Industrial Ultra-Short-Term Load Forecasting With Data Completion

Abstract

Accurate and efficient ultra-short-term load forecasting is crucial for industrial power users to have stabilized and optimized operations. In this paper, we develop novel strategies for industrial power users to handle their challenges in ultra-short-term load forecasting. Firstly, this paper proposes a two-way Genetic Algorithm Back Propagation Neural Networks (GABPNN) missing data completion model to handle data loss, which is common power load data mining. A particle swarm optimization-supporting vector regression (PSO-SVR) algorithm is further used to integrate the two-way completion results with better accuracy. In addition, the paper introduces a combined ultra-short-term load forecasting model for industrial power users. The proposed model combines the Cubature Kalman filter (CKF) prediction model with good performance in nonlinear dynamic systems and the least square support vector machine (LS-SVM) prediction model with good performance in small-scale data prediction. The grey neural network is used to integrate the two algorithms, which further improves the accuracy of ultra-short-term load forecasting. Lastly, we test the proposed strategy in case study with real industrial data and demonstrate that the proposed model has a high degree of precision in load forecasting. INDEX TERMS Cubature Kalman filter; Least squares support vector machine; Grey neural network; PSO-SVR; Combined model; Clonal immune algorithm Haoyu Jiang received the bachelor degree from the

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

“…Its performance metrics are generally also better than those of MLNN, RNN, and VARMA; however, implementing all steps and finalizing results took a long time. VARMA yielded the lowest $${R^2}$$ and the highest MAPE, indicating that VARMA is less effective than other methods in week‐ahead peak load forecasting. Although both MLNN and RNN can learn from historical data, no feature extraction was performed, and so have poorer values of $${R^2}$$, MAPE and RMSE. The execution times (seconds) required by the proposed method, hybrid CNN (max tmp) and hybrid CNN (grid search) exceed one hour to optimally tune the topology and hyperparameters of the CNN by GA (outer loop, see Section 3.5) and train parameters and synaptic weights of the hybrid CNN by Adam optimizer (inner loop, see Section 3.5), as shown in Table 2. Because the most appropriate topology and/or hyperparameters of MLNN, RNN, SVR+PSO and VARMA herein were given from [10, 16, 25, 26], the execution times shown in Table 2 for these benchmark models only involved the training time. After the proposed hybrid CNN was well‐tuned by GA and trained by Adam optimizer, 146 sets of testing samples (20% of total datasets) were studied. The testing time was only 0.39 s for these testing samples (i.e.…”

“…The execution times (seconds) required by the proposed method, hybrid CNN (max tmp) and hybrid CNN (grid search) exceed one hour to optimally tune the topology and hyperparameters of the CNN by GA (outer loop, see Section 3.5) and train parameters and synaptic weights of the hybrid CNN by Adam optimizer (inner loop, see Section 3.5), as shown in Table 2. Because the most appropriate topology and/or hyperparameters of MLNN, RNN, SVR+PSO and VARMA herein were given from [10, 16, 25, 26], the execution times shown in Table 2 for these benchmark models only involved the training time.…”

“…The accuracies of various methods are compared. The methods of interest are MLNN [10], RNN [16], support vector regression (SVR) [24] that is optimized by particle swarm optimization (PSO) [25], and the vector autoregressive moving average (VARMA) model [26]. The proposed method considers the maximum, average and minimum temperatures.…”

“…. Because the most appropriate topology and/or hyperparameters of MLNN, RNN, SVR+PSO and VARMA herein were given from[10,16,25,26], the execution times shown in Table2for these benchmark models only involved the training time.…”

Short-Term Load Forecasting Using Neural Network and Particle Swarm Optimization (PSO) Algorithm

“…In [30], a hybrid method based on Elman neural network (ENN) and PSO is proposed. Reference [31] proposes a genetic-algorithmbased backpropagation neural network (GABPNN) considering data loss. Also, a particle swarm optimizationsupporting vector regression (PSO-SVR) algorithm is further used to integrate the GABPNN results with better accuracy.…”