基于改进型极限学习机的新疆库尔勒市城市需水量预测  

Prediction of Xinjiang Kurle urban water demand based on improved limit learning machine

作　　者：司马义.阿不都热合曼 

机构地区：[1]新疆塔里木河流域巴音郭楞管理局,新疆库尔勒841000

出　　处：《水资源开发与管理》2017年第7期61-65,共5页Water Resources Development and Management

摘　　要：城市需水量系统具有大惯性、强耦合、非线性等特性,采用机理分析法,难以建立其准确的数学模型,导致预测效果差。鉴于此,该文将基于正交基函数的改进型极限学习机对城市需水量因子进行辨识,并利用经验模态分解方法确定网络隐含层节点数,建立了库尔勒市城市需水量预测模型。结果表明:模型有效性为0.9714,实测值与预测值的拟合关系比较理想,说明基于正交基函数的改进型极限学习机对城市需水量进行系统辨识是可行的。Urban water demand system is characterized by large inertia, strong coupling and nonlinearity, etc. Mechanism analysis method is adopted. It is difficult to establish an accurate mathematical model, thereby leading to poor forecast effect. Therefore, urban water demand factors are identified by the improved limit learning machine based on orthogonal basis functions in the paper. The empirical mode decomposition method is used for determining network hidden layer node quantity. An urban water demand prediction model in Korla is established. Results show that the model validity is 0. 9714. The fitting relationship between the measured value and the predicted value is more ideal. It is obvious that it is feasible to systematically identify urban water demand by improved limit learning machine based on orthogonal basis functions.

关 键 词：城市需水量 预测 极限学习机 经验模态分解 正交基函数 

分 类 号：TV213.4[水利工程—水文学及水资源]