Research On Hybrid Forecasting Method Of Heat Load Based On Neural Network And Gray System

Urban central heating is the main way of heating the northern winter,but the central heating system will consume too much energy.In order to better run and adjust the central heating system,it is necessary to predict the thermal load of the urban central heating system,so that we can control the heat load more effectively.First of all,the main factors affecting the heating load are analyzed in this paper,and three main factors are extracted from the actual data collection: outdoor average temperature,outdoor average wind speed and sunshine time.According to the related theory of gray correlation degree,it’s easily to calculate the gray correlation coefficient,in which the average outdoor temperature has the greatest influence.This conclusion coincides with the actual situation.Then this paper takes a central heating project of Beijing Hairou as the research object.Based on the gray prediction algorithm and BP neural network algorithm,this paper constructs the prediction models to study the forecast problem of the heating.We analyze the relationship between the number of initial data and the average relative error with the metabolic gray prediction model.The results show that the number of initial data has a critical value.If the number exceeds the critical value,the prediction accuracy will be reduced.At the same time,compared with the prediction accuracy of the conventional gray model and the metabolic gray model,the conclusion can be obtained intuitively.In the application of artificial neural network principle,the parameter establishment method is employed.According to the prediction results,it can be concluded that the BP neural network model has high precision.Finally,two kinds of unambiguous combinations,namely weight-based combination forecasting and prediction based on forecasting method,are used to predict the heat load data,and the better forecasting results are obtained.