支持向量机在城市用水量短期预测中的应用

为解决现有的城市用水量短期预测人工神经网络法的过学习与局部极小点等问题,通过对城市时用水量数据特征的分析,在统计学习理论和结构风险最小化准则的基础上,建立了基于支持向量机(SVM)理论的城市用水量短期预测数学模型.在算例分析中与误差逆传播(BP)神经网络预测法进行对比,发现该方法的平均预测精度提高了2%,且具有收敛速度快、泛化能力强等优点,在用水量短期预测中非常有效.

Application of Support Vector Machines in Short-Term Prediction of Urban Water Consumption

In order to overcome the over-fitting problem and the local minima problem of the artificial neural network (ANN) method in short-term prediction of urban water consumption, a new mathematical model according to the support vector machines (SVM) theory was developed. The model was based on the analysis of the characters of the hourly urban water consumption data, the statistical learning theory (SLT) , and the empirical risk minimization (ERM) principle. The experimental results indicated that the average prediction precision increased by 2 percent, compared to the back propagation (BP) neural network method, and that this model was faster in computation and had better generalization performance, which proved to be effective in short-term prediction of urban water consumption.