Forewarning of sustainable utilization of regional water resources: a model based on BP neural network and set pair analysis

Abstract

Early warning for sustainable utilization of regional water resources is an important control measure for regional water security management. To establish operable and quantitative forewarning model, in this paper, a new forewarning model for sustainable utilization of water resources based on BP neural network and set pair analysis (named BPSPA-FM for short) was established. In the proposed approach, the accelerating genetic algorithm–based fuzzy analytic hierarchy process was suggested to determine the weights of evaluation indexes, back-propagation neural network updating model was used to predict the values of the evaluation indexes, and the set pair analysis was used to determine the function values of relative membership in variable fuzzy set of the samples. BPSPA-FM was applied to early warning for sustainable utilization of regional water resources of Yuanyang Hani terrace in Yunnan Province of China. The results show that the states of sustainable utilization in this system were near the critical value between nonalarm and slight alarm from 1990 to 2000, the states of the system fell into slight alarm and were rapidly close to intermediate alarm from 2001 to 2004, and the states of the system were predicted to be near the critical value between slight alarm and intermediate alarm from 2005 to 2010. The main alarm indexes of the system were utilization ratio of water in agriculture, control ratio of surface water, per capita water supply, per unit area irrigation water and per capita water consumption. BPSPA-FM can take full advantage of the changing information of the evaluation indexes in adjacent periods and the relationship between the samples and the criterion grades. The results of BPSPA-FM are reasonable with high accuracy. BPSPA-FM is general and can be applied to early warning problems of different natural hazards systems such as drought disaster.