of this study is to predict natural gas consumption and price in residential and commercial building into five big cities (California, Washington, New-York, Texas, and Florida) in the United States. Three different machine learning algorithms such as linear model (LM), Support Vector Machine regression (SVMR), and Random Forest regression (RFR) have been used. Four statistical tests (ANOVA, Chi-square, regression, and Minitab tests) have allowed to select among the eleven weather parameters those that affected significantly the performance of natural gas. Finally, after applied these four tests, only minimum, mean, maximum air temperature and mean air speed have been recognized as principal parameters having a direct impact on the natural gas consumption. To decide on the success of these algorithms, four different statistical metrics (Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Regression coefficient R², and mean square error (MSE)) were discussed in this study. The results showed that linear model and Random forest regression could be applied to predict the natural gas consumption with a good accuracy, despite this, Random Forest regression model is the best fitting model among all the three models used. It is followed by LMR, and SVMR, respectively.
... As a result, new techniques for time-series analysis have become among the most sought after in machine learning where we seek to learn temporal trends and correlations from time-series data to perform classification [1], anomaly detection [2,3], regression [4], forecasting [5,6] and to generate synthetic time-series instances [7]. Focusing on classification, classical machine learning has provided a zoo of algorithms where the present state-of-the-art is centered around deep learning. ...
Supervised time-series classification garners widespread interest because of its applicability throughout a broad application domain including finance, astronomy, biosensors, and many others. In this work, we tackle this problem with hybrid quantum-classical machine learning, deducing pairwise temporal relationships between time-series instances using a time-series Hamiltonian kernel (TSHK). A TSHK is constructed with a sum of inner products generated by quantum states evolved using a parameterized time evolution operator. This sum is then optimally weighted using techniques derived from multiple kernel learning. Because we treat the kernel weighting step as a differentiable convex optimization problem, our method can be regarded as an end-to-end learnable hybrid quantum-classical-convex neural network, or QCC-net, whose output is a data set-generalized kernel function suitable for use in any kernelized machine learning technique such as the support vector machine (SVM). Using our TSHK as input to a SVM, we classify univariate and multivariate time-series using quantum circuit simulators and demonstrate the efficient parallel deployment of the algorithm to 127-qubit superconducting quantum processors using quantum multi-programming.
... The Autoregressive Moving Average (ARIMA) method is one of the most fundamental time series techniques. This technique uses time series principles to predict future outcomes as a linear equation based on input data and taking prediction error into account (Deb, Zhang, Yang, Lee & Shah, 2017). Because of its great accuracy and simplicity, the ARIMA algorithm has been widely employed in energy research (Prado, Minutolo & Kristjanpoller, 2020). ...
With the growth of population, many countries face the challenge of supplying energy resources. One approach to managing and planning these resources is to predict energy demand. This study presented an integrated approach by applying six Machine Learning (ML) algorithms (ANN, AR, ARIMA, SARIMA, SARIMAX, and LSTM) and mathematical programming to predict energy demand in Iran up to 2040. The data relating to electricity generation and fuel consumption in power plants, electricity imports and exports, and seven major energy-consuming sectors in Iran (residential, commercial, industrial, transportation, public, agriculture, and others) are collected. The data employed to forecast energy demand in each sector with ML algorithms and prediction accuracy indices evaluated the algorithms' prediction accuracy in every sector. Then, the optimization model for prediction accuracy improvement is introduced. The ML algorithms results are employed as inputs to the integrated model and executed by two PSO and Grey-Wolf Optimizer algorithms for different sectors. The energy demand in these seven sectors until 2040 is predicted, and five prediction accuracy metrics are used to validate the integrated optimization results. The outcomes of the proposed method in all sectors reflect its more accurateness than ML algorithms, such that the MAPE index equals 0.002-0.012 and 0.004-0.013 for the proposed model executed by the PSO and Grey-Wolf Optimizer algorithms. In general, the PSO algorithm indicates a 75.65% growth in the total energy demand of all sectors, and the Grey-Wolf Optimizer algorithm forecasts a 82.94% growth.
... With the rise of computer power and deep learning techniques, there has been a lot of progress in the field of time-series analysis and prediction. Several reviews outline the state of the art when it comes to time-series forecasting algorithms [53][54][55] and time-series classification algorithms [56][57][58], some with the aim to make them interpretable [59][60][61]. ...
Mechanistic models have been used for centuries to describe complex interconnected processes, including biological ones. As the scope of these models has widened, so have their computational demands. This complexity can limit their suitability when running many simulations or when real-time results are required. Surrogate machine learning (ML) models can be used to approximate the behaviour of complex mechanistic models, and once built, their computational demands are several orders of magnitude lower. This paper provides an overview of the relevant literature, both from an applicability and a theoretical perspective. For the latter, the paper focuses on the design and training of the underlying ML models. Application-wise, we show how ML surrogates have been used to approximate different mechanistic models. We present a perspective on how these approaches can be applied to models representing biological processes with potential industrial applications (e.g., metabolism and whole-cell modelling) and show why surrogate ML models may hold the key to making the simulation of complex biological systems possible using a typical desktop computer.
... For nonlinear problems in energy prediction and evaluation, methods such as random forest (RF) [Dudek, 2015], support Vector Regression (SVR) [Chen et al., 2017], gradient boost [Patnaik et al., 2021], Gaussian process [Shepero et al., 2018], or fuzzy systems [Efendi et al., 2015] are used. However, due to their greater flexibility in constructing forecasting models in the presence of features such as nonlinearity, high volatility, and uncertainty of load profiles, artificial intelligence models generally outperform conventional statistical models in terms of forecasting accuracy [Deb et al., 2017;Bianchi et al., 2017;Nti et al., 2020]. Therefore, artificial intelligence methods are increasingly used for load forecasting, while conventional statistical models are still used in certain applications. ...
Electricity load forecasting is crucial for effectively managing and optimizing power grids. Over the past few decades, various statistical and deep learning approaches have been used to develop load forecasting models. This paper presents an interpretable machine learning approach that identifies load dynamics using data-driven methods within an operator-theoretic framework. We represent the load data using the Koopman operator, which is inherent to the underlying dynamics. By computing the corresponding eigenfunctions, we decompose the load dynamics into coherent spatiotemporal patterns that are the most robust features of the dynamics. Each pattern evolves independently according to its single frequency, making its predictability based on linear dynamics. We emphasize that the load dynamics are constructed based on coherent spatiotemporal patterns that are intrinsic to the dynamics and are capable of encoding rich dynamical features at multiple time scales. These features are related to complex interactions over interconnected power grids and different exogenous effects. To implement the Koopman operator approach more efficiently, we cluster the load data using a modern kernel-based clustering approach and identify power stations with similar load patterns, particularly those with synchronized dynamics. We evaluate our approach using a large-scale dataset from a renewable electric power system within the continental European electricity system and show that the Koopman-based approach outperforms a deep learning (LSTM) architecture in terms of accuracy and computational efficiency. The code for this paper has been deposited in a GitHub repository, which can be accessed at the following address github.com/Shakeri-Lab/Power-Grids.
Energy predicting gains attention for its ability to manage and control energy consumption in a building. The multiple linear regression model is known for its simplicity and effective when dealing with electricity consumption. In this work, the authors have utilized the multiple linear regression (MLR) model to predict the hourly electricity energy consumption, in winter, for school buildings. For the case study, school buildings in the South of France are used. In this model, nine predictor variables are considered, namely, (1) level of indoor CO2, (2) indoor temperature, (3) indoor humidity, (4) outdoor temperature, (5) outdoor humidity, (6) global solar radiation, (7) day index (weekday/weekend), (8) time index (occupied/non-occupied), and (9) building net floor area. The first order and two-way interaction models are constructed using all predictors. The coefficient of determination (R2) is a model evaluation metric that assesses the relationship between the values of the desired outcomes and those that the model predicts. The results show that the two-way interaction model has better R2 for both training set (R2 = 74%) and testing set (R2 = 77%). However, this model gives underestimated results for higher values of electricity consumption starting from 30 kWh/h. It is also not reliable for one of the buildings as the R2 is only 55% and the inaccuracy rate is 69%. Overall, this model is a starting point for future work to improve its predicting ability by adding other influential explanatory variables.
Recent studies have shown great performance of Transformer-based models in long-term time series forecasting due to their ability in capturing long-term dependencies. However, Transformers have their limitations when training on small datasets because of their lack in necessary inductive bias for time series forecasting, and do not show significant benefits in short-time step forecasting as well as that in long-time step as the continuity of sequence is not focused on. In this paper, efficient designs in Transformers are reviewed and a design of decomposing residual convolution neural networks or DRCNN is proposed. The DRCNN method allows to utilize the continuity between data by decomposing data into residual and trend terms which are processed by a designed convolution block or DR-Block. DR-Block has its strength in extracting features by following the structural design of Transformers. In addition, by imitating the multi-head in Transformers, a Multi-head Sequence method is proposed such that the network is enabled to receive longer inputs and more accurate forecasts are obtained. The state-of-the-art performance of the presented model are demonstrated on several datasets.
Background
The COVID-19 pandemic has significantly altered the global health and medical landscape. In response to the outbreak, Chinese hospitals have established 24-hour fever clinics to serve patients with COVID-19. The emergence of these clinics and the impact of successive epidemics have led to a surge in visits, placing pressure on hospital resource allocation and scheduling. Therefore, accurate prediction of outpatient visits is essential for informed decision-making in hospital management.
Objective
Hourly visits to fever clinics can be characterized as a long-sequence time series in high frequency, which also exhibits distinct patterns due to the particularity of pediatric treatment behavior in an epidemic context. This study aimed to build models to forecast fever clinic visit with outstanding prediction accuracy and robust generalization in forecast horizons. In addition, this study hopes to provide a research paradigm for time-series forecasting problems, which involves an exploratory analysis revealing data patterns before model development.
Methods
An exploratory analysis, including graphical analysis, autocorrelation analysis, and seasonal-trend decomposition, was conducted to reveal the seasonality and structural patterns of the retrospective fever clinic visit data. The data were found to exhibit multiseasonality and nonlinearity. On the basis of these results, an ensemble of time-series analysis methods, including individual models and their combinations, was validated on the data set. Root mean square error and mean absolute error were used as accuracy metrics, with the cross-validation of rolling forecasting origin conducted across different forecast horizons.
ResultsHybrid models generally outperformed individual models across most forecast horizons. A novel model combination, the hybrid neural network autoregressive (NNAR)-seasonal and trend decomposition using Loess forecasting (STLF), was identified as the optimal model for our forecasting task, with the best performance in all accuracy metrics (root mean square error=20.1, mean absolute error=14.3) for the 15-days-ahead forecasts and an overall advantage for forecast horizons that were 1 to 30 days ahead.
Conclusions
Although forecast accuracy tends to decline with an increasing forecast horizon, the hybrid NNAR-STLF model is applicable for short-, medium-, and long-term forecasts owing to its ability to fit multiseasonality (captured by the STLF component) and nonlinearity (captured by the NNAR component). The model identified in this study is also applicable to hospitals in other regions with similar epidemic outpatient configurations or forecasting tasks whose data conform to long-sequence time series in high frequency exhibiting multiseasonal and nonlinear patterns. However, as external variables and disruptive events were not accounted for, the model performance declined slightly following changes in the COVID-19 containment policy in China. Future work may seek to improve accuracy by incorporating external variables that characterize moving events or other factors as well as by adding data from different organizations to enhance algorithm generalization.
Purpose
This study aims to propose the use of time series autoregressive integrated moving average (ARIMA) models to predict gas path performance in aero engines. The gas path is a critical component of an aero engine and its performance is essential for safe and efficient operation of the engine.
Design/methodology/approach
The study analyzes a data set of gas path performance parameters obtained from a fleet of aero engines. The data is preprocessed and then fitted to ARIMA models to predict the future values of the gas path performance parameters. The performance of the ARIMA models is evaluated using various statistical metrics such as mean absolute error, mean squared error and root mean squared error. The results show that the ARIMA models can accurately predict the gas path performance parameters in aero engines.
Findings
The proposed methodology can be used for real-time monitoring and controlling the gas path performance parameters in aero engines, which can improve the safety and efficiency of the engines. Both the Box-Ljung test and the residual analysis were used to demonstrate that the models for both time series were adequate.
Research limitations/implications
To determine whether or not the two series were stationary, the Augmented Dickey–Fuller unit root test was used in this study. The first-order ARIMA models were selected based on the observed autocorrelation function and partial autocorrelation function.
Originality/value
Further, the authors find that the trend of predicted values and original values are similar and the error between them is small.
Short-term load forecasting is an essential component of a power grid management system that performs the electric load profile. It is challenging as it depends on nonlinear and stochastic parameters. Although a good number of research works have been published on s short-term load forecasting but still accurate and effective techniques are needed. In this paper, we perform a comparative analysis of different deep neural network architectures frequently employed in short-term load forecasting in recent years. Different state of the art strategy such as preprocessing, time-dependent feature extraction, feature engineering, feature selection, and feature transformation, are explored. The better performance of hybrid models based on 1D-CNN, LSTM, and GRU is shown using three benchmark datasets. Standard error matrices such as MAE, MAPE, RMSE, and R\(^{2}\) represent these results. The best-performing hybrid models are then compared with existing works from other papers using similar datasets.KeywordsShort-term load forecastingElectric load predictionDeep learningDeep neural networkConvolutional neural networkLong short-term memoryGated recurrent unitHybrid deep neural network.
Deep learning models have achieved extensive popularity due to their capability for providing an end-to-end solution. But, these models require training a massive amount of data, which is a challenging issue and not always enough data is available. In order to get around this problem, a few shot learning methods emerged with the aim to achieve a level of prediction based only on a small number of data. This paper proposes a few-shot learning approach that can successfully learn and predict the electricity consumption combining both the use of temporal and spatial data. Furthermore, to use all the available information, both spatial and temporal, models that combine the use of Recurrent Neural Networks and Graph Neural Networks have been used. Finally, with the objective of validate the approach, some experiments using electricity data of consumption of thirty-six buildings of the University of Alicante have been conducted.KeywordsFew-shot learningGraph neural networksElectricity consumptionPattern recognition
Indoor environmental quality (IEQ) can impact human health, well-being, and productivity. This paper reviews the couplings between IEQ which consists of thermal, acoustic, visual comfort, and indoor air quality, and parameters such as human comfort and productivity, with the overarching aim of ensuring a healthy environment while reducing energy consumption in buildings. At current rates of population growth, it is anticipated that the tropics would be home to over half of the world's people by 2050, hence, special attention is paid to studies conducted in tropical climates to prepare a comprehensive review with a specific climatic context. Notably,we highlight the need for more data-driven IEQ research in tropical regions, the importance of broadening the scope of and maintaining uniformity in IEQ standards, and the significance of an adaptive and sustainable built environment as we move ahead. A discussion highlighting the existing challenges and opportunities – especially the integration of AI – for future research in the area is also presented.
Globally, the construction industry is experiencing an increase in energy demand, which has significant environmental and economic repercussions. To address these issues, it is now possible for buildings, vehicles, and renewable energy sources to collaborate and function as an advanced, integrated, and environmentally favorable system that meets the high energy demands of contemporary buildings. To attain maximum efficiency, however, it is necessary to create reliable energy demand forecasting models. In this research, by introducing the energy model of a neighbourhood with buildings with solar panels and electric vehicles, the final balance of energy production and consumption for each building and the whole neighbourhood as a micro grid is predicted. DesignBuilder is used to model neighbourhood buildings, and K-Nearest neighbor (KNN), Regression Support Vector (SVR), Adaptive Boosting (AdaBoost), and Deep neural networks (DNN) algorithms in machine learning are used to predict the final energy balance. a comparative analysis of the performance of the KNN, SVR, AdaBoost, and DNN algorithms was conducted to determine which algorithm is the most effective in predicting energy balance. Finally, the Root Mean Square Error (RMSE) has been used to validate the prediction models. The results show that the KNN, SVR, AdaBoost, and DNN algorithms had RMSE values of 0.56, 0.92, 0.95, and 0.53, respectively. Among these algorithms, the DNN and KNN algorithms had more accurate results than the other used algorithms and as a result of this research, An accurate forecast of neighbourhood energy balance was made. This study takes a novel approach by developing a model that takes into account an integrated system of houses, solar cells, and electric consumption for each building in a neighborhood, which can help to optimize energy consumption and reduce environmental impact.
It is of great importance to build an accurate model for the multi‐step forecasting of household power consumption. In recent years, more and more researchers have focused on adopting hybrid models to execute forecasting due to the irregularity and nonlinearity of power data. However, existing forecasting models usually make predictions directly on original data. This paper introduces secondary decomposition algorithm used to decompose primal data. We use singular spectrum analysis (SSA) to decompose the original series into several subseries, utilize variational mode decomposition (VMD) optimized by whale optimization algorithm to decompose the subseries with the highest frequency into several intrinsic mode functions . Then all subseries obtained from SSA and VMD are fed into long short‐term memory model to get predictions. In order to confirm the validity of proposed model, this paper performs several experimental analyses. The results of experiments show that the proposed model effectively improves the accuracy of forecasting. © 2023 Institute of Electrical Engineer of Japan and Wiley Periodicals LLC.
Energy market liberalization brings new opportunities, since large consumers have direct access to energy trading to buy energy for the next day. However, that requires a good estimation of the expected amount of energy and its hourly distribution in advance. On the other hand, smart energy meters are being installed in many facilities with the aim of achieving holistic submetering systems. These systems consist of a set of meters structured in several levels, so that there are hierarchical relations among upstream and downstream meters. This information could be exploited for achieving accurate one-day-ahead energy predictions. However, submetering systems might be incomplete due to unavailable meters or lost energy. In this paper, we propose a hierarchical prediction method for incomplete submetering systems that is based on 2D convolutional neural network (2D CNN) and is able to perform day ahead prediction of power consumption. This method exploits the hierarchical relations among meters and considers periodicity in order to forecast the power consumption for the next day. The proposed hierarchical method has proved to be more accurate and fast to forecast power consumption in incomplete submetering systems than using an individual predictions.KeywordsHierarchical predictionPower consumptionSubmetering systemsConvolutional Neural Networks
In the last few years, control engineers have started to use artificial neural networks (NNs) embedded in advanced feedback control algorithms. Its natural integration into existing controllers, such as programmable logic controllers (PLCs) or close to them, represents a challenge. Besides, the application of these algorithms in critical applications still raises concerns among control engineers due to the lack of safety guarantees. Building trustworthy NNs is still a challenge and their verification is attracting more attention nowadays. This paper discusses the peculiarities of formal verification of NNs controllers running on PLCs. It outlines a set of properties that should be satisfied by a NN that is intended to be deployed in a critical high-availability installation at CERN. It compares different methods to verify this NN and sketches our future research directions to find a safe NN.KeywordsVerification of neural networksPLCsControl system
Time Series Forecasting (TSF) is well established in domains dealing with temporal data to predict future events yielding the basis for strategic decision-making. Previous research indicated that forecasting models are vulnerable to adversarial attacks, that is, maliciously crafted perturbations of the original data with the goal of altering the model’s predictions. However, attackers targeting specific outcomes pose a substantially more severe threat as they could manipulate the model and bend it to their needs. Regardless, there is no systematic approach for targeted adversarial learning in the TSF domain yet. In this paper, we introduce targeted attacks on TSF in a systematic manner. We establish a new experimental design standard regarding attack goals and perturbation control for targeted adversarial learning on TSF. For this purpose, we present a novel indirect sparse black-box evasion attack on TSF, n Vita. Additionally, we adapt the popular white-box attacks Fast Gradient Sign Method (FGSM) and Basic Iterative Method (BIM). Our experiments confirm not only that all three methods are effective but also that current state-of-the-art TSF models are indeed susceptible to attacks. These results motivate future research in this area to achieve higher reliability of forecasting models.KeywordsAdversarial LearningTime SeriesTargeted AttackForecasting
Energy assessments and studies for the coming years in buildings show that the effects of climate change, particularly the problem of global warming, play an important role in increasing the accuracy of these calculations. A similar educational building in two different cities, Tehran and Tabriz, was considered in this study. The goal was to investigate and assess the impact of new and old data on predicting energy consumption in the coming years. The Typical Meteorological Year data was prepared using a Typical Meteorological Year data Creator software. To simulate energy consumption, the Energy Plus software was also used. Taking into account the hot and cold seasons in the forecasted years (2020–2050–2080), this issue has been proven so that for the city of Tehran in the year 2080 compared to 2020, there will be a 24% reduction in energy consumption in the heating part and a 5% increase in the cooling part. In addition, for the city of Tabriz in 2080, we will see a 19% reduction in heating energy consumption and a 7% increase in cooling energy consumption. It should be noted that, using more recent data produces estimates that are more precise than those produced using older data. Furthermore, as temperatures rise in the coming years as a result of global warming, heating will decrease and cooling will increase in Tabriz and Tehran.
The architectural design of buildings has a significant impact on the consumption
of non-renewable resources. In developed countries, the energy consumption in the
building sector is more than 333 of the total energy. Over the past decade, energy
consumption in various fields, including educational spaces, has constantly increased. Impact of indoor environment quality on schools is higher than its impact on offices. Schools are important for educating future generations and green schools pave the way for a sustainable green community.
Moreover, standards of governments are not binding and also have complexities to build green buildings. The main purpose of the current study is to create a framework based on Iran's requirements and regulations, which users can easily evaluate their design alternatives at the early stages of design.
A Dataset is generated by simulating 880555 different configurations for four room types: class, office, library, and gym. Models are configured with valid intervals for the rotation angle, WWR, shading depth, wall, window and roof U-values, and natural and mechanical ventilation. The inputs of simulations are used from ASHRAE and Iran’s regulations in order to create comfortable conditions for different cold, hot, and humid climates. Outputs of the simulations include the study of energy and long-term thermal comfort based on 99 indicators, which 8 indicators are related to energy and 6 indicators are related to thermal comfort. The outputs are computed by Honeybee and EnergyPlus. We employ machine learning algorithms on our dataset to obtain an optimal prediction model. In addition, an unseen dataset with 9521 datapoints was used to assess the performance of predicting models.
Different machine learning algorithms are used to develop a suitable prediction model. Based on the experiments, artificial neural networks and support vector regression models demonstrate superior performance on prediction of different labels. On average, the NN and SVR models perform with 83 and 13 on cross-validation data respectively.
We hope that the use of these methods will lead to the production of online software that does not require an accurate modeling or an expert. By considering the standards, the results of analyzing alternatives in the early stage of design, will be offered in a short time and acceptable accuracy. In addition, the users can choose the spaces themselves and form geometry of school. This tool will guide the designer towards the green school.
Buildings tend to not operate as intended, and a pronounced gap often exists between measured and predicted environmental and energy performance. Although the causes of this ‘performance gap’ are multi-faceted, issues surrounding data integration are key contributory factors. The distributed nature of the Architecture, Engineering and Construction (AEC) industry presents many challenges to the effective capture, integration and assessment of building performance data. Not all building data can be described semantically, nor is it feasible to create adapters between many different software tools. Similarly, not all building contextual data can easily be captured in a single product-centric model.
Buildings have in recent years been the target of a number of energy efficiency improvement strategies given that they are a major energy end-use sector in most countries. Whilst new buildings due to legislations, increasingly address sustainability and improved energy efficiency considerations, the refurbishment process of older buildings still presents a number of challenges. Advancement in Information and Communication Technology, particularly the application of low-cost Wireless Sensors and Actuators Network does however provide the opportunity to harness yet unrealized energy reduction in existing buildings. This paper presents results from an experimental study evaluating the performance and energy saving potentials of such off-the-shelve, low-cost wireless sensors and actuators network in an existing office building for occupancy detection and occupancy-driven lighting control. The study demonstrates that in addition to improved occupancy information obtainable from Wireless Sensors and Actuators Network, worthwhile savings in the energy consumption of the lighting systems can as well be achieved.
The purposes of this research are to find a model to forecast the electricity consumption in a household and to find the most suitable forecasting period whether it should be in daily, weekly, monthly, or quarterly. The time series data in our study was individual household electric power consumption from December 2006 to November 2010. The data analysis has been performed with the ARIMA (Autoregressive Integrated Moving Average) and ARMA (Autoregressive Moving Average) models. The suitable forecasting methods and the most suitable forecasting period were chosen by considering the smallest value of AIC (Akaike Information Criterion) and RMSE (Root Mean Square Error), respectively. The result of the study showed that the ARIMA model was the best model for finding the most suitable forecasting period in monthly and quarterly. On the other hand, ARMA model was the best model for finding the most suitable forecasting period in daily and weekly. Then, we calculated the most suitable forecasting period and it showed that the method were suitable for short term as 28 days, 5 weeks, 6 months and 2 quarters, respectively.
Data mining has become an essential tool during the last decade to analyze large sets of data. The variety of techniques it includes and the successful results obtained in many application fields, make this family of approaches powerful and widely used. In particular, this work explores the application of these techniques to time series forecasting. Although classical statistical-based methods provides reasonably good results, the result of the application of data mining outperforms those of classical ones. Hence, this work faces two main challenges: (i) to provide a compact mathematical formulation of the mainly used techniques; (ii) to review the latest works of time series forecasting and, as case study, those related to electricity price and demand markets.
Predicting short term electricity load accurately is critical to facilitate demand side management in the building sector. For buildings that have electricity sub-metering systems installed, it is possible to predict both the total electricity load and the loads of individual building service systems (air conditioning, lighting, power, and other equipment). In this paper, a method based on Support Vector Machine (SVM) is proposed to predict the loads at system level. For each type of system, 24 SVM models (one model per hour) were trained and deployed to predict the hourly electricity load. The inputs for the prediction method are simply weather predictions and hourly electricity loads in two previous days. A case study shows that the proposed method outperforms three other popular data mining methods (ARIMAX, Decision Tree, and Artificial Neural Network) in both CV_RMSE and N_MBE. Thus, SVM method is suggested for predicting system level electricity loads of public buildings.
The aim of this study is to provide a precise model for one-month-ahead forecast of electricity demand in residential sector. In this study, a total of 20 influential variables are taken into account including monthly electricity consumption, 14 weather variables, and 5 social variables. Based on support vector regression and fuzzy-rough feature selection with particle swarm optimization algorithms, the proposed method established a model with variables that relate to the forecast without ignoring some of these variables one may inevitably lead to forecasting errors. The proposed forecasting model was validated using historical data from South Korea. Its time period was from January 1991 to December 2012. The first 240 months were used for training and the remaining 24 for testing. The performance was evaluated using MAPE, MAE, RMSE, MBE, and UPA values. Furthermore, it was compared with that obtained from the artificial neural network, auto-regressive integrated moving average, multiple linear regression models, and the methods proposed in the previous studies, and found superior for every performance measure considered in this study. The proposed method has an advantage over the previous methods because it automatically determines appropriate and necessary variables for a reliable forecast. It is expected that the proposed model can contribute to more accurate forecasting of short-term electricity demand in residential sector. The ability to accurately forecast short-term electricity demand can assist power system operators and market participants in ensuring sustainable electricity planning decisions and secure electricity supply to the consumers.
This paper tests and compares two types of modelling to predict the same time series. A time series of electric load was observed and, as a case study, we opted for the metropolitan region of Bahia State. The combination of three exogenous variables were attempted in each model. The exogenous variables are: the number of customers connected to the electricity distribution network, the temperature and the precipitation of rain. The linear model time series forecasting used was a SARIMAX. The modelling of computational intelligence used to predict the time series was a Fuzzy Inference System. According to the evaluation of the attempts, the Fuzzy forecasting system presented the lowest error. But among the smallest errors, the results of the attempts also indicated different exogenous variables for each forecast model.
The importance of the residential class in the consumption of electricity in the Brazilian Electric System (BES) can be recognized by its quantitative size, as it, in 2013, concentrates 27% of the total consumption and 85% among all consumers. Also, in this class are the main public policies such as subsidies for consumer units inhabited by low-income families, labelling and increased energy efficiency of appliances used in the home and others. This paper aims to model and forecast the Brazilian residential energy consumption, up to 2050, with Pegels exponential smoothing techniques. In addition to the forecasts with the best model in sample, an optimization procedure of the model's hyper parameters is carried out in order to adjust the projections provided by the Energy Research Company (ERC). The results obtained show that it is possible to predict satisfactorily the electricity consumption for the proposed horizon with minimum error in sample. And the exercise of optimization proved to be important for providing level and trend equations for the official expectations regarding the residential electricity consumption in Brazil.
As a popular data driven method, artificial neural networks (ANNs) have been widely applied in building energy prediction field for decades. To improve the short term prediction accuracy, this paper presents a kind of optimized ANN model for hourly prediction of building electricity consumption. An improved Particle Swarm Optimization algorithm (iPSO) is applied to adjust ANN structure's weights and threshold values. The principal component analysis (PCA) is used to select the significant modeling inputs and simplify the model structure. The investigation utilizes two different historical data sets in hourly interval, which are collected from the Energy Prediction Shootout contest I and a campus building located in East China. For performance comparison, another two prediction models, ANN model and hybrid Genetic Algorithm-ANN (GA-ANN) model are also constructed in this study. The comparison results reveal that both iPSO-ANN and GA-ANN models have better accuracy than that of ANN ones. From the perspective of time consuming, the iPSO-ANN model has shorter modeling time than GA-ANN method. The proposed prediction model can be thought as an alternative technique for online prediction tasks of building electricity consumption.
Electrical load forecasting plays a vital role in order to achieve the concept of next generation power system such as smart grid, efficient energy management and better power system planning. As a result, high forecast accuracy is required for multiple time horizons that are associated with regulation, dispatching, scheduling and unit commitment of power grid. Artificial Intelligence (AI) based techniques are being developed and deployed worldwide in on Varity of applications, because of its superior capability to handle the complex input and output relationship. This paper provides the comprehensive and systematic literature review of Artificial Intelligence based short term load forecasting techniques. The major objective of this study is to review, identify, evaluate and analyze the performance of Artificial Intelligence (AI) based load forecast models and research gaps. The accuracy of ANN based forecast model is found to be dependent on number of parameters such as forecast model architecture, input combination, activation functions and training algorithm of the network and other exogenous variables affecting on forecast model inputs. Published literature presented in this paper show the potential of AI techniques for effective load forecasting in order to achieve the concept of smart grid and buildings.
This study presents building energy forecasting methodology for cooling load for three institutional buildings. These buildings belong to a university campus in Singapore. The daily energy consumption for cooling load is obtained for a period of two years and the daily variation is analyzed. The energy consumption is initially divided into five classes and the class numbers are used as inputs to develop a forecasting model. The model is developed using two machine learning tools. The tools used are Artificial Neural Network (ANN) and Adaptive Neuro Fuzzy Interface System (ANFIS). The division of data for training and testing the model is nearly 60% and 40% respectively. The results show that both ANN and ANFIS forecast the cooling load energy consumption of the three buildings with good accuracy. The correlation coefficient between measured and predicted consumption for training data are well above 0.98. The same is well above 0.96 for testing data. It is noted that such a methodology can be positively extended to other institutional buildings in the campus.
Short-term load forecasting (STLF) plays a decisive role in electric power system operation and planning. Accurate load forecasting not only reduces the generation costs of power systems, but also serves to maximize profit for participants in electricity markets. In recent years, power markets have grown more deregulated and competitive, adding to the complexity and uncertainties of load, and making it more difficult for conventional techniques to accurately forecast the load. To improve the accuracy of load forecasting, this paper suggests a hybrid method, called Gray-Fuzzy-Markov Chain Method (GFMCM), comprising three stages. In the first stage, daily load is forecasted by Gray model, with its training deviations classified, in a second stage, by fuzzy-set theory, and finally, fed into Markov chain model to predict future relative errors that might be supplied by the Gray model. The proposed approach has been verified by the historical data of power consumption in Ontario, PJM and Iranian electricity markets. The obtained forecasts by GFMCM proved to have better prediction properties compared to the other forecasting techniques, such as Gray models, specifically GM(1,1) and GM(1,2), ARIMA time series, wavelet-ARIMA and multi-layer perceptron (MLP) neural network.
This study proposes using a random forest model for short-term electricity load forecasting. This is an ensemble learning method that generates many regression trees (CART) and aggregates their results. The model operates on patterns of the time series seasonal cycles which simplifies the forecasting problem especially when a time series exhibits nonstationarity, heteroscedasticity, trend and multiple seasonal cycles. The main advantages of the model are its ability to generalization, built-in cross-validation and low sensitivity to parameter values. As an illustration, the proposed forecasting model is applied to historical load data in Poland and its performance is compared with some alternative models such as CART, ARIMA, exponential smoothing and neural networks. Application examples confirm good properties of the model and its high accuracy.
Demand planning for electricity consumption is a key success factor for the development of any countries. However, this can only be achieved if the demand is forecasted accurately. In this research, different forecasting methods-autoregressive integrated moving average (ARIMA), artificial neural network (ANN) and multiple linear regression (MLR)-were utilized to formulate prediction models of the electricity demand in Thailand. The objective was to compare the performance of these three approaches and the empirical data used in this study was the historical data regarding the electricity demand (population, gross domestic product: GDP, stock index, revenue from exporting industrial products and electricity consumption) in Thailand from 1986 to 2010. The results showed that the ANN model reduced the mean absolute percentage error (MAPE) to 0.996%, while those of ARIMA and MLR were 2.80981 and 3.2604527%, respectively. Based on these error measures, the results indicated that the ANN approach outperformed the ARIMA and MLR methods in this scenario. However, the paired test indicated that there was no significant difference among these methods at α = 0.05. According to the principle of parsimony, the ARIMA and MLR models might be preferable to the ANN one because of their simple structure and competitive performance.
Electricity load forecasting plays a key role in operation of power systems. Since the penetration of distributed and renewable generation is increasingly growing in many countries, Short-Term Load Forecast (STLF) of micro-grids is also becoming an important task. A precise STLF of the micro-grid can enhance the management of its renewable and conventional resources and improve the economics of energy trade with electricity markets. As a consequence of the highly non-smooth and volatile behavior of the load time series in a micro-grid, its STLF is even a more complex process than that of a power system. For this purpose, a new prediction method is proposed in this paper, in which a Self-Recurrent Wavelet Neural Network (SRWNN) is applied as the forecast engine. Moreover, the Levenberg-Marquardt (LM) learning algorithm is implemented and adapted to train the SRWNN. In order to demonstrate the efficiency of the proposed method, it is examined on real-world hourly data of an educational building within a micro-grid. Comparisons with other load prediction methods are provided.
One of the largest threats to future development is climate change. Apparently, the building sector has been the largest source of greenhouse gas production. One prospective solution to this is “green building” that aims to provide environmentally sustainable building in terms of design, construction and maintenance. However, green buildings symbolize the next stage of buildings, and the recent growth of new green building constructions is inadequate to overcome the negative impact of existing buildings. One logical solution to reduce the environmental impact of the existing buildings is through green retrofitting. Yet there is lack of systematic review on the existing body of knowledge on green retrofitting which is critical for future research. This paper aims to critically review the existing literature on green retrofitting and to identify contemporary research trends. Additionally, with the view to the current challenges, barriers, obstacles or problems to green retrofitting, this study highlights the needs to identify the Critical Success Factors (CSFs) for successful implementation of green retrofit projects.
It is important for building owners and operators to manage the electrical energy consumption of their buildings. As electrical energy is the major form of energy consumed in a commercial building, the ability to forecast electrical energy consumption in a building will bring great benefits to the building owners and operators. This paper provides a review of the building electrical energy consumption forecasting methods which include the conventional and artificial intelligence (AI) methods. The significant goal of this study is to review, recognize, and analyse the performance of both methods for forecasting of electrical energy consumption. Compared to using a single method of forecasting, the hybrid of two forecasting methods can possibly be applied for more precise results. Regarding this potential, the swarm intelligence (SI) method has been reviewed to be hybridized with AI. Published literature presented in this paper shows that, the hybrid of SVM and SI methods has indeed presented superior performance for forecasting building electrical energy consumption.
The pressing global environmental issues are fostering a rapid change in the energy and sustainability policies for the built environment. New paradigms are emerging, such as “Nearly Zero Energy Building” (nZEB), and resource efficiency is progressively becoming a crucial topic in the building sector, implying an appropriate consideration of performance over the whole life cycle. However, empirical evidences show how, very often, the gap between the predicted (design phase) and measured (operation phase) performance is very large, due to errors committed during all the phases of building life cycle. This performance gap determines a problem of credibility in the building industry and, more in general, in sustainability oriented practices. Therefore, design and operation practices should evolve in order to be able to cope with performance uncertainty determined, for example, by evolution of climate conditions, variability of behavioural patterns and performance degradation of technological components.
Building energy use prediction plays an important role in building energy management and conservation as it can help us to evaluate building energy efficiency, conduct building commissioning, and detect and diagnose building system faults. Building energy prediction can be broadly classified into engineering, Artificial Intelligence (AI) based, and hybrid approaches. While engineering and hybrid approaches use thermodynamic equations to estimate energy use, the AI-based approach uses historical data to predict future energy use under constraints. Owing to the ease of use and adaptability to seek optimal solutions in a rapid manner, the AI-based approach has gained popularity in recent years. For this reason and to discuss recent developments in the AI-based approaches for building energy use prediction, this paper conducts an in-depth review of single AI-based methods such as multiple linear regression, artificial neural networks, and support vector regression, and ensemble prediction method that, by combining multiple single AI-based prediction models improves the prediction accuracy manifold. This paper elaborates the principles, applications, advantages and limitations of these AI-based prediction methods and concludes with a discussion on the future directions of the research on AI-based methods for building energy use prediction.
In this paper, we focus on the prediction method of building energy consumption time series. The building energy consumption data can be regarded as a time series, which is usually nonlinear and non-stationary. Traditional time series analysis model has lower prediction accuracy. Then the machine learning method, especially support vector regression algorithm always has better performance to deal with non-stationary and nonlinear time series. So the support vector regression algorithm is applied to develop building energy consumption time series model. The model is applied in different buildings. Experimental results show the prediction accuracy of the model is better than traditional time series analysis model.
The intelligent building, a major smart city research and development domain, has grown beyond the scope of automation, currently focusing on the occupant centered approaches and the ancillary services offered to the power grid. The Internet of Things (IoT) introduced in the recent years are enabling those prospects. However, due to numerous legacy automation systems already in the construction market, the IoT based intelligent building approach is frequently challenged. This paper assesses the opportunities along with the criticism for a fully IoT enabled and controllable intelligent building against the well established, legacy automation systems in a fair and transparent approach. In the second part, it proposes an interoperable intelligent building design for the creation of advanced building managements schemes, by integrating the assets of current automation tools and the emerging innovations. This unification and the progressive phasing out of legacy technologies, would catalyze the market of intelligent buildings in the context of smart cities by not limiting their potential solely on new constructions and by mitigating the investment if legacy systems are already installed.
Building intelligent system has become one of the standard configurations in a public building even in a residential building. Intelligent buildings are expected to provide high-efficient, convenient, comfortable and energy-saving built environment so as to bring additional benefits. However, problems like sensor faults and control strategy flaws may result in low performance and high energy consumptions and maintenance costs are needed as well. Thus it is needed to synthetically evaluate the costs and benefits of intelligent systems to help building investors to make decisions. This paper proposes a cost-benefit evaluation method for building intelligent systems. Life cycle net present value (NPV) of all the costs and benefits, including tangible and intangible, is used as an index to evaluate the performance of the building intelligent systems. The proposed method puts emphases on the quantitative evaluation of intangible benefits using Analytical Hierarchy Process (AHP) method to determine the weights of each aspects of intangible benefit. The building energy consumptions are taken as an another emphasis as well. The proposed method is applied to evaluate three actual buildings to check its feasibility and effectiveness. Sensitivity analysis of NPV to energy consumption of the case study buildings shows that the energy consumption increase or decrease caused by different performances of intelligent systems can influence the positive or negative of NPV, which implies the building intelligent systems are worth of investment or not.
Electricity load forecasting is crucial for effective operation and management of buildings. Support Vector Regression (SVR) have been successfully used in solving nonlinear regression and time series problems related to building energy consumption forecasting. As the performance of SVR heavily depends on the selection of its parameters, differential evolution (DE) algorithm is employed in this study to solve this problem. The forecasting model is developed using weighted SVR models with nu-SVR and epsilon-SVR. The DE algorithm is again used to determine the weights corresponding to each model. A case of time series energy consumption data from an institutional building in Singapore is used to elucidate the performance of the proposed model. The proposed model can be used to forecast both, half-hourly and daily electricity consumption time series data for the same building. The results show that for half-hourly data, the model exhibits higher weight for nu-SVR, whereas for daily data, a higher weight for epsilon-SVR is observed. The mean absolute percentage error (MAPE) for daily energy consumption data is 5.843 and that for half-hourly energy consumption is 3.767 respectively. A detailed comparison with other evolutionary algorithms show that the proposed model yields higher accuracy for building energy consumption forecasting.
Household air pollution is ranked the 9th largest Global Burden of Disease risk (Forouzanfar et al., The Lancet 2015). People, particularly urban dwellers, typically spend over 90% of their daily time indoors, where levels of air pollution often surpass those of outdoor environments. Indoor air quality (IAQ) standards and approaches for assessment and control of indoor air require measurements of pollutant concentrations and thermal comfort using conventional instruments. However, the outcomes of such measurements are usually averages over long integrated time periods, which become available after the exposure has already occurred. Moreover, conventional monitoring is generally incapable of addressing temporal and spatial heterogeneity of indoor air pollution, or providing information on peak exposures that occur when specific indoor sources are in operation. This article provides a review of the new air pollution sensing methods to determine IAQ and discusses how real-time sensing could bring a paradigm shift in controlling the concentration of key air pollutants in billions of urban houses worldwide. However, we also show that besides the opportunities, challenges still remain in terms of maturing technologies, or data mining and their interpretation. Moreover, we discuss further research and essential development needed to close gaps between what is available today and needed tomorrow. In particular, we demonstrate that awareness of IAQ risks and availability of appropriate regulation are lagging behind the technologies.
Traditionally, short-term electricity price forecasting has been essential for utilities and generation companies. However, the deregulation of electricity markets created a competitive environment and the introduction of new market participants, such as the retailers and aggregators, whose economic viability and profitability highly depends on the spot market price patterns. The aim of this study is to examine artificial neural network (ANN) based models for Day-ahead price forecasting. Specifically, the models refer to the sole application of ANNs or to hybrid models, where the ANN is combined with clustering algorithm. The training data are clustered in homogenous groups and for each cluster, a dedicated forecaster is employed. The proposed models are characterized by comprehensive operation and by high level of flexibility; different inputs can be taken under consideration and different ANN topologies can be examined. The models are tested on a data set that consists of atypical price patterns and many outliers. This approach makes the price forecasting problem a more challenging task, providing evidence that the proposed models can be considered as useful and robust forecasting tools to the actual needs of market participants, including the traditional generation companies and self-producers, but also the retailers/suppliers and aggregators.
A mixture of support vector machines (SVMs) is proposed for time series forecasting. The SVM mixture is composed of a two-stage architecture. In the first stage, self-organizing feature map (SOM) is used as a clustering algorithm to partition the whole input space into several disjointed regions. A tree-structured architecture is adopted in the partition to avoid the problem of predetermining the number of partitioned regions. Then, in the second stake, multiple SVMs, also called SVM mixture, that best fit partitioned regions are constructed by finding the most appropriate kernel function and the optimal free parameters of SVMs. The experiment shows that the SVMs mixture achieves significant improvement in the generalization performance in comparison-with the single SVMs model. In addition, the SVMs mixture also converges faster and use fewer support vectors.
Titanium alloys such as TÎ6A14V have been popular choices as bulk materials for loadbearing orthopedic implants due to their resistance to corrosion and excellent biocompatibility. To enhance the biological properties and wear resistance of TÎ6A14V, samples of the titanium alloy were tested against titanium composites formed with TÎ6A14V with 2 and 5wt. % tricalcium phosphate. Results show that the calcium phosphate tends to form along the grain boundaries of the titanium alloy matrix, which causes the surface hardness to rise from an average of 177 HV to as high as 542 HV. The increased strength of the material also significantly reduced the amount of material removed during wear testing and preserved a relatively smooth wear track surface to reduce frictional forces from shear. Contact angle measurements with distilled water decreased as the concentration of TCP increased, implying greater hydrophilicity with the composite materials. Simulated body fluid testing for three days show that the presence of TCP in the material accelerated the growth rate of apatite crystals. The results from the contact angle measurements along with the SBF study show that the TCP composite can enhance biological responses of the composites within the body.
This paper study the used of moving averages (MA) and Exponential smoothing techniques (EST) for load forecasting. The case study was in Universiti Teknologi PETRONAS (UTP),Malaysia. The study was divided by two types of load forecasting namely Semester On (SOn)and Semester Off (SOf). Later, MA and ESMT being used to forecast the usage load for bothSOn and SOf. The results indicated that ESMT gives better forecasting compared to MA in terms of less measurements of error e.g. Mean Absolute Percentage Error (MAPE).
The enhancement of load forecasting has become one of the core research topics in the energy field. Because power load has both time-variant and nonlinear characteristics, different types of methods, neural networks (NN) in particular, have been applied to power load forecasting. This study proposes a real-valued genetic algorithm (RGA)- based neural network with support vector machine (NN-SVM) model to predict the power load in both short-term and mid-term forecasting by using a radial-basis-function neural network (RBFNN), SVM and RGA. The model consists of two stages. In short-term load forecasting (STLF), the first stage applies the RBFNN to predict monthly variations, and the second stage trains the SVM through hourly data to obtain the final forecast. Similar operations are used in mid-term load forecasting (MTLF). In the process of SVM training and NN learning, RGA is used to find the optimal parameters. The results of several experiments show that this new model performs more accurately and stably than some conventional models including RBFNN, RGA-SVM, Karman filter in STLF. Also it is able to function well in MTLF.
This paper presents a detailed investigation and analysis of the energy consumption characteristics of three institutional buildings in Singapore. Building information, energy consumption data of the air-conditioning system, and energy consumption data of the plug loads were collected separately. Identification models are developed to predict the real daily energy consumption data. Developed models involve three specific functions to represent the variability of the daily occupancy, the additional occupancy due to visitors and the variation of outdoor air temperature. The performance of the developed identification model is very satisfactory and fits very well with the real energy consumption data. Based on the identification model, key factors which are influencing the energy consumption in institutional buildings are identified as the variability of the daily occupancy, hence a methodology is developed to calculate the occupancy in the building. The identified parameters are used as inputs into deterministic energy simulation programs, like Energy Plus, to perform detailed energy analysis. The whole methodology developed has improved significantly the accuracy of the energy simulation modelling of institutional buildings and has permitted to understand and evaluate the major energy characteristics of these buildings.
In this paper five exponential smoothing methods are considered for load forecasting for lead times from a half-hour-ahead to a year-ahead. Forecasting load demand with high accuracy is required to prevent energy wasting and system failure. For this purpose, a half-hourly load demand of Malaysia for one year, from September 01, 2005 to August 31, 2006 was used. The mean absolute percentage error (MAPE) is used for comparing the forecasting accuracy. Time series of load demand recorded at half-hourly intervals contains more than one seasonal pattern, which are the intraday and intraweek seasonal cycles. The forecasts produced by the Holt-Winters Taylor (HWT) exponential smoothing outperformed the traditional Holt-Winters and modified Holt-Winters exponential smoothing methods.
Short-term load forecasting of building electricity usage is of great importance for anomaly detection on electricity usage pattern and management of building energy consumption in an environment where electricity pricing is dynamically determined based on the peak energy consumption. In this paper, we present a data-driven forecasting model for day-ahead electricity usage of buildings in 15-minute resolution.
By using variable importance analysis, we have selected key variables: day type indicator, time-of-day, HVAC set temperature schedule, outdoor air dry-bulb temperature, and outdoor humidity as the most important predictors for electricity consumption. This study proposes a short-term building energy usage forecasting model based on an Artificial Neural Network (ANN) model with Bayesian regularization algorithm and investigates how the network design parameters such as time delay, number of hidden neurons, and training data effect on the model capability and generality.
The results demonstrate that the proposed model with adaptive training methods is capable to predict the electricity consumption with 15-minute time intervals and the daily peak electricity usage reasonably well in a test case of a commercial building complex.
Electric load forecasting is an important issue for power utility, associated with the management of daily operations such as energy transfer scheduling, unit commitment, and load dispatch. Inspired by strong non-linear learning capability of support vector regression (SVR), this paper presents a SVR model hybridized with the differential empirical mode decomposition (DEMD) method and auto regression (AR) for electric load forecasting. The differential EMD method is used to decompose the electric load into several detail parts associated with high frequencies (intrinsic mode function (IMF)) and an approximate part associated with low frequencies. The electric load data from the New South Wales (NSW, Australia) market and the New York Independent System Operator (NYISO, USA) are employed for comparing the forecasting performances of different alternative models. The results illustrate the validity of the idea that the proposed model can simultaneously provide forecasting with good accuracy and interpretability.
Long-term power load forecasting is of major importance for power suppliers to define the future power consumption of a given region. However, it is not easy to contend with the uncertainty of the long-term load. In order to effectively forecast the long-term load, a collaborative principal component analysis and fuzzy feed-forward neural network (PCA-FFNN) approach is proposed in this study. The difference between this and existing methods is that the collaborative PCA-FFNN approach takes into account the different points of view in a more efficient way, and therefore the results obtained are more comprehensive and more in-depth. In the proposed methodology, a group of domain experts is formed. These domain experts are asked to configure their own PCA-FFNNs to forecast the long-term load based on their views. A collaboration mechanism is therefore established. To facilitate the collaboration process and to derive a single representative value from these forecasts, the partial-consensus fuzzy intersection and radial basis function network (PCFI-RBF) approach is used. The effectiveness of the proposed methodology is illustrated with a case study.