Search

Short-term load forecasting (STLF) of heterogeneous multi-agents plays a significant role in smart grid. Faced with special difficulties of multi-agent STLF due to the high heterogeneity, uncertainty and volatility, traditional local methods usually make predictions based on load aggregation or clustering, the complexity of which will rise greatly with the increase of agent types. It has become more and more difficult for traditional methods to meet the STLF demand of smart grid. Meanwhile, global time series forecasting method emerges gradually in many fields, which can make predictions for many agents only by one model with much lower complexity. However, there is few researches on the multi-agent global STLF. This study proposes a global deep-Long Short-Term Memory (LSTM) STLF model based on a pretraining method. The model fully applies the information such as electricity consumption pattern, weather and calendar in a structural and orderly way. The empirical results show that our model can effectively predict the daily load of heterogeneous households, with prediction accuracies of 87.9–90.2% on the test set across various tasks. Compared with the base model, our model achieves an 8.7% higher accuracy with a much faster convergence speed. Furthermore, the fluctuations of accuracy on different tasks are within 2.3%, showing that our model is robust. The superiority of the global method is proved through the comparison with local method in the empirical results. This study is expected to make contribution to global STLF of heterogeneous households and providing experience for the global prediction of heterogeneous time series based on deep-LSTM and pretraining method. [ABSTRACT FROM AUTHOR]

An accurate short-term load forecasting plays an important role in modern power system's operation and economic development. However, short-term load forecasting is affected by multiple factors, and due to the complexity of the relationships between factors, the graph structure in this task is unknown. On the other hand, existing methods do not fully aggregating data information through the inherent relationships between various factors. In this paper, we propose a short-term load forecasting framework based on graph neural networks and dilated 1D-CNN, called GLFN-TC. GLFN-TC uses the graph learning module to automatically learn the relationships between variables to solve problem with unknown graph structure. GLFN-TC effectively handles temporal and spatial dependencies through two modules. In temporal convolution module, GLFN-TC uses dilated 1D-CNN to extract temporal dependencies from historical data of each node. In densely connected residual convolution module, in order to ensure that data information is not lost, GLFN-TC uses the graph convolution of densely connected residual to make full use of the data information of each graph convolution layer. Finally, the predicted values are obtained through the load forecasting module. We conducted five studies to verify the outperformance of GLFN-TC. In short-term load forecasting, using MSE as an example, the experimental results of GLFN-TC decreased by 0.0396, 0.0137, 0.0358, 0.0213 and 0.0337 compared to the optimal baseline method on ISO-NE, AT, AP, SH and NCENT datasets, respectively. Results show that GLFN-TC can achieve higher prediction accuracy than the existing common methods. [ABSTRACT FROM AUTHOR]

Short-term electricity load forecasting plays a crucial role in modern container terminal. In this work, we design a short-term forecasting approach aimed at port load under the framework of patch learning. Firstly, singular spectrum analysis is applied to obtain denoised and noise features, respectively; then, a patch learning model based on the long short-term memory network is employed to address such a time-series forecasting problem. LSTM network and BiLSTM are considered as the global models to process denoised and noisy data, respectively, and convolutional neural network is selected as the patch model. Furthermore, an endpoint detection strategy is designed for adaptively identifying the positions of patches. The performance of the proposed model is tested and verified on a real Chinese container terminal load dataset. Experimental results show that the proposed approach, compared with state-of-the-art load forecasting models, has the greatest performance with respect to seven evaluation criteria. [ABSTRACT FROM AUTHOR]

In this paper, a novel swarm intelligence optimization algorithm is presented based on the sparrow search algorithm, namely, an intensified sparrow search algorithm (ISSA). Specifically, the newly proposed neighbor search strategy takes into account both exploring the entire feasible solution space as much as possible in the iterative process and effectively preventing the weakening of exploration ability in the late iteration. In addition, a new foraging method called the saltation learning strategy is put forward to improve the search capability of the scrounger. Firstly, the effectiveness of the ISSA is comprehensively evaluated by competition functions, known as CEC-BC-2017. The simulation results show that the ISSA substantially improves the convergence accuracy of the basic SSA and also outperforms three SSA-based variants and six state-of-art optimization algorithms. Then, to further demonstrate the real-world application potential, the ISSA is successfully used in two engineering design problems (including the pressure vessel and the welded beam designs). Finally, the proposed ISSA is employed to optimize the hyper-parameters of the long short-term memory (LSTM) network, which leads to a novel ISSA-LSTM model. The developed ISSA-LSTM model is applied to the short-term load forecasting of the power system. The experimental results show that the mean absolute percentile error (MAPE), root mean square error (RMSE) and mean absolute error (MAE) values of the proposed ISSA-LSTM model are 1.2778%, 1.2171 and 0.9267 respectively, which is superior to several LSTM-based variants. [ABSTRACT FROM AUTHOR]

Accurate short-term load forecasting is crucial for the steady operation of the power system and power market schedule planning. The extraction of features and training of prediction models are challenging as the load series is extremely volatile and nonlinear. To address the above issues, we propose a deep bidirectional long short-term memory (DBiLSTM) network based on variational mode decomposition (VMD) and an attention mechanism, in which the model hyperparameters are optimized using the improved particle swarm optimization (IPSO) technique. In this study, the mode number k of the VMD is determined by the ratio of residual energy following decomposition. Subsequently, the DBiLSTM is stacked using multiple layers of BiLSTM for a more precise representation of time-series data and the capturing of information at different scales, thereby enabling nonlinear load sequence forecasting and enhancing the accuracy. Finally, the IPSO uses nonlinear decreasing inertia weights to overcome the drawbacks of premature convergence and local optima. The effectiveness and progress of the proposed method are evaluated using the power load dataset from the ninth electrical attribute modeling competition test questions. [ABSTRACT FROM AUTHOR]

Power load forecasting is an important part of smart grid, and its accuracy will directly affect the control and planning of power system operation. In the context of electricity market reform, real-time electricity prices affect users' electricity consumption patterns. A short-term load forecasting model based on support vector regression (SVR) with whale optimization algorithm (WOA) considering real-time electricity price is proposed in this paper. Meta-heuristics are very promising in optimizing the parameters of SVR, and the WOA algorithm is used to determine the appropriate combination of SVR's parameters to accurately establish a forecasting model. The initial value of the original WOA algorithm lacks ergodicity, and has defects such as easy to fall into local optimum and low convergence accuracy. Chaos mechanism and elite opposition-based learning strategy are introduced into WOA to balance the exploration and exploitation of the algorithm and improve the algorithm convergence speed. Numerical examples involving two power load datasets show that the proposed model can achieve better forecasting performance in comparison with other models, such as SVR, BPNN. At the same time, it proves that the forecasting accuracy with electricity price is higher than that without electricity price. [ABSTRACT FROM AUTHOR]

Accurate electricity demand forecasting can provide a timely and effective reference for economic control and facilitate the secure production and operation of power systems. However, electricity data are well known for their nonlinearity and multi-seasonal features, making it challenging to construct forecasting models. This study investigates the combination of singular spectrum analysis to facilitate the construction of decomposition-based forecasting approaches for electricity load. First, we demonstrate and emphasize the importance of separability for specifically extracting different features hidden in the original data; moreover, only by using the separable feature subseries, the constructed individual model can capture the inner and distinct characteristics of original series more effectively. Second, this study decomposes the electricity load into several significant features using singular spectrum analysis. Each feature series is predicted separately to construct aggregate results. In particular, we propose SSA-based period decomposition to not only perform separable decomposition but also overcome the border effect, which has received little attention in previous work. Finally, to verify the effectiveness of the proposed method, we conduct an empirical study and compare the performance of the discussed models. The empirical results show that the proposed approach can obtain the expected forecasting performance and is a reliable and promising tool for extracting different features. [ABSTRACT FROM AUTHOR]

Energy systems face challenges due to climate change, distributed energy resources, and political agenda, especially distribution system operators (DSOs) responsible for ensuring grid stability. Accurate predictions of the electricity load can help DSOs better plan and maintain their grids. The study aims to test a systematic data identification and selection process to forecast the electricity load of Danish residential areas. The five-ecosystem CSTEP framework maps relevant independent variables on the cultural, societal, technological, economic, and political dimensions. Based on the literature, a recurrent neural network (RNN), long-short-term memory network (LSTM), gated recurrent unit (GRU), and feed-forward network (FFN) are evaluated and compared. The models are trained and tested using different data inputs and forecasting horizons to assess the impact of the systematic approach and the practical flexibility of the models. The findings show that the models achieve equal performances of around 0.96 adjusted R2 score and 4–5% absolute percentage error for the 1-h predictions. Forecasting 24 h gave an adjusted R2 of around 0.91 and increased the error slightly to 6–7% absolute percentage error. The impact of the systematic identification approach depended on the type of neural network, with the FFN showing the highest increase in error when removing the supporting variables. The GRU and LSTM did not rely on the identified variables, showing minimal changes in performance with or without them. The systematic approach to data identification can help researchers better understand the data inputs and their impact on the target variable. The results indicate that a focus on curating data inputs affects the performance more than choosing a specific type of neural network architecture. [ABSTRACT FROM AUTHOR]

Aiming at the problem of strong randomness and low forecasting accuracy in short-term electric load, a method based on empirical wavelet transform and random forest is proposed. In this method, the noise is extracted and stripped by wavelet transform, and the original data are decomposed into several groups of low or high frequencies. The empirical wavelet transform is more adaptive, and details are demonstrated accurately. These decomposed modes are used as characteristic variables to forecast with random forest. This method has three advantages: (1) Because of the instability of electric data, the empirical wavelet decomposition can be used to characterize the non-stationary signal characteristics. (2) Empirical wavelet transform has more advantages in time domain analysis because of its correlation to signal removal and the tendency of noise whitening. (3) More adaptive details with empirical mode (noise is no exception) are expressed in the random forest, and it still demonstrates accurate results even after data features are lost. The Australian electricity data in different periods are used for case analysis. The results compared with other methods have shown that the model could reveal effectively the influence of random noise and improve the accuracy and reliability of short-term load forecasting. [ABSTRACT FROM AUTHOR]

Short-term load forecasting plays a major role in energy planning. Its accuracy has a direct impact on the way power systems are operated and managed. We propose a new Clustering-based Similar Pattern Forecasting algorithm (CSPF) for short-term load forecasting. It resorts to a K-Medoids clustering algorithm to identify load patterns and to the COMB distance to capture differences between time series. Clusters' labels are then used to identify similar sequences of days. Temperature information is also considered in the day-ahead load forecasting, resorting to the K-Nearest Neighbor approach. CSPF algorithm is intended to provide the aggregate forecast of Portugal's national load, for the next day, with a 15-min discretization, based on data from the Portuguese Transport Network Operator (TSO). CSPF forecasting performance, as evaluated by RMSE, MAE and MAPE metrics, outperforms three alternative/baseline methods, suggesting that the proposed approach is promising in similar applications. [ABSTRACT FROM AUTHOR]

With the continuous development of smart grids, short-term power load forecasting has become increasingly important in the operation of power markets and demand-side management. In order to explore the influence of temperature and holidays on seasonal loads, this paper proposes a short-term SVM power load forecasting method based on K-Means clustering. The method includes the steps of selecting similar days, data preprocessing, SVM prediction model training and parameter adjustment. Among them, the selection of similar days uses K-Means to group seasonal load data into two categories according to temperature characteristics, as the input data to explore the effect of temperature on seasonal load. And divide the data into holidays and working days as the model input data to discover the impact of holidays on seasonal loads by using calendar rules. In order to verify the load forecasting effect of the proposed method, several experiments were carried out on two actual residential load data and two data online, and compared with the LSTM and decision tree load forecasting models in terms of prediction accuracy evaluation index and running time. The results show that the model constructed in this paper has 39.75% improved to the conventional methods for the accuracy and 128.89% improved for the running time. [ABSTRACT FROM AUTHOR]

The accuracy level of short-term load forecasting (STLF) affects the power department's arrangements for unit start-up, shutdown, overhaul, and load dispatching. However, the existing algorithms do not fully consider load volatility and difficulty in setting the algorithm parameters. In this regard, this paper proposes a hybrid model which combines complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), gated recurrent unit (GRU) with improved gray wolf optimizer (IGWO), namely CEEMDAN–IGWO–GRU (CIG) hybrid algorithm. Firstly, the details and trend information of the load signal are separated by the CEEMDAN algorithm to suppress the interference of load fluctuation. Then, the GRU network optimized by IGWO parameters is used to predict each component, separately. Finally, the complete load forecasting results are obtained by reconstructing the forecasting result of each component. The power load data of a certain area under study is used to verify the CIG model, and the experimental results are compared with other existing algorithms. The experimental results show that the load forecasting results of the CIG model get the high-precision evaluation of 0.6997%, 52.4685, and 38.1891 MW in MAPE, RMSE, and MAE, respectively. Therefore, the parameter optimization ability of the IGWO algorithm can effectively improve the prediction accuracy, and the CIG method can availably suppress the impact of load fluctuation on prediction and has a powerful nonlinear fitting ability. In conclusion, CIG has great potential in establishing a power load forecasting model. [ABSTRACT FROM AUTHOR]

Short-term load forecasting (STLF) is essential for developing reliable and sustainable economic and operational strategies for power systems. This study presents a forecasting model combining cascade forward neural network (CFNN) and intuitionistic fuzzy time series (IFTS) models for STLF. The proposed cascading intuitionistic fuzzy time series forecasting model (C-IFTS-FM) offers the advantage of CFNN using the links of both linear and nonlinear to model fuzzy relations between inputs and outputs. Moreover, it offers a more reliable and realistic approach to uncertainty, taking notice of also the degree of hesitation. C-IFTS-FM works in univariate structure when it uses only hourly load data, and in bivariate structure when it uses hourly load data and hourly temperature time series together. The conversion of time series into IFTS is realized with intuitionistic fuzzy c-means (IFCM). Thus, the membership and non-membership values for each data point are produced. In modelling process, membership and non-membership values, in addition to actual lagged observations, are used as input of the CFNNs. The effectiveness of C-IFTS-FM on test sets for both structures was discussed comparatively via different error criteria, in addition, the convergence time was examined, and also the fit of forecasts and observations was presented with different illustrations. Among different combinations of hyperparameters, in the best case, approximately 86% better accuracy is achieved than the best of the others, while even in the case of the worst of hyperparameters combination, the accuracy was improved by over 20% for the PSJM data sets. For HEXING, CHENGNAN, and EUNITE data sets, these progress rates reached approximately 90% in the best case. [ABSTRACT FROM AUTHOR]

Sustainable energy systems are characterised by an increased integration of renewable energy sources, which magnifies the fluctuations in energy supply. Methods to to cope with these magnified fluctuations, such as load shifting, typically require accurate short-term load forecasts. Although numerous machine learning models have been developed to improve short-term load forecasting (STLF), these models often require large amounts of training data. Unfortunately, such data is usually not available, for example, due to new users or privacy concerns. Therefore, obtaining accurate short-term load forecasts with little data is a major challenge. The present paper thus proposes the latent space-based forecast enhancer (LSFE), a method which combines transfer learning and data augmentation to enhance STLF when training data is limited. The LSFE first trains a generative model on source data similar to the target data before using the latent space data representation of the target data to generate seed noise. Finally, we use this seed noise to generate synthetic data, which we combine with real data to enhance STLF. We evaluate the LSFE on real-world electricity data by examining the influence of its components, analysing its influence on obtained forecasts, and comparing its performance to benchmark models. We show that the Latent Space-based Forecast Enhancer is generally capable of improving the forecast accuracy and thus helps to successfully meet the challenge of limited available training data. [ABSTRACT FROM AUTHOR]

Highly accurate short-term load forecasting (STLF) is essential in the operation of power systems. However, the existing predictive methods cannot achieve an effective balance between prediction accuracy and computational cost. Furthermore, the prediction residual is rarely used to improve the predictive accuracy in STLF. This paper proposes a novel decomposition-based ensemble model for the STLF task. First, an optimized empirical wavelet transform (OEWT) is developed to rationally decompose the STLF load by combining the approximate entropy method with the empirical wavelet transform. Particularly, OEWT improves both prediction accuracy and computational cost in STLF. Second, a new hybrid machine learning method (named master learner) is proposed by rationally combining long short-term memory networks (LSTMs) with broad learning system (BLS) in STLF, effectively strengthening the predictive accuracy without significantly increasing the computational cost. Third, a residual learning model (named residual learner) is developed in the master learner to extract the effective predictive information from residual results, further improving the prediction accuracy in STLF. Fourth, an auxiliary learner is proposed by introducing another BLS to connect the input and output of the proposed model, enhancing the predictive robustness. The proposed decomposition-based ensemble model is compared with state-of-the-art and traditional models in STLF. Experimental results show that the model not only has high predictive accuracy and robustness but also low computational cost. [ABSTRACT FROM AUTHOR]

Electrical load forecasting, mainly short-term load forecasting (STLF), plays a vital role in efficient power system planning by making it more intelligent, sustainable, and reliable. However, due to the presence of skewness and irregularities in the observed data, it becomes a challenging task to improve the accuracy of STLF. To handle this, we propose a new model, named Singular Spectrum Analysis-Long Short Term Memory (SSA-LSTM). SSA is used to remove the high-frequency components (high volatility data) from the load series. Based on the resultant series obtained from SSA, LSTM model forecasts the final load. We have used six publicly available datasets to evaluate the performance of our proposed model. With Kaggle Global Energy Forecasting Competition (GEFCom2012) dataset, the generated analytical results show 59.83% improvement over Tao's benchmark model and 12.46%, 12.48%, 4.17%, 5.42%, 5.68%, 3.28%, 4.73% improvement over refined multi-linear regression, Moving average and Exponential smoothing (MA & ES), Support Vector Regression (SVR) model, feed-forward neural network (FNN), deep belief network (DBN), LSTM and Singular Spectrum Analysis (SSA) respectively. Our model has outperformed MA & ES, SVR, FNN, DBN, ensemble deep learning method (EDBN), SSA and LSTM in South Australia, Tasmania, Queensland, and Victoria electrical load dataset by 11.38%, 23.54%, 3.05%, and 1.39% respectively using RMSE error metric. [ABSTRACT FROM AUTHOR]

Short-term load forecasting with high accuracy is essential to power systems. Because power loads involve high volatility and uncertainty, it is challenging to accurately perform short-term load forecasting (STLF). To solve this problem, this paper proposes a decomposition-based approximate entropy cooperation long short-term memory (DB-AEC-LSTM) model for STLF. In DB-AEC-LSTM, a complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) method is first introduced to generate the multiple electric load time series into many cooperation sub-series and decrease the reconstruction errors. Then, an Approximate Entropy Cooperation ensemble Long Short-term Memory Model is developed by using approximate entropy (ApEn) to construct an effective cooperative relationship between different time sub-series groups, greatly improving the predictive accuracy. By rationally combined the effective technologies ApEn, CEEMDAN, and AEC-LSTM, the proposed DB-AEC-LSTM can obtain competitive predictive performance in STLF. Several short-term load forecasting datasets are performed to check the predictive performance of DB-AEC-LSTM. Experimental results show that DB-AEC-LSTM has better predictive accuracy and satisfactory robustness compared with state-of-the-art and conventional predictive models. [ABSTRACT FROM AUTHOR]

A short-term electrical load forecasting model is proposed in this work. The proposed model is based on independent component analysis (ICA), discrete wavelet transform, Clustering, analysis of variance (ANOVA), and support vector regression (SVR) (called as IDCA-SVR). In the first step, ICA is used to decompose the given signal to its basic components. In the second step, a multi-scale decomposition is applied to each component to transform it into the corresponding subsequences. In the third step, all subsequences are clustered by utilizing the K-means clustering algorithm. Then, in the fourth step, the ANOVA test is applied to each cluster. The cluster with the highest intra-cluster correlation is selected. Finally, the electrical load is forecasted using SVR. The proposed IDCA-SVR method generates the most informative subsequences of the load time series with the lowest noise and outliers containing multi-scale information with both high and low frequencies. The proposed method is evaluated on three datasets of Poland, Bosnia and Herzegovina, and Canada (AMPds). The results show that the proposed method significantly decreases the forecasting error. Different measures such as mean absolute percent error (MAPE) and mean square error (MSE) are used to evaluate the forecasting results. According to done experiments, the proposed method forecasts the electrical load of Poland (training dataset) with MAPE = 1.7113 and MSE = 3.4621e−04, Poland (testing dataset) with MAPE = 2.3382 and MSE = 6.5692e−04, Bosnia and Herzegovina with MAPE = 1.8475e−04 and MSE = 1.0742e−09, and AMPds with MAPE = 0.499 and MSE = 1.4335e−07. According to the evaluation results, the proposed method works well on three datasets of Poland, Bosnia and AMPds compared to other forecasting methods. [ABSTRACT FROM AUTHOR]

With the continuous development of smart grid construction and the gradual improvement of power market operation mechanisms, the importance of power load forecasting is continually increasing. In this study, a short-term load prediction method based on the fuzzy optimization combined model of load feature recognition was designed to address the problems of weak generalization ability and poor prediction accuracy of the conventional feedforward neural network prediction model. First, the Douglas–Peucker algorithm and fuzzy optimization theory of load feature recognition were analyzed, and the combined prediction model was constructed. Second, data analysis and pre-processing were performed based on the actual historical load data of a certain area and the corresponding meteorological and calendar rule information data. Finally, a practical example was used to test and analyze the short-term load forecasting effect of the fuzzy optimization combined model. The calculation results proved that the presented fuzzy optimization combined model of load feature recognition outperformed the conventional model in terms of computational efficiency and specific performance; therefore, the proposed model supports further development of actual power load prediction. [ABSTRACT FROM AUTHOR]

With the development of smart grid, household load forecasting played an important role in power system operations. However, the household load forecasting is often inefficient due to its high volatility and uncertainty. Consequently, a multi-objective LSTM ensemble model based on the DBN combination strategy, is proposed in this paper. This method first builds a deep learning framework based on the LSTM network in order to generate several sub-models. With the diversity and accuracy of the sub-models as the objective functions, the improved MOEA/D algorithm is then used to optimize the parameters, in order to enhance the overall performance of the sub-models and ensure their differences. Finally, a DBN-based combination strategy is used to combine the single forecasts in order to form the ensemble forecast, and reduce the adverse effects of model uncertainty and data noise on the prediction accuracy. The experimental results show that the proposed method has several advantages in prediction accuracy and generalization capacity, compared with several current intelligent prediction methods. [ABSTRACT FROM AUTHOR]

Electrical load forecasting is a key player in building sustainable power systems and helps in efficient system planning. However, the irregular and noisy behavior in the observed data makes it difficult to achieve better forecasting accuracy. To handle this, we propose a new model, named singular spectrum analysis-long short- term memory (SSA-LSTM). SSA is a signal processing technique used to eliminate the noisy components of a skewed load series. LSTM model uses the outcome of SSA to forecast the final load. We have used five publicly available datasets from the Australian Energy Market Operator (AEMO) repository to assess the performance of the proposed model. The proposed model has superior forecasting accuracy compared to other existing state-of-the-art methods [persistence, autoregressive (AR), AR-exogenous, ARMA-exogenous (ARMAX), support vector regression (SVR), random forest (RF), artificial neural network (ANN), deep belief network (DBN), empirical mode decomposition (EMD-SVR), EMD-ANN, ensemble DBN, and dynamic mode decomposition (DMD)] for half-hourly and one day ahead load forecasting using RMSE and MAPE error metrics. [ABSTRACT FROM AUTHOR]

The forecast of electricity demand in recent years is becoming increasingly relevant because of market deregulation and the introduction of renewable resources. To meet the emerging challenges, advanced intelligent models are built to ensure precise power forecasts for multi-time horizons. The use of intelligent forecasting algorithms is a key feature of smart grids and an effective tool of resolving uncertainty for better cost and energy efficiency decisions like scheduling the generations, reliability and power optimization of the system, and economic smart grid operations. However, prediction accuracy in forecasting algorithms is highly demanded since many important activities of power operators like load dispatch depend upon the short-term forecast. This paper proposes a model for the estimation of the consumption of electricity in Agartala, Tripura in India, which can accurately predict the next 24 h of load with and estimation of load for 1 week to 1 month. A number of specific characteristics in the city have been analysed in order to extract variables that could affect the pattern of electricity consumption directly. In addition, the present paper shows the way to significantly improve the accuracy of the prediction through ensemble machine learning process. We demonstrated the performance of individual Random forest and XGBoost along with their ensemble. The RF and XGBoost ensemble obtained an accuracy with an improvement of 15–29%. The analyses or findings also provide interesting results in connection with energy consumption. [ABSTRACT FROM AUTHOR]

Since their inception, convolutional neural networks (CNNs) have been shown to have powerful feature extraction and learning capabilities, and the creation of deep residual networks (DRNs) was a milestone in the development of CNNs. However, residual networks mostly use convolution structures, which are widely applied to image recognition and classification problems. Therefore, when facing a load forecasting problem that involves nonlinear regression, will a DRN using a convolution structure still achieve great results? To answer this question, we present a network based on a DRN with a convolution structure to carry out short-term load forecasting, and we mainly focus on the effects of DRNs with different depths, widths and block structures for dealing with nonlinear regression problems. Through multiple sets of controlled experiments, we obtain the best network architecture and the corresponding hyperparameters for short-term load forecasting. The experimental results demonstrate that the model has higher prediction accuracy than existing models, and the DRN with a convolution structure can handle load forecasting while still achieving state-of-the-art results. [ABSTRACT FROM AUTHOR]

Electricity is the most important source of energy that is exploited nowadays; it is essential for the economic development and the social stability, and this implies the need to model systems that keeps a perfect balance between supply and demand. This task depends heavily on identifying the factors that affect power consumption and improving the precision of the forecasted model. This paper presents a novel convolutional neural network (CNN) for short-term load forecasting (STLF); studies have been conducted to identify the different factors that affect the power consumption in Algeria (North Africa), and these studies helped to determine the inputs to the model. The proposed CNN uses a two-dimensional input unlike the conventional one-dimensional input used for STLF, and the results given by the CNN were compared to other artificial intelligence methods and demonstrated good results for both: one-quarter-ahead and 24-h-ahead forecast. [ABSTRACT FROM AUTHOR]

The accuracy of short-term electricity load forecasting is of great interest since it allows avoiding unexpected blackouts and lowering operating costs. In this paper, we aim to implement the artificial neural networks to model and forecast the half-hourly electric load demand in Tunisia over the period 2000–2008. To improve the quality of forecasts, the proposed artificial neural network model uses not only past electric load values as inputs, but also climatic and calendar variables. To determine the optimal structure of the neural network model, this paper employs the pattern search algorithm. Moreover, the neural network model is equipped with the Levenberg–Marquardt learning algorithm. Our findings confirm the performance of this algorithm to the view of evaluation indicators since the mean absolute percentage error values range between 1.1 and 3.4%. The analysis also shows the superiority of the Levenberg–Marquardt algorithm compared to the resilient back propagation algorithm and the conjugate gradient algorithm. In the light of the current research, we stress the aptness of the proposed artificial neural network model in forecasting short-term electricity demand. [ABSTRACT FROM AUTHOR]

Many forecasting models are built on neural networks. The key issues in these models, which strongly translate into the accuracy of forecasts, are data representation and the decomposition of the forecasting problem. In this work, we consider both of these problems using short-term electricity load demand forecasting as an example. A load time series expresses both the trend and multiple seasonal cycles. To deal with multi-seasonality, we consider four methods of the problem decomposition. Depending on the decomposition degree, the problem is split into local subproblems which are modeled using neural networks. We move from the global model, which is competent for all forecasting tasks, through the local models competent for the subproblems, to the models built individually for each forecasting task. Additionally, we consider different ways of the input data encoding and analyze the impact of the data representation on the results. The forecasting models are examined on the real power system data from four European countries. Results indicate that the local approaches can significantly improve the accuracy of load forecasting, compared to the global approach. A greater degree of decomposition leads to the greater reduction in forecast errors. [ABSTRACT FROM AUTHOR]

Hybrid load forecasting models analyze linear and nonlinear components separately. If hybrid models were integrated with factor and feature selection algorithms, they would improve significantly. In the hybrid model proposed by this paper, the initial data were decomposed by an empirical mode decomposition (EMD) model. The linear component was analyzed through the autoregressive integrated moving average (ARIMA) method and the nonlinear component by a neural network (NN) and weighted by the improved flower pollination algorithm (IFPA). With the nonlinear component, the input load demand variable was decomposed by a wavelet transform (WT). In this paper, the improved grasshopper optimization algorithm (IGOA) and the principal component analysis (PCA) were employed to determine the input feature and input factor, respectively. Therefore, the proposed model was composed of EMD, IGOA, PCA, ARIMA, IFPA, NN, and WT algorithms. Finally, Iran's Electricity Market (IEM) data were used to show improvements in the precision of the proposed forecasting model. [ABSTRACT FROM AUTHOR]

Electrical load forecasting plays an important role in the regular planning of power systems, in which load is influenced by several factors that must be analysed and identified prior to modelling in order to ensure better and instant load balancing between supply and demand. This paper proposes a two-stage approach for short-term electricity load forecasting. In the first stage, a set of day classes of load profiles are identified using K-means clustering algorithm alongside daily temperature estimation profiles. The proposed estimation method is particularly useful in case of lack of historical regular temperature data. While in the second stage, the stacked denoising autoencoders approach is used to build regression models able to forecast each day type independently. The obtained models are trained and evaluated using hourly electricity power data offered by Algeria's National Electricity and Gas Company. Several models are investigated to substantiate the accuracy and effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

This paper tackles a new challenge in power big data: how to improve the precision of short-term load forecasting with large-scale data set. The proposed load forecasting method is based on Spark platform and "clustering–regression" model, which is implemented by Apache Spark machine learning library (MLlib). Proposed scheme firstly clustering the users with different electrical attributes and then obtains the "load characteristic curve of each cluster", which represents the features of various types of users and is considered as the properties of a regional total load. Furthermore, the "clustering–regression" model is used to forecast the power load of the certain region. Extensive experiments show that the proposed scheme can predict reasonably the short-term power load and has excellent robustness. Comparing with the single-alone model, the proposed method has a higher efficiency in dealing with large-scale data set and can be effectively applied to the power load forecasting. [ABSTRACT FROM AUTHOR]

Accurate short-term load forecasting (STLF) is crucial for reliable operation of a power system. Back propagation neural network (BPNN) is widely used in the forecasting field because of its powerful approximation capability. However, due to a variety of unstable factors, electrical time series often exhibit highly noisy and nonlinear characteristics. Usually, a large deviation will be produced when employing single BPNN to capture the complex data pattern. To solve this problem, this paper proposes a new hybrid forecasting approach that combines ensemble empirical mode decomposition (EEMD), chaotic self-adaptive flower pollination algorithm (CSFPA) and BPNN. EEMD is employed to decompose the original load series with the purpose of reducing the forecasting complexity. Developed CSFPA uses logistic equation to produce the chaotic initial population. In addition, aiming at providing a better optimization capability, CSFPA calculates the self-adaptive switch probability at each iteration. The best initial weights and biases of BPNN are provided by the optimization result of CSFPA. The performance of the proposed method is validated by two real-world load data sets from different electricity markets. The numerical results demonstrate that the proposed method outperforms three advanced methods; it is an effective and promising method for STLF. [ABSTRACT FROM AUTHOR]

The intermittency and uncertainty of wind power may pose danger on the safety on the power system, thus the research of wind speed and wind power prediction methods has been widely concerned. The keys of wind power forecasting are the forecasting model selection and model optimization. In this paper, the least squares support vector machine (LSSVM) is chosen as the wind speed and wind power prediction model and improved ant colony optimization (IACO) algorithm is used to optimize the most important parameters which influence the least squares support vector machine (LSSVM) regression model. In the IACO-LSSVM method, the kernel parameter σ and regularization parameter γ were considered the position vector of ants, the improved pheromone updating method is used to effectively solve the contradiction between expanding search and finding optimal solution. A multi-input–multi-output (MIMO) short-term wind speed and wind power forecasting model is built and applied in a wind farm of Gansu province in order to predict wind speed and wind power. For comparative study, PSO-LSSVM model and SVM model are used for forecasting. Prediction analysis results show that the IACO-LSSVM model can achieve higher prediction accuracy and confirm the effectiveness and feasibility of the method. [ABSTRACT FROM AUTHOR]

Electricity, an indispensable resource in daily life and industrial production, is hard to store, so accurate short-term load forecasting (STLF) plays a vital role in resource allocation, capital budgeting of power companies, energy deployment and government control. In recent decades, the strong dependency relationships of time series have been considered in many researches, but the discrete information has not proven to be very useful in their experiments. In general, while discrete information is weak, it can provide macro trends compared to the micro trends of continuous information. In this research, we aim to combine macro and micro information by continuous and discrete time series to generate multiple time series (MTS). The MTS comprise four information sequences: short-term, cycle, long short-term and cross-long short-term. These MTS are used to build a STLF system using a recurrent neural network (RNN) model that can learn sequential information between continuous and discrete series. Therefore, the RNN model with MTS information can improve the forecasting performance for short-term load forecasting. The experimental results show that our proposed forecasting system outperforms the state-of-the-art approach. [ABSTRACT FROM AUTHOR]

This paper presents short-term load forecasting models, which are developed by using fuzzy logic and adaptive neuro-fuzzy inference system (ANFIS). Firstly, historical data are analyzed and weekdays are grouped according to their load characteristics. Then, historical load, temperature difference and season are selected as inputs. In general literature, fuzzy logic hourly load forecasts are tested in the range a few days or a few weeks. Unlike previous studies, the hourly load forecast is carried out for 1 year. This paper shows that fuzzy logic can give good results in very large test data sets for 1 year. Besides, for countries with large areas, the temperature data taken from only one point would lead to increase the forecasting errors. Therefore, the average of temperature for six cities having the maximum power consumption is weighted average. The mean absolute percentage errors of the fuzzy logic and ANFIS models in terms of prediction accuracy are obtained as 2.1 and 1.85, respectively. The results show that the proposed fuzzy logic and ANFIS models are capable of load forecasting efficiently and produce very close values to the actual data and are the alternative way for short-term load forecasting in Turkey. [ABSTRACT FROM AUTHOR]

We offer a neural network model for forecasting the next day's hourly electric load of a city. We use a few ambient temperature account methods in the research to see how each of them affects the forecasting accuracy. Optimal meta-parameters are determined to tune the neural network to give best forecasts. Among such meta-parameters are the data history depth, data seasonality radius and regularization parameter of neural network weights. A multilayer perceptron is used to make forecasts. It is shown that the electric load can be forecasted most accurately when an additional neural network forecasts hourly ambient temperatures using actual hourly temperatures of the previous day and the weather station's temperature predictions for the forecast day. [ABSTRACT FROM AUTHOR]

Short-term electric load forecasting (STLF) is an essential tool for power generation planning, transmission dispatching, and day-to-day utility operations. A number of techniques are used and reported in the literature to build an accurate forecasting model. Out of them Artificial Neural Networks (ANN) are proven most promising technique for STLF model building. Many learning schemes are being used to boost the ANN performance with improved results. This motivated us to explore better optimization approaches to devise a more suitable prediction technique. In this study, we propose a new hybrid model for STLF by combining greater optimization ability of artificial bee colony (ABC) algorithm with ANN. The ABC is used as an alternative learning scheme to get optimized set of neuron connection weights for ANN. This formulation showed improved convergence rate without trapping into local minimum. Forecasting results obtained by this new approach have been presented and compared with other mature and competitive approaches, which confirms its applicability in forecasting domain. [ABSTRACT FROM AUTHOR]

Aiming to the disadvantages of short-term load forecasting with empirical mode decomposition (EMD) such as mode mixing and many high-frequency random components, a new short-term load forecasting model based on ensemble empirical mode decomposition (EEMD) and sub-section particle swarm optimization (SS-PSO) is proposed in this paper. Firstly, the load sequence is decomposed into a limited number of intrinsic mode function (IMF) components and one remainder by EEMD, which can avoid the mode mixing problem of traditional EMD. Then, through calculating and observing the spectrum of decomposed series, some low-frequency IMFs are extracted and reconstructed. Other IMFs can be forecasted with appropriate forecasting models. Since IMF1 is main random component of the load sequence, the linear combination model is adopted to forecast IMF1. Because the weights of the linear combination model are very important to obtain high forecasting accuracy, SS-PSO is proposed and used to optimize the linear combination weights. In addition, the factors such as temperature and weekday are taken into consideration for short-term load forecasting. Simulation results show that accuracy of the load forecasting model proposed in the paper is higher than that of BP neural network, RBF neural network, support vector machine, EMD and their combinations. [ABSTRACT FROM AUTHOR]

This paper proposes a self-similar local neuro-fuzzy (SSLNF) model with mutual information-based input selection algorithm for the short-term electricity demand forecasting. The proposed selfsimilar model is composed of a number of local models, each being a local linear neuro-fuzzy (LLNF) model, and their associated validity functions and can be interpreted itself as an LLNF model. The proposed model is trained by a nested local liner model tree (NLOLIMOT) learning algorithm which partitions the input space into axis-orthogonal sub-domains and then fits an LLNF model and its associated validity function on each sub-domain. Furthermore, the proposed approach allows different input spaces for rule premises (validity functions) and consequents (local models). This appealing property is employed to assign the candidate input variables (i.e., previous load and temperature) which influence short-term electricity demand in linear and nonlinear ways to local models and validity functions, respectively. Numerical results from short-term load forecasting in the New England in 2002 demonstrated the accuracy of the SSLNF model for the STLF applications. [ABSTRACT FROM AUTHOR]

Due to deregulation of electricity industry, accurate load forecasting and predicting the future electricity demand play an important role in the regional and national power system strategy management. Electricity load forecasting is a challenging task because electric load has complex and nonlinear relationships with several factors. In this paper, two hybrid models are developed for short-term load forecasting (STLF). These models use 'ant colony optimization (ACO)' and 'combination of genetic algorithm (GA) and ACO (GA-ACO)' for feature selection and multi-layer perceptron (MLP) for hourly load prediction. Weather and climatic conditions, month, season, day of the week, and time of the day are considered as load-influencing factors in this study. Using load time-series of a regional power system, the performance of ACO + MLP and GA-ACO + MLP hybrid models is compared with principal component analysis (PCA) + MLP hybrid model and also with the case of no-feature selection (NFS) when using MLP and radial basis function (RBF) neural models. Experimental results and the performance comparison with similar recent researches in this field show that the proposed GA-ACO + MLP hybrid model performs better in load prediction of 24-h ahead in terms of mean absolute percentage error (MAPE). [ABSTRACT FROM AUTHOR]

Short-term load forecasting is a popular topic in the electric power industry due to its essentiality in energy system planning and operation. Load forecasting is important in deregulated power systems since an improvement of a few percentages in the prediction accuracy will bring benefits worth of millions of dollars. In this study, a promising variant of genetic programming, namely gene expression programming (GEP), is utilized to improve the accuracy and enhance the robustness of load forecasting results. With the use of the GEP technique, accurate relationships were obtained to correlate the peak and total loads to average, maximum and lowest temperatures of day. The presented model is applied to forecast short-term load using the actual data from a North American electric utility. A multiple least squares regression analysis was performed using the same variables and same data sets to benchmark the GEP models. For more verification, a subsequent parametric study was also carried out. The observed agreement between the predicted and measured peak and total load values indicates that the proposed correlations are capable of effectively forecasting the short-term load. The GEP-based formulas are relatively short, simple and particularly valuable for providing an analysis tool accessible to practicing engineers. [ABSTRACT FROM AUTHOR]

A new support vector machine (SVM) optimized by an improved particle swarm optimization (PSO) combined with simulated annealing algorithm (SA) was proposed. By incorporating with the simulated annealing method, the global searching capacity of the particle swarm optimization (SAPSO) was enhanced, and the searching capacity of the particle swarm optimization was studied. Then, the improved particle swarm optimization algorithm was used to optimize the parameters of SVM ( c, σ and ε). Based on the operational data provided by a regional power grid in north China, the method was used in the actual short term load forecasting. The results show that compared to the PSO-SVM and the traditional SVM, the average time of the proposed method in the experimental process reduces by 11.6 s and 31.1 s, and the precision of the proposed method increases by 1.24% and 3.18%, respectively. So, the improved method is better than the PSO-SVM and the traditional SVM. [ABSTRACT FROM AUTHOR]

An improved wavelet neural network algorithm which combines with particle swarm optimization was proposed to avoid encountering the curse of dimensionality and overcome the shortage in the responding speed and learning ability brought about by the traditional models. Based on the operational data provided by a regional power grid in the south of China, the method was used in the actual short term load forecasting. The results show that the average time cost of the proposed method in the experiment process is reduced by 12.2 s, and the precision of the proposed method is increased by 3.43% compared to the traditional wavelet network. Consequently, the improved wavelet neural network forecasting model is better than the traditional wavelet neural network forecasting model in both forecasting effect and network function. [ABSTRACT FROM AUTHOR]

Electric load forecasting has received an increasing attention over the years by academic and industrial researchers and practitioners due to its major role for the effective and economic operation of power utilities. The aim of this paper is to provide a collective unified survey study on the application of computational intelligence (CI) model-free techniques to the short-term load forecasting of electric power plants. All four classes of CI methodologies, namely neural networks (NNs), fuzzy logic (FL), genetic algorithms (GAs) and chaos are addressed. The paper starts with some background material on model-based and knowledge-based forecasting methodologies revealing a number of key issues. Then, the pure NN-based and FL-based forecasting methodologies are presented in some detail. Next, the hybrid neurofuzzy forecasting methodology (ANFIS, GARIC and Fuzzy ART variations), and three other hybrid CI methodologies (KB-NN, Chaos-FL, Neurofuzzy-GA) are reviewed. The paper ends with eight representative case studies, which show the relative merits and performance that can be achieved by the various forecasting methodologies under a large repertory of geographic, weather and other peculiar conditions. An overall evaluation of the state-of-art of the field is provided in the conclusions. [ABSTRACT FROM AUTHOR]

In this paper we present the Bayesian Combined Predictor (BCP), a probabilistically motivated predictor for time series prediction. BCP utilizes local predictors of several types (e.g., linear predictors, artificial neural network predictors, polynomial predictors etc.) and produces a final prediction which is a weighted combination of the local predictions; the weights can be interpreted as Bayesian posterior probabilities and are computed online. Two examples of the method are given, based on real world data: (a) short term load forecasting for the Greek Public Power Corporation dispatching center of the island of Crete, and (b) prediction of sugar beet yield based on data collected from the Greek Sugar Industry. In both cases, the BCP outperforms conventional predictors. [ABSTRACT FROM AUTHOR]

This paper presents a novel method for short-term load forecasting based on local polynomial regression. Before applying the local regression, data mining algorithm selects historic load sequences satisfying known factors that are characterising required load model. Further on, the selected sequences are pre-processed with robust location estimator ( M- estimator) in order to reduce serial correlation and to eliminate outliers in historic data. On pre-processed load data we applied locally a truncated Taylor expansion to approximate functional relationship between load and load-affecting factors. Two methods for selecting optimal smoothing parameters (window size and polynomial degree) for local approximations are presented in the paper. These algorithms offer to us close insight into trade-off between bias and variance of the local approximations. In that way, they are able to help in selecting smoothing parameters locally (for each local fit) to fulfil the load modelling requirements. An example is presented at the end of this paper that clearly demonstrates the main features of this method. [ABSTRACT FROM AUTHOR]

This paper proposes a novel neural model to the problem of short-term load forecasting. The neural model is made up of two self-organizing map nets – one on top of the other. It has been successfully applied to domains in which the context information given by former events plays a primary role. The model was trained and assessed on load data extracted from a Brazilian electric utility. It was required to predict once every hour the electric load during the next 24 hours. The paper presents the results, and evaluates them. [ABSTRACT FROM AUTHOR]