摘要
城市日用水量预测是城市供水管网系统动态模拟的基础和前提,用神经网络对城市日用水量预测的非线性回归组合模型求解的过程中,提出了采用新型的仿生算法——蚂蚁算法来训练神经网络的权值.此方法简化了训练过程,避免了BP算法易陷于局部极值等问题.将经过蚂蚁算法训练的神经网络应用到S.X市日用水量预测模型中,显示了此网络模型具有良好的预测能力,验证了基于蚂蚁算法的神经网络在城市日用水量的预测中具有有效性和可行性.
Daily water demand forecasting is an important factor of affecting the precision of hydraulic models. Neural network is used to solve the nonlinear regression combination model of daily water demand forecasting, and a novel bionic algorithm, ant algorithm, is used to train neural network weights. Applying the trained neural network in the daily water demand forecasting model of SX city, the results show that very nice effects are obtained. This approach simplifies neural network training and overcomes the limitation of BP algorithm.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2005年第1期60-62,共3页
Journal of Harbin Institute of Technology
基金
国家十五科技攻关项目(2002BA107B05)
关键词
城市日用水量
神经网络
蚂蚁算法
自相关系数
随机数据序列
Backpropagation
Correlation methods
Forecasting
Genetic algorithms
Hydraulic models
Mathematical models
Neural networks
Nonlinear systems
Regression analysis
Water supply systems