Heating Load Of Changchun Donghui Heating Area Forecasting Algorithm Based On BP Improved By PSO

The problems of energy extremely nervous and serious pollution of industrial today, if we adopted the central heating that we can reduce energy and environmental pollution, which can improve the efficiency of economic development. In recent years, central heating has become the main form of heating, which will also become a research focus in the thermal power generation in the future. With the reform and innovation of central heating, household metering heating system are constantly promotion, which puts forward higher requirements for central heating and heat supply network effective utilization. So the heating load forecasting will bring important meaning to the whole heat-supply system.This paper introduced the significance of central heating and significance of the heating load forecasting, also the domestic and foreign development status of heating load forecasting. The pipeline for water pressure, temperature, outdoor temperature, sunshine, temperature and so on all kinds of heating load between the related parameters of the detecting data do not exist correlation. Some common algorithms about heating load forecasting are analyzed. Through the characteristics of these algorithms, combined with the reaction time of heating load are relatively long, and the nonlinear characteristics of various parameters data. We selected neural network algorithm for the heating load forecasting. BP neural network algorithm has strong nonlinear mapping ability, self- learning ability and a wide range of adaptability. But BP neural network has some limitations, such as the initial weight settings are set randomly, the neural network can easily lead to instability or trapping in local optimal easily; or it is hard to get the convergence value; or it will lead to poor prediction accuracy. In order to optimize the algorithm, the particle swarm optimization is introduced. The swarm optimization algorithm can optimize algorithm of BP network structure, and increase the convergence speed. The optimization BP neural network algorithm can adjustment of connection weights and threshold quickly and improve the search ability.Application of the particle swarm optimization algorithm of BP neural network model to predict the thermal load, the first thing was to do the data pretreatment, then removed the abnormal heating load data; completed the missing data and normalized different dimension of the data, so as to improve the prediction accuracy of the model. After pretreatment, the data was input to the model, based on the optimized BP neural network for heating load prediction.Finally, chose Changchun thermal power plant heating load data in winter was input to the system model for short-term load forecasting. Simulation results show that applied optimization neutral network can modify parameters and structure more quickly and have more prediction accuracy. The result proved that the algorithm has the fast convergence; moreover, the average relative error of the improved model is lower than primitive algorithm.