Research On Heating Load Forecasting Based On Neural Network And Particle Swarm Optimization

As the main infrastructure for winter heating in northern cities,central heating systems consume a lot of energy and cause pollution to the environment,so it is important to achieve energy-efficient operation.The current form of heating in China is still crude,making it difficult to achieve the requirement of heating on demand,which causes energy wastage and also reduces the experience of heat users.Accurate heating load prediction is an important way to achieve energy-efficient operation of central heating systems.In this paper,a residential-type heat station of a central heating system in Beijing is used as the research object,and the research on heating load prediction is conducted based on BP neural network.The effectiveness of the heating load prediction model is compared and analysed in terms of the influencing factors,the structure of the neural network and the division of the heating cycle.A particle swarm algorithm is used to optimise the BP neural network to further improve the accuracy of the heating load prediction in view of the shortcomings of the initial weight setting of the BP neural network.The heating load of a heat station is affected by multiple factors such as weather,system and random factors.Too many factors input into the neural network model will only increase the complexity of the model.In this paper,a correlation analysis method is used to calculate the Pearson correlation coefficient between each factor and the heating load.The results show that outdoor temperature,building thermal inertia and atmospheric pressure have a large influence on the heating load and are initially used as features to be introduced into the model.While having good non-linear mapping capability,the BP neural network also has the advantages of strong generalisation capability and high error tolerance,making it very suitable for the prediction of heating loads.On the basis of BP neural network prediction,the influence of input features is firstly analysed to establish models with different input features;afterwards,a double-implicit neural network model is established to compare with a single-implicit layer neural network model;finally,the heating week is divided into the early end and mid-term to establish a neural network model for simulation experiments,and the prediction results are analysed by comparing the evaluation indexes of the prediction results,namely the mean absolute percentage error,the mean By comparing the mean absolute percentage error,mean absolute error and root mean square error of the prediction results,the model with the highest accuracy,MAPE,was 3.25%.To further improve the accuracy,a particle swarm algorithm(PSO)was used to optimise the BP neural network.The particle swarm algorithm has powerful optimisation finding capability and can find the best initial weights and thresholds for the BP neural network.In this paper,a PSO-BP neural network is established to predict the heating load of a heat station.The prediction results show that the PSO-BP neural network can further improve the prediction accuracy,and the MAPE of the two models decreased by 0.17% and0.24% respectively compared to the standard BP neural network.The prediction of heat supply load of heat stations based on BP neural network can achieve high accuracy and meet the requirements of practical application,verifying the feasibility of the neural network prediction method,which can guide the economic operation of heat stations and provide a reference for the accurate prediction of heat supply load,and has practical application significance.