Power Load Forecasting Based on Data Analysis and Neural networks

Abstract

Load forecasting plays a vital role in power system planning, and it is closely related to the problems such as power transmission and power dispatching. However, the power demand between power companies and consumers is varying from time to time, and it is difficult to forecast the load precisely. Usually, the power load is relative to economy, weather, season, temperature and other factors. In order to improve the precision of load forecasting, it is significant to grasp main characteristics and factors influencing the power consumption. This paper proposes a load forecasting approach combining the principal component analysis (PCA) with neural networks (NN). The PCA distills prime factors to reduce the inputs of NN, and it also compresses the sample space reasonably. The NN is used to forecast power load with a deep neural network structure which includes eight inputs and two hide layers. This paper uses some practical data of power system to test the performances of the proposed load forecasting approach, and the results show that the proposed approach forecasts power load more precisely than the traditional approach in some aspects.