Research On Urban Water Supplies Combination Forecasting Model Based On Driven Of The Time Series

By water supplies forecasting,it can provide data basis for the water supplies scheduling,and can provide guidance for the investment decision of the municipal water supply infrastructure improve utilization efficiencies of water resources,and promote social harmony and healthy development.In order to further improve the predictive accuracy of urban water supply,this paper focuses on the issues of the daily and monthly water supplies forecasting,through collecting the measured data of 1#,2#plant as the research object.To carry out the research which is based on the sequential drive of urban water supply combination forecast model research,obtains the related research results and conclusions are as follows:This paper introduces the basic principle of water supply forecasting and studies the predictability of the water supplies series(four time series collected from two scales water supply companies);Firstly,This paper introduces the basic principle of driver based on time sequence prediction model and data processing method,then analyzes the predictability of time series of the water supply.Through power spectrum and the maximum lyapunov index from the perspective of qualitative and quantitative analysis respectively,daily or monthly water supply time series are obvious chaos characteristic,indicated that the water supply time series predictability.Based on time-series driver chaos forecast method does not need to consider all the factors that impact load,only find the rule of the system from the sequential workload is small,not easily disturbed.Four models do not need the other factors that affect water supply(such as,population,climate,holidays,etc.),using only water supply timing,greatly reducing the workload of data collection,sorting,etc,the practical value is bigger.?Study on prediction model based on chaos theory;Based on chaos theory prediction model is the most widely used new technology forecasting model,It is one of the effective means to solve the prediction problem of nonlinear(chaotic)system,possesses the advantages of high forecast precision.Analysis compared prediction accuracy of the Chaotic global method,local-region method and the maximum Lyapunov index method,the results show that Chaos local-region method than the Chaotic global method and the maximum Lyapunov index method prediction accuracy is higher,and contrast phase synchronization,chaotic prediction error fluctuations smaller local method.Moreover,in order to further improve the prediction accuracy,discussed weighted first-order local-region method adjacent points optimization problem,evolutionary tracing method considered neighboring phase points evolution track,distinguish the true and false neighboring phase points,than the traditional selection methods and Information criterion to obtain the more accurate prediction results.?Study on Water supplies forecasting base on BP neural network,RBF neural network,and GRNN neural network;Back propagation and radical basis function neural network(three neural network model)which is the most mature and most widely used technology in the present is used as the basic model model of water supplies forecasting.The back propagation and radical basis function neural network theory and Model building steps introduced and analyzed the advantages and disadvantages of neural network,and proposed modified method.The results show that neural network can better grasp the water supply system trend,the radical basis function neural network prediction accuracy is higher than back propagation neural network,and radical basis function neural network the prediction fluctuation is small,but the disadvantage of neural networks is complex parameters design,prediction results is uncertainty.?The combination model research based on new technology prediction method;This paper introduces the characteristics of the combination model and the necessity of water supply system prediction model combination,then author introduces the principle of chaos algorithm combined with neural network model.Finally,in the example analysis,first using the weight one-rank local-region forecasting model,BP,RBF and GRNN neural network forecasting model,individual projections for daily or monthly water supply.Then,Three kinds of neural network and weight one-rank local-region forecasting model combination respectively.The results show that the combination forecast model accuracy is higher than the corresponding single forecasting model,the combination model of the weight one-rank local-region forecasting method and GRNN neural network model prediction accuracy is the highest,it have better stability and shorter operation time,proposed combination forecast method is feasible.In this paper,the combined forecasting model not only provides an effective new method for urban water supply forecasting,and also for the development of other related research provides a new way of thinking.All forecasting methods are compiled the corresponding program and apply to the software copyright,patent,which laid a foundation for practical use in the future.