Study On Water Demand Forecast Of Harbin

Water resources have a major meaning in the development of the city. Researching the water demand prediction methods in urban which is short of water to improve prediction accuracy and reduce the predicted risk, not only can treat the result as the basis of urban water supply system to optimize, and provide a reference for the planning and construction of water supply system, but also can improve the efficiency of water use, improve the urban ecological environment and promote the healthy and harmony development of social.On the basis of investigating and studying a large number of related documents and summary of previous research experience, this paper studied water demand forecasting methods of Harbin.According to the Statistical Yearbook and documentation, Harbin urban water demand was divided into ecological water demand, domestic water demand and production water demand. After analyzing the traditional water demand forecasting methods, three methods were selected. They are gray model, BP neural network and support vector machine.First, the raw data was tested if it met the gray forecast feasibility. The original data sequence passed the law of quasi-smooth and quasi-exponentially test. In the forecasting process, because the model prediction error made by gray model is relatively large, the results were needed to be corrected. After analyzing, the Markov chain was selected to adjust the predicted results. The results showed the improved gray model conquer a certain extent of the shortcomings of the gray model and the model prediction accuracy was significantly improved.When using BP neural network to predict, the original data was mapped to the interval [0.1,0.6] to make the model good extrapolation. Using the model to predict the ecological water demand, production water demand and living water demand, good results was achieved.When using support vector machine to predict, RBF kernel function was selected. Cross validation method was used to calculate the root mean square error of the model, and a wide range of search method was used to find the optimal parameters. Then as a basis to build the model of support vector regression machine to predict Harbin urban ecological water demand, domestic water demand and production water demand, the effects was good.In order to improve the accuracy of model predictions, and integrate the advantages of a variety of models, combination forecasting model was used. At the request of the weights, linear programming method which goal was minimizing the sum of the absolute value was used to obtain the optimal weights of the objective function, effectively prevent certain solving weight is not the optimal weight. Experimental results showed that combination forecasting model results was significantly better than a single prediction model, which effectively combines the advantages of several prediction models. So combination forecasting model was an effective prediction method.Finally, the predicted results were analyzed. For how to further improve the prediction accuracy, a number of recommendations was given according to the model predictions. And according to the forecast results, some recommendations were given to the relevant departments.The result of this paper can be provided to researchers of related fields as theoretical reference and expanded applied to other areas. In practice, it can provide the basis for the relevant departments of Harbin to make the implementation of scientific management and macro-guidance in order to take control policies timely and appropriately to provide effective protection for the required water supply of Harbin.