Urban Short-term Water Supply Combined Forecasting Model Based On LSSVM-ARIMA

Through the short-term water supply forecast,the optimal control of the water supply system can be achieved,and the supply and demand balance(to meet customer demand for water)and energy conservation(to save energy consumption)purposes can be achieved.In order to further improve the prediction accuracy of short-term urban water supply,this paper takes the measured data collected from a water plant in a city as the research object,and studies the LSSVM-ARIMA based short-term urban water supply combination prediction model from the perspectives of 15-minute and hourly water supply prediction.The relevant research results and conclusions are as follows:(1)Analysis and research on short-term water supply prediction characteristics: first,standard deviation denoising method was used to preprocess the water supply time series,and invalid information of water supply time series was removed.The standard deviation denoising method in this paper includes 1 times standard deviation denoising method,2 times standard deviation denoising method and 3 times standard deviation denoising method.After processing,there are 8 kinds of water supply time series(including original 15 minutes and time supply time series).Then the predictability and individual differences of each water supply sequence are analyzed.In this study,the maximum Lypunov index and power spectrum were used for quantitative and qualitative analysis of water supply sequences,respectively.It was determined that the 15-minute and time-dependent water supply sequences in this study had chaotic characteristics,which indicated that the water supply sequence in this study was predictable and had practical significance for its prediction.In this study,through the analysis of the characteristics of the water supply sequence,it is found that the water supply sequence has the characteristics of linearity,non-linearity,periodicity andlarge local fluctuation.Therefore,ARIMA model and support vector machine model are selected as the basic models of the combination model.(2)Support vector machine(SVM)is a new nonlinear technology,which performs well in prediction field and has a high accuracy.Through the in-depth study of support vector machine,it is found that the standard support vector machine model is more suitable for predicting small sample sequences.Due to the large sample size of instance water supply sequence,this study adopts the least square support vector machine(LSSVM)model to predict the time sequence of instance water supply for 15 minutes and at the same time.LSSVM model is more suitable for fitting prediction of large sample sequences on the basis of retaining the advantages of standard SVM model.The results show that LSSVM can effectively predict the overall change trend of the water supply time series of the instance,and the prediction accuracy of LSSVM on the 15-minute water supply time series is higher than that of the water supply time series,which indicates that LSSVM model has better fitting effect on the water supply time series with strong nonlinear trend.For LSSVM model,moderate denoising can improve the prediction accuracy of the model.(3)Research based on ARIMA prediction model: the most commonly used time series prediction model--ARIMA model is adopted to predict the15-minute and water-supply time series of an instance.ARIMA model has the advantages of fast convergence and strong robustness.The results show that the ARIMA model can track the overall trend change of the water supply sequence,but the local fitting effect is poor.The prediction accuracy of ARIMA model for the 15-minute water supply sequence is worse than that for the time-supply sequence,which indicates that ARIMA model is more suitable for predicting the time series with strong linear trend.For ARIMA model,denoising can improve the prediction accuracy.(4)Based on the study of short-term water supply combination prediction model: according to the characteristics of information independence and detachability in the time series of water supply of an example,ARIMA model and LSSVM model are combined in series mode,and LSSVM-ARIMAcombination prediction model is established at last.According to the evaluation index of the model,the prediction accuracy of the combined model and each single model is analyzed and compared.The results show that the prediction accuracy of the combined model is significantly higher than that of the single model,which proves that the combined prediction method proposed in this study is feasible.For the combined model,two standard deviation denoising can improve the prediction accuracy of the model most effectively.The combination model proposed in this paper is easy to implement,has strong adaptive ability and high prediction accuracy,and is especially suitable for real-time control of water distribution device by water supply system.The combined forecasting model proposed in this paper not only provides a new method for short-term water supply forecasting,but also provides a new idea for other related research.The prediction methods mentioned in this paper are programmed and applied for software copyright,which lays a solid foundation for the practical application in the future.