【摘要】 针对城市需水量预测中时间序列的非线性特性及传统BP网络预测收敛速度慢易陷入局部极小值等问题,将Chaos理论和BP神经网络理论相结合,提出了一种基于Chaos-BP理论的城市短期需水量COBP(Chaos Back Propagtion)预测模型。利用重构相空间的嵌入维数确定COBP网络的结构,通过混沌优化搜索,找到BP神经网络权值的全局最优值,并对其输出的"尖点"预测值进行混沌参数控制,实现城市短期需水量的预测。仿真分析表明,与传统预测模型相比,COBP预测模型所需训练数据样本少,收敛速度快、易达到全局最小值,预测结果整体误差的指标良好,呈现良好的综合预测性能。更多还原

【Abstract】 In view of the nonlinear characteristics of time series in urban water demand forecasting and the problem of slow convergence rate of traditional BP neural network and local minimum,the Chaos theory and BP neural network theory are combined.This paper presents a forecasting model of short term urban water demand COBP(Chaos Back Propagation)based on Chaos-BP theory. The structure of COBP network is determined by embedding dimension of reconstructed phase space,the global optimal value of BP neural network weight is found by chaos optimization search,and chaotic parameter control is carried out on the prediction value of "cusp" of BP neural network,so that the forecast of urban short-term water demand can be realized. The simulation results show that compared with the traditional prediction model,the COBP prediction model requires fewer training data samples,faster convergence speed,easier to reach the global minimum value,and the overall error index of the prediction results is good,showing a good comprehensive prediction performance.更多还原