Evaluation and Predication of Groundwater QualityBased on Fuzzy Model and BP Neural Networks

Abstract

Groundwater quality is an important indicator to be considered for utilizing water resources; as a result, when developing and protecting water resources, evaluation and predication on the water quality shall be taken as the reference. In this paper, analysis for the over-standard contents in groundwater was conducted for groundwater in Xi"an by applying multiple super-scale method and pollution index method; in addition, fuzzy pattern recognition model was also introduced so as to make comprehensive quality evaluation for the groundwater, then the evaluated results were compared with that of the existed comprehensive index method andprobabilistic neural network model; finally, predication was carried out for progressing trends of the four major pollutants by BP neural networks. The results reveal that total solids (TS), total hardness (TH) and nitrate radical exceeded the standard higher; while the over-standard ratios for fluoride, hexavalent chrome and chloridate content stayed at the average, and low over-standard ratio was found for phenol and arsenic, and pollution indexes of all pollutants were decreasing with fluctuations. Furthermore, Polluted areas were mainly at the urban area and sewage irrigated area in northern suburb. In conclusion, temporal evaluation for groundwater based on fuzzy pattern recognition method proves that groundwater quality in Xi"an had been obviously improved in the duration 1985-2011; the evaluation was more reliable compared with the other two methods. Besides, water quality predication in accordance with BP neutral networks indicates that optimal subordinate degree for most groundwater in Xi"an within 2012-2017 was of level II, which implies that groundwater quality was getting better.