Daily Peak Load Forecasting Research Articles

Application of Seasonal Trend Decomposition using Loess and Long Short-Term Memory in Peak Load Forecasting Model in Tien Giang

Daily peak load forecasting is critical for energy providers to meet the loads of grid-connected consumers. This study proposed a Seasonal Trend decomposition using Loess combined with Long Short-Term Memory (STL-LTSM) method and compared its performance on peak forecasting of electrical energy demand with Convolutional Neural Network and LSTM (CNN-LSTM), Wavenet, and the classic approaches Artificial Neural Network (ANN) and LSTM. The study evaluated the models using demand data from the power system in Tien Giang province, Vietnam, from 2020 to 2022, considering historical demand, holidays, and weather variables as input characteristics. The results showed that the proposed STL-LSTM model can predict future demand with lower Base Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). Therefore, the proposed method can help energy suppliers make smart decisions and plan for future demand.

RAID: Robust and Interpretable Daily Peak Load Forecasting via Multiple Deep Neural Networks and Shapley Values

Accurate daily peak load forecasting (DPLF) is crucial for informed decision-making in energy management. Deep neural networks (DNNs) are particularly apt for DPLF because they can analyze multiple factors, such as timestamps, weather conditions, and historical electric loads. Interpretability of machine learning models is essential for ensuring stakeholders understand and trust the decision-making process. We proposed the RAID (robust and interpretable DPLF) model, which enhances DPLF accuracy by recognizing daily peak load patterns and building separate DNN models for each day of the week. This approach was accessible for energy providers with limited computational resources, as the DNN models could be configured without a graphics processing unit (GPU). We utilized scikit-learn’s MLPRegressor for streamlined implementation, Optuna for hyperparameter optimization, and the Shapley additive explanations (SHAP) method to ensure interpretability. Applied to a dataset from two commercial office buildings in Richland, Washington, RAID outperformed existing methods like recurrent neural networks, Cubist, and HYTREM, achieving the lowest mean absolute percentage error values: 14.67% for Building 1 and 12.74% for Building 2. The kernel SHAP method revealed the influence of the previous day’s peak load and temperature-related variables on the prediction. The RAID model substantially improved energy management through enhanced DPLF accuracy, outperforming competing methods, providing a GPU-free configuration, and ensuring interpretable decision-making, with the potential to influence energy providers’ choices and promote overall energy system sustainability.

A Hybrid Model of VMD-EMD-FFT, Similar Days Selection Method, Stepwise Regression, and Artificial Neural Network for Daily Electricity Peak Load Forecasting

Daily electricity peak load forecasting is important for electricity generation capacity planning. Accurate forecasting leads to saving on excessive electricity generating capacity, while maintaining the stability of the power system. The main challenging tasks in this research field include improving forecasting accuracy and reducing computational time. This paper proposes a hybrid model involving variational mode decomposition (VMD), empirical mode decomposition (EMD), fast Fourier transform (FFT), stepwise regression, similar days selection (SD) method, and artificial neural network (ANN) for daily electricity peak load forecasting. Stepwise regression and similar days selection method are used for input variable selection. VMD and FFT are applied for data decomposition and seasonality capturing, while EMD is employed for determining an appropriate decomposition level for VMD. The hybrid model is constructed to effectively forecast special holidays, which have different patterns from other normal weekdays and weekends. The performance of the hybrid model is tested with real electricity peak load data provided by the Electricity Generating Authority of Thailand, the leading power utility state enterprise under the Ministry of Energy. Experimental results show that the hybrid model gives the best performance while saving computation time by solving the problems in input variable selection, data decomposition, and imbalance data of normal and special days in the training process.

Week‐ahead daily peak load forecasting using genetic algorithm‐based hybrid convolutional neural network

Daily peak load forecasting is crucial for the operation of bulk power systems, including economic dispatch and unit commitment. It is also essential for peak load shaving and load management in distribution systems. In power markets, peak load forecasting helps participants develop bidding strategies. This paper proposes a new method, using a hybrid convolutional neural network (CNN) that is cascaded with a fully-connected network, for making week-ahead daily peak load forecasts. The proposed method uses three loops to obtain the optimal CNN: The outer loop performs crossover/mutation operations and tournament selection to produce chromosomes to optimize the network topology and hyperparameters (such as kernel size) of the hybrid CNN by genetic algorithms; the middle loop deals with the order of chromosomes; the inner loop optimizes the synaptic weights and parameters (e.g. values of a kernel) using Adam optimizer. Daily peak load data and corresponding meteorological data for Taiwan are explored. Simulation results show that the proposed method outperforms the traditional CNN, multi-layer neural network, recurrent neural network, support vector regression and vector autoregressive moving average model.

LSTM enhanced by dual-attention-based encoder-decoder for daily peak load forecasting

Daily peak load forecasting is a challenging problem in the filed of electric power load forecasting. Since the nonlinear and dynamic of influence factors and their sequential dependencies are significant for modeling daily peak load, a prediction model based on long short-term memory (LSTM) enhanced by dual-attention-based encoder-decoder is presented. Functioned as the specific encoder and decoder, LSTM is utilized to participate in the nonlinear dynamic temporal modeling. The encoder-decoder is used for information utilization of both the influence factors and daily peak load. Moreover, a dual-attention mechanism, which is inserted into the encoder-decoder, is designed to take into account the effects of different influence factors and time nodes on the daily peak load simultaneously. It is benefit for the above mechanism design to analyze the characteristics of daily peak load precisely and to achieve more accurate prediction results. Comprehensive experiments are performed based on a real set of one provincial capital city in eastern China. The case study shows that the proposed methodology provides the most accurate results with an average MAPE 2.07%, an average RMSE 133 MW and an average MAE 326.6 MW.

Interpretable Short-Term Electrical Load Forecasting Scheme Using Cubist

Daily peak load forecasting (DPLF) and total daily load forecasting (TDLF) are essential for optimal power system operation from one day to one week later. This study develops a Cubist-based incremental learning model to perform accurate and interpretable DPLF and TDLF. To this end, we employ time-series cross-validation to effectively reflect recent electrical load trends and patterns when constructing the model. We also analyze variable importance to identify the most crucial factors in the Cubist model. In the experiments, we used two publicly available building datasets and three educational building cluster datasets. The results showed that the proposed model yielded averages of 7.77 and 10.06 in mean absolute percentage error and coefficient of variation of the root mean square error, respectively. We also confirmed that temperature and holiday information are significant external factors, and electrical loads one day and one week ago are significant internal factors.

Deep Learning for Daily Peak Load Forecasting–A Novel Gated Recurrent Neural Network Combining Dynamic Time Warping

Daily peak load forecasting is an essential tool for decision making in power system operation and planning. However, the daily peak load is a nonlinear, nonstationary, and volatile time series, which makes it difficult to be forecasted accurately. This paper, for the first time, proposes a bespoke gated recurrent neural network combining dynamic time warping (DTW) for accurate daily peak load forecasting. The shape-based DTW distance is used to match the most similar load curve, which can capture trends in load changes. By analyzing the relationship between the load curve and the cycle of human social activities, the some-hot encoding scheme is first applied on the discrete variables to expand the features to further characterize their impact on load curves. Then, a three-layer gated recurrent neural network is developed to forecast daily peak load. The proposed algorithm is implemented on the Theano deep learning platform and tested on the loaded dataset of the European Network on Intelligent Technologies. The simulation results show that the proposed algorithm achieves satisfactory results compared with other algorithms using the same dataset in this paper.

Daily Peak Load Forecasting by Artificial Neural Network using Differential Evolutionary Particle Swarm Optimization Considering Outliers

Abstract This paper proposes an Artificial Neural Network (ANN) based daily peak load forecasting method by differential evolutionary particle swarm optimization (DEEPSO) considering outliers. When outliers exist in the training data, forecasting accuracy of daily peak load forecasting can be affected by the outliers. Therefore, engineers have removed the outliers from training data so far and it is a heavy burden for engineers. Utilization of evolutionary computation has a possibility to solve this problem. Moreover, forecasting accuracy may be improved using evolutionary computation techniques instead of the conventional stochastic gradient descent (SGD) with outliers. The proposed weights tuning method by DEEPSO is compared with the conventional weights tuning methods by SGD and PSO for verification of the efficacy of the proposed method.

Daily Peak Load Forecasting Based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and Support Vector Machine Optimized by Modified Grey Wolf Optimization Algorithm

Daily peak load forecasting is an important part of power load forecasting. The accuracy of its prediction has great influence on the formulation of power generation plan, power grid dispatching, power grid operation and power supply reliability of power system. Therefore, it is of great significance to construct a suitable model to realize the accurate prediction of the daily peak load. A novel daily peak load forecasting model, CEEMDAN-MGWO-SVM (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and Support Vector Machine Optimized by Modified Grey Wolf Optimization Algorithm), is proposed in this paper. Firstly, the model uses the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) algorithm to decompose the daily peak load sequence into multiple sub sequences. Then, the model of modified grey wolf optimization and support vector machine (MGWO-SVM) is adopted to forecast the sub sequences. Finally, the forecasting sequence is reconstructed and the forecasting result is obtained. Using CEEMDAN can realize noise reduction for non-stationary daily peak load sequence, which makes the daily peak load sequence more regular. The model adopts the grey wolf optimization algorithm improved by introducing the population dynamic evolution operator and the nonlinear convergence factor to enhance the global search ability and avoid falling into the local optimum, which can better optimize the parameters of the SVM algorithm for improving the forecasting accuracy of daily peak load. In this paper, three cases are used to test the forecasting accuracy of the CEEMDAN-MGWO-SVM model. We choose the models EEMD-MGWO-SVM (Ensemble Empirical Mode Decomposition and Support Vector Machine Optimized by Modified Grey Wolf Optimization Algorithm), MGWO-SVM (Support Vector Machine Optimized by Modified Grey Wolf Optimization Algorithm), GWO-SVM (Support Vector Machine Optimized by Grey Wolf Optimization Algorithm), SVM (Support Vector Machine) and BP neural network to compare with the CEEMDAN-MGWO-SVM model and analyze the forecasting results of the same sample data. The experimental results fully demonstrate the reliability and effectiveness of the CEEMDAN-MGWO-SVM model proposed in this paper for daily peak load forecasting, which shows the strong generalization ability and robustness of the model.

Identification of nonlinear Hammerstein system using mixed integer-real coded particle swarm optimization: application to the electric daily peak-load forecasting

This paper investigates the modeling of a class of dynamic systems using nonlinear Hammerstein (NLH) model composed of a memory-less polynomial block cascaded to an autoregressive with exogenous input (ARX) time-series block. The model thus defined is known as NLHARX. Both the integer orders and the real coefficients of the model are identified simultaneously in a unified framework using a new algorithm based on a mixed coded integer-real particle swarm optimization. Unlike classical identification methods which assume the orders to be known in advance, the proposed approach is new since it estimates both the real and integer design parameters while minimizing the error between the outputs of the system and the model. The usefulness and the effectiveness of the proposed approach have been demonstrated through extensive simulations. Two illustrative examples are included in this paper: an empirical example and an application to the forecasting of the daily peak-load of Hail region, Saudi Arabia. Future works will be devoted to the identification of more complex dynamic systems, such as Hammerstein–Wiener and the application to the prediction of time-series related to water and energy.

Daily Peak Load Forecasting by Taguchi's T Method

SUMMARYIn this paper, a new technique of daily peak load forecasting is proposed. At first, we propose nonlinear correction T method as improvement technique of T method that is one of the multivariate analysis techniques, called Mahalanobis–Taguchi (MT) system, suggested in Quality Control. Moreover, we examine the daily peak load forecasting technique, which is based on data in the last year on a forecasting day, by T method and nonlinear correction T method. Furthermore, we examine the technique, which is based on the data of most recent several weeks of a forecasting day, by the T method and Nonlinear correction T method. In addition, the effectiveness of the proposed method by comparing the multiple regression analysis is discussed.

Locally-Weighted Polynomial Neural Network for Daily Short-Term Peak Load Forecasting

Electric load forecasting is essential for effective power system planning and operation. Complex and nonlinear relationships exist between the electric loads and their exogenous factors. In addition, time-series load data has non-stationary characteristics, such as trend, seasonality and anomalous day effects, making it difficult to predict the future loads. This paper proposes a locally-weighted polynomial neural network (LWPNN), which is a combination of a polynomial neural network (PNN) and locally-weighted regression (LWR) for daily short-term peak load forecasting. Model over-fitting problems can be prevented effectively because PNN has an automatic structure identification mechanism for nonlinear system modeling. LWR applied to optimize the regression coefficients of LWPNN only uses the locally-weighted learning data points located in the neighborhood of the current query point instead of using all data points. LWPNN is very effective and suitable for predicting an electric load series with nonlinear and non-stationary characteristics. To confirm the effectiveness, the proposed LWPNN, standard PNN, support vector regression and artificial neural network are applied to a real world daily peak load dataset in Korea. The proposed LWPNN shows significantly good prediction accuracy compared to the other methods.

タグチのT法を用いた翌日最大電力需要予測

In this paper, a new technique of daily peak load forecasting is proposed. At first, we propose Non-linear correction T method as improvement technique of T method that is one of the multivariate analysis techniques, called MT system, suggested in Quality Control. Moreover, we examine the daily peak load forecasting technique, which is based on data in the last year on a forecasting day, by T method and Non-linear correction T method. Furthermore, we examine the technique, which is based on the data of most recent several weeks of a forecasting day, by the T method and Non-linear correction T method. In addition, the effectiveness of the proposed method by comparing the multiple regression analysis is discussed.

非線形補正T法を用いた翌日最大電力需要予測

非線形補正T法を用いた翌日最大電力需要予測

Feature selection for daily peak load forecasting using a neuro-fuzzy system

Accurate electrical daily peak load forecasting (DPLF) is essential for power system management in order to prevent overloading and grid failure. Fuzzy neural networks have been successfully applied to load forecasting due to their nonlinear mapping and generalized behavior. In this paper, a neuro-fuzzy based DPLF (N-DPLF) model with a feature selection method is proposed for DPLF. The load data is clustered into seven subsets according to the season and day type. For each subset, the four features with the highest salience ranks are selected. After training N-DPLF model, the formed BSWs (bounded sum of weighted fuzzy membership functions) in accordance with the selected features denote characteristics of these features. The N-DPLF model provides explicit BSWs in hyperboxes, instead of the uncertain black box nature of neural network models, so that the selected features can be interpreted by the visually constructed BSWs. The N-DPLF model with a feature selection method shows a mean absolute percentage error (MAPE) of 1.86 % using Korea Power Exchange data over 1-year period.

Daily Peak Load Forecasting of Next Day using Weather Distribution and Comparison Value of Each Nearby Date Data

By the development of industry, in recent years; dependence to electric energy is growing year by year. Therefore, reliable electric power supply is in need. However, to stock a huge amount of electric energy is very difficult. Also, there is a necessity to keep balance between the demand and supply, which changes hour after hour. Consequently, to supply the high quality and highly dependable electric power supply, economically, and with high efficiency, there is a need to forecast the movement of the electric power demand carefully in advance. And using that forecast as the source, supply and demand management plan should be made. Thus load forecasting is said to be an important job among demand investment of electric power companies. So far, forecasting method using Fuzzy logic, Neural Net Work, Regression model has been suggested for the development of forecasting accuracy. Those forecasting accuracy is in a high level. But to invest electric power in higher accuracy more economically, a new forecasting method with higher accuracy is needed. In this paper, to develop the forecasting accuracy of the former methods, the daily peak load forecasting method using the weather distribution of highest and lowest temperatures, and comparison value of each nearby date data is suggested.

Combined use of unsupervised and supervised learning for daily peak load forecasting

Abstract In this paper, we have aimed to present a hybrid neural network model for daily electrical peak load forecasting (PLF). Since peak loads usually follow similar patterns, classification of data improves the accuracy of the forecasts. Several factors in peak load, e.g. weather temperature, relative humidity, wind speed and cloud cover, were introduced into the model in order to enhance forecast quality. Most classification attempts in the literature have been intuitive and empty of justification. In this paper, we have proposed a novel approach for clustering data by using a self-organizing map. The Davies–Bouldin validity index was introduced to determine the best clusters. A feed forward neural network (FFNN) has been developed for each cluster to provide the PLF. Eight training algorithms have also been used in order to train the proposed FFNNs. Applying principal component analysis (PCA) decreased the dimensions of the network’s inputs and led to simpler architecture. To evaluate the effectiveness of the proposed hybrid model (PHM), forecasting has been performed by developing a FFNN that uses the un-clustered data. The results proved the superiority and effectiveness of the PHM. Linear regression (LR) models have also been developed for PLF, and the results indicated that the PHM produces considerably better forecasts than those of LR models. Furthermore, the results show that the suggested clustering approach significantly improves the forecasting results on regression analysis too.

Improving electric load forecasts using network committees

Accurate daily peak load forecasts are important for secure and profitable operation of modern power utilities, with deregulation and competition demanding ever-increasing accuracies. Machine learning techniques including neural and abductive networks have been used for this purpose. Network committees have been proposed for improving regression and classification accuracy in many disciplines, but are yet to be widely applied to load forecasting. This paper presents a formal approach to apply the technique using historical load and temperature data spanning multiple years, with individual committee members trained on different years. Correlation among data for successive years is investigated and methods to enhance independence between member models for improving committee performance are described. Both neural and abductive networks implementations are presented and compared. An abductive network three-member committee was developed on data for three successive years and evaluated on the fourth year. Compared to a monolithic model trained on the same full three-year data, the committee reduces the mean absolute percentage error from 2.52% to 2.19%. The corresponding reduction in the mean of the absolute error from 70 MW to 61 MW is statistically significant at the 95% confidence level.

Daily Peak Load Forecasting by Structured Representation on Genetic Algorithms for Non-linear Function Fitting

Daily Peak Load Forecasting by Structured Representation on Genetic Algorithms for Non-linear Function Fitting

Artificial neural network-based peak load forecasting using conjugate gradient methods

The daily electrical peak load forecasting (PLF) has been done using the feed forward neural network (FFNN)-based upon the conjugate gradient (CG) back-propagation methods, by incorporating the effect of 11 weather parameters, the previous day peak load information, and the type of day. To avoid the trapping of the network into a state of local minima, the optimization of user-defined parameters, namely, learning rate and error goal, has been performed. The training dataset has been selected using a growing window concept and is reduced as per the nature of the day and the season for which the forecast is made. For redundancy removal in the input variables, reduction of the number of input variables has been done by the principal component analysis (PCA) method of factor extraction. The resultant dataset is used for the training of a 3-layered NN. To increase the learning speed, the weights and biases are initialized according to the Nguyen and Widrow method. To avoid over fitting, an early stopping of training is done at the minimum validation error.