Heat Load Forecast Method Research Of The Central Heating System

China's northern cities and towns in winter heating required for heating energy consumption in the social energy consumption accounted for a large proportion.With the increasing emphasis on energy conservation,most areas have adopted a central heating method.However,the central heating system covers a wide range of areas,including a large number of hot users and complex large heat pipe network,the system of many factors,internal correlation,lag time is long,there is a serious non-linear,control and adjustment is very difficult,Therefore,it is very important to predict the actual heat demand of the heat users during the operation regulation of the heating system.In order to predict the thermal load of the central heating system more accurately,this paper analyzes the influencing factors of the heat load.Based on the input variables and the evaluation criteria of the forecasting model,the neural network and the support vector machine In the prediction of heat load,different optimization and improvement models are established,and the performance of heat load forecasting is studied in detail.For the neural network algorithm,this paper establishes the BP neural network prediction model,and then uses the wavelet analysis theory to improve the model and established the wavelet neural network prediction model.For the support vector machine algorithm,this paper establishes the support vector machine regression prediction model,and uses the particle swarm optimization algorithm to optimize the parameters of the support vector machine forecasting model,and thus builds the particle swarm support vector machine prediction model.In order to improve the learning ability of support vector machine,this paper adopts the dynamic multi-population particle swarm optimization support vector machine algorithm,and establishes a dynamic load forecasting model based on dynamic multi-population particle swarm support vector machine.The support vector machine(SVM)algorithm is more advanced than the neural network in dealing with the high-dimensional mathematical problems related to the heating load,the results show that the SVM algorithm is more advanced than the neural network;The search ability of dynamic multi-population particle swarm optimization in parameter optimization is obviously superior to particle swarm optimization;The prediction accuracy of the optimized prediction model is higher than that of the original prediction model;The predictive model of the thermal load forecasting model of the central heating system established by the support vector machine and its optimization algorithm is better than the prediction model established by using the neural network and its optimization algorithm.Based on the analysis and comparison of each model and based on the measured data of heating,the comprehensive evaluation factors are used.The dynamic multi-population particle swarm support vector machine thermal load forecasting model is stable and has high prediction accuracy.Accurate and effective for the scientific production of heating enterprises to provide an effective reference for the distribution of heat,scheduling to provide the necessary basis.