An Improved Quantum Particle Swarm Algorithm Optimized Regularized Extreme Learning Machine for Short-Term Load Forecasting

Abstract

In order to screen out the nonlinear relationship in massive load data more quickly and improve the accuracy of short-term load forecasting model, the genetic quantum particle swarm optimization algorithm (GAQPSO) is proposed to optimize the input weight and hidden layer deviation of the regularized extreme learning machine (RELM) based on the shortcomings of quantum particle swarm optimization algorithm (QPSO) in dealing with complex high-dimensional parameter optimization problems, which forms a hybrid short-term load forecasting model called GAQPSO-RELM. Meanwhile, when the input features are selected, the influences of historical load, temperature, time interval and type of date are fully considered, so the accuracy of the short-term load forecasting model is further improved. The experimental result shows that the proposed short-term load forecasting model has a higher accuracy than the QPSO-RELM model and the common RELM model. Besides, it can better reflect the changeable trend of the daily load curve, which verifies the effectiveness of the proposed predictive model.