A Short-term Load Forecasting Model Based on Neural Network Considering Weather Features

Abstract

In recent years, the grid load has been significantly affected by meteorological factors, showing strong nonlinearity and unpredictability, which brings great difficulties to load forecasting. Therefore, the use of meteorological factors for short-term load forecasting has become an indispensable factor to improve the application of smart grid. This paper first collects and analyzes the data of grid load, then disposes the abnormal load data. Secondly, meteorological factors such as temperature, humidity, wind speed, daily radiation, and rainfall are analyzed one by one, and the trend of load changes with temperature and humidity is obtained from this. Besides, these meteorological factors are coupled, and a comprehensive weather perception index is proposed to express the influence of the above factors on human skin perception. Finally, with the comprehensive weather perception index and load data as input, a BP neural network model is established for load forecasting, and actual calculation examples prove the high accuracy and rapidity of the method.