Research On The Heating Load Combination Forecasting Based On Particle Swarm Optimization

The accurate predication of heating load is of great practical significance to improve system management of the central heating and the heating quality, environmental protection, and energy conservation. It is difficult to get satisfactory results by using one forecasting model, however, combined forecasting method,which utilizes various information provided by every forecasting model to get an appropriate combination forecasting model,can effectively improve the fitting ability of forecasting model to improve the accuracy of prediction.In this paper, we study on the particle swarm optimization and the combination prediction theory based on the prior study, then apply both of the researches in the heating load forecast, and established the combination forecast model for short-term heating load forecast at last. The determination of weight for single predication is critical in the process of building the combined forecasting model. This work mainly focuses on single forecasting model such as least square method, time series method, RBF neural networks method according to the features of heating load forecast, then made the historical data by the heating system measured data of some thermal energy station in the city of Daqing, established the objective function under the constraints to achieve its smallest in terms of least squares criterion to Calculates the weight of the single forecasting model. Finally we get a certain forecast model, and use the data from daily run of a thermal substation in Daqing to check out the forecasting effect of model.The objective function created by least squares criterion is a constrained optimization problem, while particle swarm optimization has the advantages of easy solving process and less computational cost. Thereby,particle swarm optimization is an important tool to solve such a problem, and it is chosen to seek for the weight of heating load forecasting model. For the particle swarm algorithm is easy to fall into local optimal solution, slow convergence in the Later stage of evolution and other shortcomings, the article has improved standard particle swarm optimization on three aspects.In this paper, the combination of forecasting and particle swarm optimization solves the problem that the weight coefficient is difficult to determine. These made heating load forecasting become more scientific, and have instructional significance on heating load forecasting.