Study On Eutrophication Model Of Bohai Bay Based On Fuzzy Theory And Neural Network

Along with the rapid development of economic round the Bohai Sea region and high exploitation of coast, seawater eutrophication problem of Bohai Bay becomes increasingly serious, which has heavily restricted sustainable development of coastal economic. Establishing of assessment model and prediction model for eutrophication will have great significance for prevention, repairing and management of coastal water eutrophication.Applying information technologies of statistics theory, fuzzy theory, neural network and genetic algorithms etc, and using MATLAB software package, this thesis emphasizes on the research of seawater eutrophication assessment model of Bohai Bay and prediction model of chlorophyll value by researching the relationship of environmental factors and phytoplankton biomass.Through analyzing main factors related to eutrophication in Bohai Bay and synthesizing physical, chemical and biological indicators, the method of comprehensive fuzzy assessment for seawater eutrophication is put forward. The results transformed by weighted-average principle show that eutrophication is serious near Tianjin port and the outlets. In order to avoid the effect of subjective factors,a method of fuzzy rules extraction based on database is applied, then the fuzzy logic prediction model of chlorophyll value is established.In addition, this thesis establishes seawater eutrophication assessment model based on BPNN of Bohai Bay. This model uses enough training samples for the BP neural network, and checks the generalization of network by testing samples. The difference of BP assessment results and fuzzy assessment results is not significance. Based on the research of genetic algorithms, a method of optimization of BP structure and initial values with two-grade genetic algorithms is proposed, and then a genetic neural network prediction model of chlorophyll value in red tide monitoring area of Bohai Bay is built in this thesis. The result is better by optimized prediction model than by traditional BP prediction model.This paper also studies on the combination of fuzzy logic and neural network- ANFIS (adaptive neuro-fuzzy inference system). The subtractive clustering is used to build initial fuzzy inference system with monitoring data; and then ANFIS is adopted to build prediction model of chlorophyll value. An appointed staion is simulated and most of the results are close to fact values.