Research On Heat Load Prediction Of Central Heating System Based On Improved Genetic Algorithm

The central heating system is a basic livelihood project in northern China,which has long been valued for its high energy consumption and close connection with the winter haze problem.At the same time,with the rapid development of China’s social economy,the improvement of people’s quality of life also puts forward higher requirements for the central heating industry.The research focus of central heating in the new period is mainly on energy saving and emission reduction and meeting the needs of users,which put forward requirements for accurate load forecast of central heating.The heat exchange station is an important part of the central heating system.In the actual heating system operation control,the short-term heat load prediction of the heat exchange station based on historical operation data plays an important role.In this paper,the short-term heat load prediction of heat exchange station is taken as the object.In view of the difficulty of heat load prediction with time-varying,time lag,uncertainty and strong coupling,a heat load prediction model based on improved genetic algorithm combined with LSTM neural network is proposed to obtain a relatively ideal Predicted heat load.The main content of this article is as follows,First,comprehensively analyze the influencing factors of heat load.The relationship between various influencing factors and heat load is studied,and the internal connection is determined through correlation analysis and significance test,which makes the subsequent modeling input feature selection have a judgment basis.Second,construct a short-term heat load prediction model based on LSTM neural network.Considering the outstanding performance of the long-term and short-term memory neural network in time series forecasting problems,the heat exchange station based on the long-and short-term memory neural network was constructed with the date type,weather characteristics and heat load sequence as input and the short-term heat load prediction value as output.Short-term heat load prediction model,and compared with the commonly used modeling algorithms in the research field,such as wavelet neural network(WNN),GA-BP,support vector machine(SVR),etc.The experimental results show that the LSTM-based heat load prediction model has Good performance.Third,to achieve short-term heat load forecast model optimization based on improved genetic algorithm.Considering the LSTM neural network modeling process,the choice of model hyperparameters usually depends on the experience of researchers,or a large number of experimental results,the performance of the training model is not guaranteed and unstable,so a combination of intelligent optimization methods is adopted To assist LSTM modeling to achieve better model performance.The paper conducted a lot of data experiments on genetic algorithm(GA),proposed a dynamic assisted individual oriented crossover(DAIOX)operator,and proved that the improved genetic algorithm has better optimization performance,Further combining the genetic algorithm based on dynamically assisted individual directed crossover operator with LSTM network to construct a thermal load prediction model with good performance.Finally,develop a visualization platform for thermal load prediction based on.Net.The development of the visualization platform for heat load prediction is completed through the.Net platform,which makes the prediction method of the subject research have practical significance.