基于灰色关联分析的GA-BP网络需水预测模型研究

针对城市需水预测涉及因素众多、不同地区影响因子不尽相同且多寡不一及影响因子的选择直接决定需水量预测的结果与实际是否相符等问题,提出了灰色关联分析法、遗传算法和BP神经网络相结合的需水预测模型,并以南京市为例,通过灰色关联分析法筛选出主要影响因素,采用遗传算法优化BP神经网络,构建基于灰色关联分析的GA-BP神经网络需水预测模型。实例应用结果表明,该模型用于需水预测能够比较全面地考虑需水量影响因子,与传统BP网络相比,GA-BP网络预测精度更高,训练速度更快,可作为资料时间序列较短情况下一种较好的需水预测方法。

Study of GA-BP Neural Network for Water Requirement Prediction Based on Gray Correlation Analysis

Urban water requirement prediction normally involves many factors, which are different from place to place, and the choice of influencing factors will directly decide whether the results of water requirement prediction match the actual or not. Hence, this paper proposed a water requirement prediction model which combined the gray correlation analysis method, genetic algorithm (GA) and BP neural network. Taking Nanjing City as an example, based on gray correlation analysis, the main factors are screen out, and the BP neural network optimized by genetic algorithm was employed to establish GA-BP neural network water requirement prediction model. The results show that the established model takes a comprehensive consideration of water demand influencing factors, and it is prior to that of the traditional BP neural network in terms of accuracy and training speed. Therefore, the proposed model can provide a better water demand forecasting method in the case of short time series of data.