Short-term Load Forecasting Based on the Improved BAS Optimized Elman Neural Network

Abstract

To improve the accuracy of short-term load forecasting, an Elman neural network short-term load forecasting model based on an improved beetle antennae search (BAS) algorithm was proposed. Firstly, the classical function Rosenbrock was optimized by particle swarm optimization (PSO) algorithm, BAS algorithm, and improved BAS algorithm which convergence faster by comparison. Further, the improved BAS algorithm was used to optimize the short-term power load prediction model of the Elman neural network, and the stability of the algorithm was improved by taking into account the shortcoming of the traditional BAS algorithm, which was prone to fall into local convergence. We used ground load data in the European Intelligent technology Competition for analysis and prediction. Then we used respectively traditional Elman neural network and the Elman optimized by the improved BAS algorithm to predict and analyze the power load for a month, and obtain the residual and residual rates between the load and the real value. Example results and comparative analysis show that the operation time of the improved BAS-Elman neural network is about 15.8% shorter than that of the traditional BAS network, and the prediction accuracy is about 1.2% higher than original Elman short-term power load forecasting.