基于径向基函数的城市日用水量预测方法  

Prediction Method for Based on Radial Basis Function for Daily Water Consumption

作　　者：张宏伟[1] 岳琳[1] 王亮[2] 

机构地区：[1]天津大学环境科学与工程学院,天津300072 [2]天津工业大学材料科学与化学工程学院,天津300160

出　　处：《天津大学学报》2006年第4期486-489,共4页Journal of Tianjin University(Science and Technology)

基　　金：国家自然科学基金资助项目(50278062;50578108).

摘　　要：结合城市日用水量影响因素的特点和变化规律,分析探讨了城市日用水量预测模型的求解方法．建立日用水量和其相关因素之间的预测模型,分别采用径向基函数(RBF)网络算法与支持向量机(SVM)回归法求解该预测模型．RBF网络具有结构自适应确定,输出不依赖初始权值的优良特性;SVM回归法采用结构风险最小化准则 (SRM),以统计学习理论作为理论基础,运算速度快,泛化能力强,预测精度高．通过分析验证的结果,证明了该日用水量预测模型的可行性,采用RBF和SVM两种求解方法均能得到满意的结果．Combined with the influencing factors and characteristics of municipal daily water demand, the solution to the consumption model was analyzed. Forecast model for municipal daily water consumption and its influencingfactors was set up, and then radial basis function （RBF） network and support vector machines （SVM） were adopted to solve the model. RBF network has such advantages that the output is independent the initial weight value and the structure can be determined SVM algorithm embodies the structural risk minimization by adaptation. Based on statistical learning theory, the （SRM） principle, which is more rapid more accurate, and has higher generalized performance. Analysis of the experimental results proves that the prediction model of municipal daily water consumption is feasible; the RBF network and SVM can both get satisfactory results.

关 键 词：城市日用水量 径向基网络 支持向量机 泛化能力 预测模型 

分 类 号：X321[环境科学与工程—环境工程] TU991.31[建筑科学—市政工程]