Prediction Research of Heating Load in Residential Area Based on BP Neural Network

Abstract

This paper analyzed the influence factors of heating load in residential area. The BP neural network algorithm was used for forecast, and BP neural network structure were built. The input sample was outdoor temperature, outdoor wind speed, indoor heat coefficient and heat supply in the last few days. The network structure output was heating load value on the day. Besides, the heating load of a residential district was predicted and verified by day unit and hour unit heating load, and the error of heating load forecasting with different methods were compared. The result shows that the regulation of heating load changes in residential area is periodic. the accuracy of heating load forecasting using hours as a unit is higher than using the day as a unit.