基于GABP技术的PCA弥苴河水质预测

Water quality prediction of Miju River based on PCA and GABP

针对传统的水质预测方法中由于因子的多重相关性而造成的预测精度偏低的问题,提出了一种将主成分分析法(PCA)和遗传算法优化的BP神经网络(GABP)相耦合的水质预测方法.利用主成分分析法提取对水质因子影响较强的综合成分,克服了传统水质预测方法中信息冗余的问题.在对大理弥苴河水质进行大量实际监测的基础上,分别采用PCA-GABP神经网络,GABP神经网络以及传统的BP神经网络3种模型的方法,建立了弥苴河水质高锰酸盐指数的的预测模型.通过数据预处理,筛选了600组数据进行训练学习和测试.通过对3个模型的预测误差分析对比,可以得出PCA-GABP神经网络预测模型精度更高.

This paper proposed a method of water quality prediction which combined Principlal Component Analysis（PCA） with Back-Propagation Neural Network optimized by Genetic Algorithm（GABP） in order to solve the problem of low prediction accuracy on account of multiple correlation factor in the traditional method of water quality prediction.Overcame the information redundancy by using Principlal Component Analysis,it extracted the variable component with strong influence.Three predition models for CODMn in Miju River were estab- lished respectively by using PCA-GABP neural network, BP neural network and GABP neural network on the basis of monitoring.600 groups of data were selected for learning and testing.It was approved that the model of water quality prediction based on PCA-GABP neural network could provide better accuracy.