A Combined ARIMA-PPR Model for Short-Term Load Forecasting

Abstract

Accurate short-term load forecasting (STLF) is essential to ensure reliable and efficient operation plan , and improve the utilization of power grid . The purpose of this research is to explore and evaluate the use of a combined ARIMA-PPR model to STLF. By fully analyzing the auto-correlation of load data, we develop an autoregression integrated moving average (ARIMA) model and a projection pursuit regression (PPR) model to capture the linear and nonlinear pattern exhibited in the load series. The combined model is constructed by assigning weight coefficients to individual models, and the weight coefficients are determined by root mean square error (RMSE). The combined model is applied to the 5 min interval load series data of Sichuan Province, China to the task of half-hour ahead forecasting. The combined model outperforms two single models in a variety of performance measures. It achieves more reliable and better prediction results, because it can capture both linear and non-linear modes. . Results show that the combined model is a promising tool for STLF.