PSO RBF神经网络在城市需水量预测中的应用

为克服径向基函数（RBF）神经网络由于参数选取不当而对其收敛性能的干扰，利用粒子群优化算法（PSO）的全局搜索能力对RBF神经网络的三个参数进行寻优，建立了基于PSO RBF神经网络算法的城市需水量预测模型。结果显示，PSO RBF神经网络算法拟合某市1998～2007年需水量数据的平均相对误差为0.18%，预测2008～2010年需水量数据的平均相对误差为3.84%，耗时1.2 s；通过RBF神经网络算法拟合的误差平均值为0.28%，预测的平均相对误差为5.62%，耗时2.1 s，表明PSO RBF神经网络算法具有更高的收敛速度与精度。

Application of PSO RBF Neural Network in Urban Water Demand Prediction

In order to eliminate the impact of inappropriate RBF neural network parameters for interfering convergence, particle swarm optimization (PSO) with global search ability is applied to optimize three parameters of RBF neural network. And then the forecast model of urban water demand is established based on PSO RBF neural network. The results show that the fitting average relative error of water demand for a certain city is 0.18% from 1998 to 2007 and the forecasting average relative error is 3.84% from 2008 to 2010, the computational time consuming is 1.2s for PSO RBF neural network algorithm while the corresponding data is 0.28%, 5.62% and 2.1s for RBF neural network algorithm, respectively. It indicates that PSO RBF neural network algorithm has higher convergence and precision than RBF neural network algorithm.