摘要
为解决现有的城市用水量短期预测BP神经网络法对初始权值敏感、易陷入局部极小点和收敛速度慢等问题,通过对城市时用水量数据特征的分析,应用基于全局随机优化思想的粒子群优化(PSO)算法对BP网络的初始权值进行优化,建立了PSO-BP城市时用水量预测模型.在算例分析中与传统BP神经网络预测法进行对比,发现该方法的收敛速度明显提高,且平均预测精度提高了2%,在用水量短期预测中非常有效.
In order to overcome the over-fitting problem and the local minima problem of the BP neural network method, the PSO-BP prediction model concerned on the hourly urban water consumption was developed. The model was based on the analysis of the characters of the hourly urban water consumption data and the particle swarm optimization （PSO） algorithms with the global stochastic optimization idea. The experimental results indicated that the average prediction precision increased by 2 per cent, compared to the traditional BP method. It was also shown that this model was faster in computation and had better generalization performance, which proved to be effective in shortterm prediction of urban water consumption.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2007年第9期165-170,共6页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(5027806250578108)
关键词
城市用水量
短期负荷预测
粒子群优化
BP神经网络
预测模型
urban water consumption
short-term load prediction
particle swarm optimization
BP neural network
prediction model
