Research On Short-Term Power Load Forecasting Based On Random Vector Functional Link Network

Short-term power load forecasting is an important part of power system planning and operation,which is of great significance to ensure the safe and stable operation of power system.However,some deep learning methods of existing need high computation cost,how to establish a fast accurate predicting network model is challenging.Due to random vector functional link network(RVFL)network has the characteristics of rapid convergence.Therefore,a variety of short-term power load forecasting models based on RVFL network are established in this paper.According to the characteristics of the RVFL network,this article from three aspects to improve the accuracy of the short-term power load forecasting based on RVFL network,the main research contents and innovation points are as follows:First of all,the initialization of RVFL network is very important to the prediction model,and improper initialization will make the convergence speed and performance of the model worse.To solve the above problems,gaussian distribution and Sine activation function are adopted in this paper to better complete parameter initialization of RVFL network and Regularization method is adopted to optimize the output weight of RVFL network.Through the establishment of Re RVFL network forecasting model,The validity of the Re RVFL network forecasting model is verified by simulation and comparison experiment.Secondly,when the training data in the model is changed,the original parameters of the model are not applicable and the prediction accuracy will be reduced.In this paper,incremental learning(IL)method is added to Re RVFL network to establish short-term power load prediction model of IL-Re RVFL network,so that the parameters of the model can be kept updated.Through simulation and comparison experiments,it is shown that the model can further improve the accuracy of short-term power load prediction.Finally,in view of the power load data belongs to a kind of typical time series,it has the characteristics of nonlinear nonstationary.In this paper,the empirical mode decomposition(EMD)is used to establish the short-term power load prediction model of EMD-IL-Re RVFL.Due to the EMD has the disadvantage of mode mixing,it can not achieve the desired effect and even make the effect worse.in this paper,further uses improved EMD decomposition algorithm: Ensemble Empirical Mode Decomposition(EEMD)and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN),and set up EEMD-IL-Re RVFL forecasting model and CEEMDAN-IL-Re RVFL forecasting model.Through the simulation experiments show that the two based on improved EMD and IL-Re RVFL network integration the combined model has higher prediction accuracy.