Heating Load Forecast Based On Support Vector Machine Regression And On Control Strategy Optimization

The nonlinear characteristics of central heating system,such as multi-dimensional,high delay and strong coupling,cause the mismatch between the heat supply of heat source and the demand of users.Accurate heating load prediction is the key to realize heat supply on demand and improve energy efficiency in heating.This paper takes a residential heating system in Shijiazhuang city as the research object for heating load prediction.It takes the measured data of the whole heating season from 2016 to 2017 as the sample,and conducts the daily cumulative heating load prediction.First of all,it analyzes the main factors of heat load.Then,combining with the actual data collected,indoor temperature factor is added into the prediction models,and the input variables of the model are extracted: the accumulated heat consumption of the previous three days,the daily average outdoor temperature,and the daily average indoor temperature.In order to select an accurate heating load forecasting model,BP neural network model,multivariable linear regression model and support vector machine regression model are established on the basis of determining the model independent variables and model evaluation indexes.The results show that the prediction model of support vector machine regression is superior to BP neural network prediction model in prediction accuracy and variable interpretation ability,and is superior to multivariate linear regression prediction model in dealing with fuzzy and complex nonlinear problems.Considering the influence of the selection of the penalty parameter C and the kernel function parameter γ on the prediction accuracy of the support vector machine,in order to further improve the learning ability and prediction accuracy of support vector machine,the support vector machine regression model based on grid search,differential evolution,grey wolf algorithm and hybrid algorithm was established,respectively.By comparing and analyzing the prediction results of different optimization models,it can be seen that the prediction accuracy of support vector machine model based on optimization algorithm is higher than that of single support vector machine regression prediction model,and the support vector machine prediction model based on hybrid algorithm is superior to the prediction model based on single differential evolution,single gray wolf optimization and grid search optimization.On the basis of determining the optimal prediction model and combining with the weather forecast for the next 3 days,a phased regulation strategy of the secondary water supply temperature based on the prediction of daily cumulative heat consumption is proposed,which provides a reference for the optimal regulation of the heating system.