飞蛾火焰优化算法-投影寻踪回归模型在需水预测中的应用

基于用水时间序列构建投影寻踪回归(PPR)需水预测模型。针对PPR模型矩阵参数难以确定的不足,利用一种新型群体智能仿生算法——飞蛾火焰优化(MFO)算法优化PPR模型矩阵参数,提出MFO-PPR预测模型,并构建MFO-BP模型作对比,以1980—2013年上海市需水预测为例,分别利用实例前20组和后10组数据对模型参数进行率定及预测。结果表明:MFO-PPR模型对实例后10 a需水预测的平均相对误差绝对值和最大相对误差绝对值分别为1.84%、4.20%,预测精度优于MFO-BP模型的2.06%、4.61%。MFO算法具有较好的全局寻优能力,将MFO算法应用于PPR模型参数寻优,可有效地提高PPR模型的预测精度。

Application of Projection Pursuit Regression Model Optimized by Moth-Flame Optimization Algorithm in Prediction of Water Demand Prediction

Based on water use time series,a projection pursuit regression (PPR) water demand forecasting model was constructed.Aiming at the problem that the PPR model matrix parameters are difficult to be determined,a new group intelligent bionic algorithm-Moth-Flame Optimization (MFO) algorithm is used to optimize the PPR model matrix parameters.The MFO-PPR model is proposed.Compared with the MFO-BP model,taking the water demand prediction of Shanghai from 1980 to 2013 as an example.The parameters of the first 20 groups and the last 10 groups were used to estimate and predict the model parameters respectively.The results show that: the absolute value of the average absolute error and the maximum relative error of water demand forecast of MFO-PPR model in the last 10 years of example are 1.84% and 4.20%,respectively,indicating that the prediction accuracy is better than that of MFO-BP model.MFO algorithm has better global optimization ability.The MFO algorithm is applied to the optimization of PPR model parameters,which can improve the prediction precision of PPR model.