Heating Load Forecast Study Based On Correlation Analysis Of Influencing Factors And Wavelet Neural Network

In the context of building energy conservation is increasingly valued,heating energy conservation as an important part of it has also been concerned.Heating as expectation and energy saving is an important means of achieving energy savings in heating,and to achieve heating as expectation and energy saving,the heating load should be predicted.Based on the operation data of a heat exchange station in a certain district of Beijing,this paper conducts a heating load forecasting study of a central heating system.The main research contents are as follows:Firstly,the heating operation parameters of the heat exchange station and the meteorological parameters of the area are recorded,and the recorded data is subjected to wavelet denoising processing.According to the denoising performance evaluation parameter SNR and root mean square error,different data select the appropriate wavelet.The denoising method eliminates errors in recorded data and provides the best denoising data for the heating load forecast.Secondly,the factors affecting the heating load and its delay law are analyzed,and the correlation between the influencing factors and the heating load is analyzed.Select the influencing factors such as outdoor temperature and solar radiation as the input parameters of the heating load prediction model.Provide a basis for constructing a suitable heating load forecast model.Then,the BP neural network forecast model is used to predict the heating load,and the average relative error between the prediction result and the real heat consumption is 7.91%.In order to improve the accuracy of heating load forecast,the wavelet neural network forecast model combined with wavelet analysis and neural network is used to predict the heating load.The average relative error between the prediction result and the real heat consumption is 6.87%.The prediction accuracy is better than the BP neural network model.There is a certain improvement.Finally,because the wavelet neural network and BP neural network are easy to fall into the local optimum problem,the particle swarm optimization algorithm is used to optimize the wavelet neural network forecast model.The average relative error between the predicted result and the true heat consumption is 5.01%.The relative error of the smaller-wave neural network forecast results decreased by 1.86%,which was 2.9% lower than that predicted by the BP neural network.The relevant research results in this paper can provide some theoretical guidance for the central heating system to achieve heating as expectation and energy saving.