基于粒子群算法优化的城市供水量预测模型研究

鉴于BP神经网络、RBF神经网络在城市供水量预测精度上的不足，利用粒子群算法优化两者相关参数，实现更高预测精度，并通过建立BP神经网络、RBF神经网络、PSO-BP神经网络、PSO-RBF神经网络分别对城市供水量数据进行仿真预测。最终测试样本统计结果显示：RBF神经网络比BP神经网络平均相对误差（MRE）低约1%，在拟合度（R2）上高约0.014；PSO-BP神经网络比BP神经网络在MRE上降低约1.25%，在R2上提高约0.05；PSO-RBF神经网络比RBF神经网络在MRE上降低约0.3%，在R2上提高约0.072。由此说明RBF神经网络比BP神经网络在城市供水量预测方面更有优势，并且利用粒子群算法优化神经网络模型参数可有效提升神经网络预测精度。

Research on the Prediction Model of Urban Water Supply Based on Particle Swarm Optimization

In view of the shortage of BP and RBF neural networks in the prediction accuracy of urban water supply， the particle swarm optimization algorithm is used to optimize the related parameters to achieve higher prediction accuracy， and BP neural network， RBF neural network and PSO-BP are established. The neural network and PSO-RBF neural network respectively simulate and predict the urban water supply data. The final test sample statistics show that the RBF neural network is about 1% lower than the BP neural network average relative error （MRE）， and about 0.014 higher than the fitness （R2）； PSO-BP neural network ratio the BP neural network was reduced by about 1.25% on the MRE and about 0.05 on the R2； the PSO-RBF neural network was reduced by about 0.3% on the MRE than the RBF neural network， and the R2 was improved on R2 by 0.072. Therefore， it is shown that RBF neural network has more advantages than BP neural network in urban water supply forecasting， and using particle swarm optimization algorithm to optimize the parameters of neural network model can effectively improve the prediction accuracy of neural network.