Short-term Power Load Research Articles

Improving the Bi-LSTM model with XGBoost and attention mechanism: A combined approach for short-term power load prediction

Short term power load forecasting plays an important role in the management and development of power systems with a focus on the reduction in power wastes and economic losses. In this paper, we construct a novel, short-term power load forecasting method by improving the bidirectional long short-term memory (Bi-LSTM) model with Extreme Gradient Boosting (XGBoost) and Attention mechanism. Our model differs from existing methods in the following three aspects. First, we use the weighted grey relational projection algorithm to distinguish the holidays and non-holidays in the data preprocessing. Secondly, we add the Attention mechanism to the Bi-LSTM model to improve the validity and accuracy of prediction. Thirdly, XGBoost is a newly-developed, well-performing prediction model, which is used together with the Attention mechanism to optimize the Bi-LSTM model. Therefore, we develop a novel, combined power load prediction model “Attention-Bi-LSTM + XGBoost” with the weight determination theory-error reciprocal method. Using two power market datasets, we evaluate our prediction method by comparing it with two benchmark models and four other models. With our prediction method, the MAPE, MAE, and RMSE for the Singapore’s power market are 0.387, 43.206, and 54.357, respectively; and those for the Norway’s power market are 0.682, 96.278, and 125.343, respectively. The test results are smaller than the results for six other models. This indicates that our prediction method outperforms the LSTM, Bi-LSTM, Attention-RNN, Attention-LSTM, Attention-Bi-LSTM, and XGBoost in effectiveness, accuracy, and practicability.

Short-term power load forecasting based on gray relational analysis and support vector machine optimized by artificial bee colony algorithm.

Short-term power load forecasting is essential in ensuring the safe operation of power systems and a prerequisite in building automated power systems. Short-term power load demonstrates substantial volatility because of the effect of various factors, such as temperature and weather conditions. However, the traditional short-term power load forecasting method ignores the influence of various factors on the load and presents problems of limited nonlinear mapping ability and weak generalization ability to unknown data. Therefore, a short-term power load forecasting method based on GRA and ABC-SVM is proposed in this study. First, the Pearson correlation coefficient method is used to select critical influencing factors. Second, the gray relational analysis (GRA) method is utilized to screen similar days in the history, construct a rough set of similar days, perform K-means clustering on the rough sets of similar days, and further construct the set of similar days. The artificial bee colony (ABC) algorithm is then utilized to optimize penalty coefficient and kernel function parameters of the support vector machine (SVM). Finally, the above method is applied on the basis of actual load data in Nanjing for simulation verification, and the results show the effectiveness of the proposed method.

Short-Term Electrical Load Forecasting Based on Time Augmented Transformer

Electrical load forecasting is of vital importance in intelligent power management and has been a hot spot in industrial Internet application field. Due to the complex patterns and dynamics of the data, accurate short-term load forecasting is still a challenging task. Currently, many tasks use deep neural networks for power load forecasting, and most use recurrent neural network as the basic architecture, including Long Short-Term Memory (LSTM), Sequence to Sequence (Seq2Seq), etc. However, the performance of these models is not as good as expected due to the gradient vanishing problem in recurrent neural network. Transformer is a deep learning model initially designed for natural language processing, it calculates input–output representations and captures long dependencies entirely on attention mechanisms which has great performance for capturing the complex dynamic nonlinear sequence dependence on long sequence input. In this work, we proposed a model Time Augmented Transformer (TAT) for short-term electrical load forecasting. A temporal augmented module in TAT is designed to learn the temporal relationships representation between the input history series to adapt to the short-term power load forecasting task. We evaluate our approach on a real-word dataset for electrical load and extensively compared it to the performance of the existed electrical load forecasting model including statistical approach, traditional machine learning and deep learning methods, the experimental results show that the proposed TAT model results in higher precision and accuracy in short-term load forecasting.

Platform-Independent Web Application for Short-Term Electric Power Load Forecasting on 33/11 kV Substation Using Regression Tree

Short-term electric power load forecasting is a critical and essential task for utilities in the electric power industry for proper energy trading, which enables the independent system operator to operate the network without any technical and economical issues. From an electric power distribution system point of view, accurate load forecasting is essential for proper planning and operation. In order to build most robust machine learning model to forecast the load with a good accuracy irrespective of weather condition and type of day, features such as the season, temperature, humidity and day-status are incorporated into the data. In this paper, a machine learning model, namely a regression tree, is used to forecast the active power load an hour and one day ahead. Real-time active power load data to train and test the machine learning models are collected from a 33/11 kV substation located in Telangana State, India. Based on the simulation results, it is observed that the regression tree model is able to forecast the load with less error.

An ensemble multi-step M-RMLSSVR model based on VMD and two-group strategy for day-ahead short-term load forecasting

Accurate prediction of the power load is one of the keys to guarantee stable operation of power construction. However, with the surge of power load, the uncertainty of multi-step prediction brings more challenges. To reduce the uncertainty of multi-step short-term power load forecasting, a short-term ensemble day-ahead forecasting model based on variational modal decomposition (VMD) and two-group multi-step strategy is proposed in this paper. Firstly, VMD is used to decompose the original sequence into different sub-sequences. Then, according to the curve characteristics of different sub-sequences, different multi-step strategies are selected to achieve the best prediction effect. Finally, the multi-step results of all sub-sequences are integrated to get the final day-ahead prediction results. Among them, the two groups of multi-step strategies are divided into single-output approaches (Direct, Recursive, DirRec) and multi-output approaches (MIMO, RecMO, DirMO). For the prediction tool of the single-output approaches, a hybrid prediction model with integrating decomposition, feature selection, optimization algorithm and LSSVR is proposed. For the prediction tool of the multi-output approaches, a hybrid algorithm based on multi-kernel RMLSSVR is proposed to enhance the diversity of MLSSVR, and the parameters and weights of multi-kernel MLSSVR are optimized by the PSO with extreme value disturbance without velocity update. Taking the hourly market load data of California as an example, three groups of comparative experiments are carried out on the proposed method. Experiments show that the accuracy of our model is improved by at least 19% compared with the best RecMO strategy.

Economic Dispatch of Microgrid Based on Load Prediction of Back Propagation Neural Network–Local Mean Decomposition–Long Short-Term Memory

To plan the work of power generation equipment, it is necessary to ensure that the power supply is sufficient and to achieve the minimum cost to ensure the safety and economy of the microgrid. Based on back propagation neural network–local mean decomposition–long short-term memory (BPNN–LMD–LSTM) load prediction, the design is based on a fixed-time consistency algorithm with random delay to predict the economic dispatch of microgrids. Firstly, the initial power load prediction sequence is obtained by continuous training of the back propagation neural network (BPNN); the residual sequence with other influencing factors is decomposed by local mean decomposition (LMD); and the long short-term memory neural network (LSTM) is used to predict the output prediction residual sequence, and the final short-term power load prediction is obtained. Based on predicting load, the fixed-time consistency algorithm with random delay is used to add supply and demand balance constraints to optimize the power distribution of the power generation units of the distributed microgrid and reduce the power generation cost of the microgrid. The results show that the prediction model has better prediction accuracy, and the scheduling algorithm based on the prediction model has a faster convergence rate to reach the lowest power generation cost.

A Novel Short-Term Power Load Forecasting Method Based on TSNE-EEMD-LSTM

In this paper, a novel power load forecasting model is proposed to fully extract the periodic characteristics of short-term load at various time scales and explore the potential correlations between influencing factors and characteristics of load components. Firstly, the t-distributed stochastic neighbor embedding algorithm is used to map sample points of high-dimensional load influencing factors to low-dimensional space, and the ensemble empirical mode decomposition algorithm is employed to split the historical load curve into multiple signal components with different frequencies. Then, several long short-term memory networks including nonlinear mapping and time series models are established to mine the relationship between low-dimensional comprehensive influencing factors and each intrinsic mode function component by utilizing different inputs. Finally, the effectiveness of the hybrid model is verified via using the short-term load dataset of 3-hour data granularity in a certain region, and the influence of key parameters of the model on the forecasting effect is discussed.

Research on Short-term Load Forecasting Based on GWO-BILSTM

In short-term power load forecasting, changes in external multi-dimensional factors will have a certain impact on the accuracy of load forecasting. In response to this problem, this paper proposes a prediction model GWO-BILSTM that combines the Grey Wolf Optimizer (GWO) and the bi-directional long short-term memory (BILSTM) network. Taking the real power load data as the data set, the high correlation parameters are selected as the input through the Pearson correlation analysis, the hyperparameters of BILSTM are optimized by the GWO algorithm, and finally the GWO-BILSTM prediction model is established based on the optimized parameters to predict the data set. The experimental results show that the mean absolute percentage error, root mean square error and mean absolute error index of the GWO-BILSTM model are better than other comparison models, which effectively improves the prediction accuracy of multi-dimensional power load data.

Short-term power load interval forecasting based on nonparametric Bootstrap errors sampling

Short-term power load forecasting plays a vital role in the planning of distribution network and the development of social economy. Many researchers have devoted their attention to construct point forecasting models. However, the traditional point forecasting regards the forecasting result as a deterministic variable. Since the deviation existed in load forecasting is simply unavoidable and significant, which has great volatility and randomness, point forecast methods are difficult to capture the fluctuation of power load. Probability forecast models are proposed to obtain the uncertain information in the load power. In this paper, firstly, some error information is obtained from the deterministic forecasting results of point forecasting; secondly, the interval of time series data is divided according to the deterministic error information; finally, the Bootstrap method is used to estimate the confidence interval of the deteministic error information to obtain the uncertainty information in the power load data. The instability and randomness of the deterministic error are otained by combining the interval forecasting method so as to improve the accuracy of power load forecasting. Therefore, probability forecasting is combined with point forecasting to obtain more accurate results. The proposed model is used to forecast power load for Queensland, Australia and capital district of New York State. The experimental results show that the proposed method performs better than other comparative models. The experimental results show that the ELM-AdaBoost model has better forecast performance in both long-term and short-term load datasets, and can overcome the seasonality of power load time series data.

Short-Term Load Forecasting Based on the CEEMDAN-Sample Entropy-BPNN-Transformer

Aiming at the problem that power load data are stochastic and that it is difficult to obtain accurate forecasting results by a single algorithm, in this paper, a combined forecasting method for short-term power load was proposed based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN)-sample entropy (SE), the BP neural network (BPNN), and the Transformer model. Firstly, the power load data were decomposed into several power load subsequences with obvious complexity differences by using the CEEMDAN-SE. Then, BPNN and Transformer model were used to forecast the subsequences with low complexity and the subsequences with high complexity, respectively. Finally, the forecasting results of each subsequence were superimposed to obtain the final forecasting result. The simulation was taken from our proposed model and six forecasting models by using the load dataset from a certain area of Spain. The results showed that the MAPE of our proposed CEEMDAN-SE-BPNN-Transformer model was 1.1317%, while the RMSE was 304.40, which was better than the selected six forecasting models.

Short-Term Power Load Forecasting Based on SAPSO-CNN-LSTM Model considering Autocorrelated Errors

Accurate power load forecasting is essential for power grid operation and dispatching. To further improve the accuracy of power load forecasting, this study proposes a new power load forecasting method. Firstly, correlation coefficients of influential variables are calculated for feature selection. Secondly, the form of input data is changed to adjust for autocorrelated errors. Thirdly, data features are extracted by convolutional neural networks (CNN) to construct feature vectors. Finally, the feature vectors are input into long short-term memory (LSTM) network for training to obtain prediction results. Moreover, for solving the problem that network hyperparameters are difficult to set, the simulated annealing particle swarm optimization (SAPSO) algorithm is used to optimize the hyperparameters. Experiments show that the prediction accuracy of the proposed model is higher compared with LSTM, CNN-LSTM, and other models.

A Temporal-Spatial Model Based Short-Term Power Load Forecasting Method in COVID-19 Context

The worldwide coronavirus disease 2019 (COVID-19) pandemic has greatly affected the power system operations as a result of the great changes of socio-economic behaviours. This paper proposes a short-term load forecasting method in COVID-19 context based on temporal-spatial model. In the spatial scale, the cross-domain couplings analysis of multi-factor in COVID-19 dataset is performed by means of copula theory, while COVID-19 time-series data is decomposed via variational mode decomposition algorithm into different intrinsic mode functions in the temporal scale. The forecasting values of load demand can then be acquired by combining forecasted IMFs from light Gradient Boosting Machine (LightGBM) algorithm. The performance and superiority of the proposed temporal-spatial forecasting model are evaluated and verified through a comprehensive cross-domain dataset.

Short-term power load probability density forecasting based on GLRQ-Stacking ensemble learning method

The high penetration rate of distributed energy brings severe challenges to the dispatch and operation of power systems. Improving the accuracy of short-term power load forecasting can enhance the management efficiency of the system, and thus enhance the economic and social benefits of the system. To that end, gradient boosting decision tree (GBDT), light gradient boosting machine (LGBM) and random forest (RF) are used as the base learners, and quantile regression long-short term memory (QRLSTM) is used as the meta-learner to construct a novel Stacking ensemble learning model (GLRQ-Stacking) in this paper. This is a novel heterogeneous ensemble framework, which cleverly combines Bagging and Boosting to achieve Stacking. Thereinto, more comprehensive forecast information is provided by QRLSTM. Afterwards, a density prediction method using Gaussian approximation of quantiles (GAQ) to modify kernel density estimation (KDE) is proposed. The reconstructed probability density curves are derived by modifying KDE. While unifying the values of metrics, the constructed prediction intervals (PIs) cover more real values. The accuracy and effectiveness of the framework are evaluated based on two German datasets from different seasons and one Electrician Mathematical Contest in Modeling (EMCM) data case. Through the comparison of the results, the superiority of the proposed ensemble framework and the modified density prediction method is verified, which could improve operations management of power systems.

Short-Term Power Load Forecasting Model Based on t-SNE Dimension Reduction Visualization Analysis, VMD and LSSVM Improved with Chaotic Sparrow Search Algorithm Optimization

Short-Term Power Load Forecasting Model Based on t-SNE Dimension Reduction Visualization Analysis, VMD and LSSVM Improved with Chaotic Sparrow Search Algorithm Optimization

Variable-Weighted Ensemble Forecasting of Short-Term Power Load Based on Factor Space Theory

Variable-Weighted Ensemble Forecasting of Short-Term Power Load Based on Factor Space Theory

Research on short-term load forecasting based on accumulated temperature effect and improved temporal convolutional network

Aiming at the uncertainty in short-term load forecasting, and considering the multi-dimensionality of load data and the continuity of time series, this paper constructs a short-term power load forecasting method based on accumulated temperature effect and improved Temporal Convolutional Network. First, the common load influencing factors in the short-term forecasting process were explored, the data were cleaned and correlated, and a quantitative model of temperature variables that considered the effect of accumulated temperature was established, such as the weighted average method and the discrete correction model. Secondly, fully consider and divide the multi-dimensional influencing factors, the time series characteristic quantity is formed by taking advantage of the Temporal Convolutional Network (TCN), extracting potential relationships between time-series data and non-time-series data, and the optimal load estimation value is obtained. Finally, the output of the TCN is fused with the non-time-series data to form a new input feature matrix, and the load prediction is completed through the Back Propagation (BP) Neural Network. The calculation example shows that the TCN-BP model combined with temperature correction can reduce the average absolute error by 1.77%. Under the same forecasting method, the load forecasting accuracy obtained by the weighted average method is higher than that of the discrete correction model. Compared with the single TCN and BP models, the method proposed in this paper increases the training time, but reduces the time loss of prediction and improves the prediction accuracy.

Application of Support Vector Regression and Time Series Method in Short-term Power Load Forecasting with Regional Difference

Aims: The aim of this study is to perform short-term load forecasting. Background: Short-term load forecasting plays a key role in power dispatching. It provides basic data for basic power generation planning and system safety analysis so that the power dispatching work is more practical and the power generation efficiency is higher. Objective: The aim of this study is to ensure the safe operation of the electricity market and relieve the pressure of supply and demand. Method: In this paper, the SVR model is used for short-term load prediction. Results: The SVR model has the advantage of minimizing the structural risk and has good generalization performance for the predicted object. At the same time, the global optimization is ensured, a lot of mapping calculation is reduced, the actual risk is reduced, and the prediction performance is improved. Conclusion: The target model has higher forecasting accuracy than other forecasting models and can effectively solve the problems of the power market.

Multivariate Adaptive Step Fruit Fly Optimization Algorithm Optimized Generalized Regression Neural Network for Short-Term Power Load Forecasting

Short-term load forecasting plays a significant role in the management of power plants. In this paper, we propose a multivariate adaptive step fruit fly optimization algorithm (MAFOA) to optimize the smoothing parameter of the generalized regression neural network (GRNN) in the short-term power load forecasting. In addition, due to the substantial impact of some external factors including temperature, weather types, and date types on the short-term power load, we take these factors into account and propose an efficient interval partition technique to handle the unstructured data. To verify the performance of MAFOA-GRNN, the power load data are used for empirical analysis in Wuhan City, China. The empirical results demonstrate that the forecasting accuracy of the MAFOA applied to the GRNN outperforms the benchmark methods.

Short-Term Power Load Forecasting of Integrated Energy System Based on Attention-CNN-DBILSTM

In view of the fact that the potential high-dimensional features in the historical sequence are difficult to be effectively extracted by traditional power load forecasting methods and the coupling factors of electricity, heat, and gas have not been considered, the correlation of electric heating and gas load is considered in this paper, and a short-term power load forecasting method for integrated energy systems based on Attention-CNN- (Convolutional Neural Network-) DBILSTM (Deep Bidirectional Long-Short-Term Memory) is proposed. First, the correlation between the multiple load influencing factors is considered, and the Pearson coefficient is used to quantitatively calculate the correlation between the multiple loads. Second, a CNN network consisting of a one-dimensional convolutional layer and a pooling layer is established. High-dimensional features reflecting the dynamic changes of the load are extracted, and the proposed feature vector is constructed in the form of time series as the input of the DBILSTM network; the dynamic change law of time series data is modeled and learned. Then, the Attention mechanism is introduced to assign different weights to the hidden state of DBILSTM through the mapping weight and the learning parameter matrix, to reduce the loss of historical information and strengthen the impact of key information, and the Dense layer is used to output the load prediction results. Finally, the influence of the correlation of multiple loads and its influencing factors on the power load forecasting results is analyzed, based on the historical load data of the integrated energy system in a certain area of Northeast China. The simulation results of the calculation example show that the prediction accuracy of the method reaches 97.99%, and the integrated energy system electric, heat, and gas load correlation coefficients as the input parameters of the Attention-CNN-DBILSTM network can reduce the average prediction error by 0.37%∼1.93%. The proposed method has been verified to effectively improve the prediction accuracy by comparison with the prediction model results of CNN-LSTM network, CNN-BILSTM network, and CNN-DBILSTM network.

Short-term power load forecasting based on BiGRU-Attention-SENet model

ABSTRACT Due to the randomness of weather conditions, the complexity of power demands and the nonlinearity of high-dimensional data, traditional regression analysis, and time series forecasting methods can no longer meet the current requirements of high-precision wind power load forecasting. Therefore, it is necessary to use an accurate short-term load forecasting method for power grid dispatching and planning. Taking the onshore wind power load data of Valencia, Spain (from January 1, 2015, to December 31, 2018), as the research object, this paper proposes a hybrid model of bidirectional gated recurrent units and squeeze-and-excitation networks (BiGRU-SENet) based on an attention mechanism, which is good at handling the nonlinearity of high-dimensional time series data, to predict short-term power loads. The results showed that (1) the attention mechanism could significantly improve the predictive performance of BiGRU by weighting the input of different time steps. (2) The GRU-CNN was more suitable for short-term load forecasting than CNN-GRU, which could make full use of time series information and not lose the key information at the beginning. (3) SENet enabled CNN to perform feature-wise recalibration, further focusing on effective features. (4) Compared with the second-best model (BiGRU-Attention-CNN), BiGRU-Attention-SENet was 12.849% and 3.096% lower in terms of the mean absolute percentage error (MAPE) and root mean square error (RMSE), respectively. In terms of computational overheads, BiGRU-Attention-SENet achieved greater improvement in prediction accuracy by adding relatively fewer computational overheads (the parameters increased by 0.338%, and the FLOPs increased by 0.621%). (5) Through fine tuning, the pre-trained model achieved almost the same accuracy as the no pre-trained one in a shorter time, which proved that BiGRU-Attention-SENet had strong generalization ability. The hybrid model proposed in this paper has application prospects in power load forecasting.