Research Of Water Consumption Prediction Method Based On Nonlinear Function

During the optimal control in city water supply, the short-term urban water consumption prediction is gist. The method and the effect directly affect the execution and the benefit of the optimal control. At present, the short-term forecasting methods are always based on regression analysis and time series analytical methods inside and outside of the country. Recently, it occurs some reports related the Artificial Neural Network prediction method for urban water consumption prediction. But in these models and methods, it exists some problems, such as estimation of the parameters includes certain fuzzy factors itself,complexity, taking time relatively in calculating, quite low forecasting precision, and so on. In this paper, the author has studied numbers the methods of urban short-term water consumption prediction and put forward new predicting methods for daily and hourly burden of water supply. The paper proposes Back Propagation Network, Radial Basis Function Network and Support Vector Machines on urban water consumption prediction. First, the paper set up forecast model on urban daily and hourly water quantity and its influential factors, based on the characteristic of urban water quantity, Then, use BP Network, RBF Network and SVM explain the prediction model. RBF network has such advantageous properties as the independence of output on initial weight value and the adaptation for determining the structure. The theory of the SVM algorithm is based on statistical learning theory, the proposed algorithm embodies the Structural Risk Minimization principle, it is more rapidness, generalized performance and accurate. Analysis of the experimental results proved that the model of urban water consumption prediction is feasible, the BP network, the RBF network and SVM all can get the satisfied result. By analyzing the precision of advanced kinds of prediction methods, the author bring forward the linear combination prediction methods and non-linear combination prediction methods in the urban water consumption.