Research Of Water Quality Evaluation Of The Sanhe Rivers In The Central Area Of Chengdu Based On BP-ANN

Evaluation of river water quality is to divide the level of pollution and determine the type of pollution, by evaluating the degree of river pollution. It can determine the degree of river pollution, forecast the development trends and provide the directional and principled programs and basis of water environmental protection.There are certain irrationalities and limitations exist in the traditional methods currently used in China’s water quality evaluation. Artificial Neural Network (ANN) can simulate some basic characteristics of the human brain. It has the advantages of massively parallel processing, distributed storage, adaptability, fault tolerance, etc. ANN can avoid the limitations of traditional evaluation methods. The learning ability and the memory functions of ANN provide a theoretical basis to solve the kind of problems of pattern recognition and classification of water quality evaluation.Based on the field survey of geomorphology, geology and surface water system in Chengdu City, combined with ANN theory, and based on the actual monitoring situations and pollution features of the Sanhe Rivers in the central area of Chengdu, a three-layer BP-ANN model with only one hidden layer has been adopted in this paper. DO, CODMn, BOD5and NH3-N are selected as the input nodes of the model. The only one output node of the model is water quality grade of evaluation results. Node numbers of hidden layer is determined as9by calculating the minimum test error and the network structure of the BP-ANN model of the water quality evaluation is4-9-1. The integrated and staging water quality grades of the Sanhe Rivers in the central area of Chengdu from the year2006to2010are evaluated by the BP-ANN model. Evaluation results show that the water quality grades of the monitoring sections of the Sanhe Rivers in the central area of Chengdu are similar in the five years, and the water quality grades are mostly concentrated in Class II-Class III, and the water quality of each section is relatively stable. In the space domain, the quality of upstream sections is better than which of downstream sections during one year. In the time domain, water quality of wet seasons is the worst, which of dry seasons is closed to common seasons, but which of dry seasons and common seasons is slightly better than the status of wet seasons during one year.Tow methods of water quality evaluation, BP-ANN method and Single Factor method, are compared. Comparison results show that:the water grade evaluated by BP-ANN model is much better than which evaluated by Single Factor method. Single factor method of water quality evaluation is a pessimistic method. Because various water quality factors are considered, results of BP-ANN evaluation method are more objective and accurate,Combined with the evaluation results, the pollution causes of the Sanhe Rivers in the central area of Chengdu are preliminarily analyzed, and the remediation measures are proposed.This study can provide a theoretical basis for pollution prevention and control of the Sanhe Rivers in the central area of Chengdu, and can also provide a useful reference for other river water environmental protection.