Study on the BP-GA model and its application in water quality assessment

Improving the convergence speed and global searching ability of Back Propagation network (BP) always occupies a significant and sophisticated subject. This paper improves the intelligent model by coupling the Genetic Algorithm (GA). Specific improvements include initializing multiple sets of BP's weights and thresholds, setting a smaller BP training epoch and building the fitness function by inputting verification data for BP. The process of building the BP-GA model is explained in detail. Take the water quality assessment in Baiyangdian wetland as an example. The output of this model is compared with the one by using the fuzzy synthetic evaluation method. Result shows that the trained BP-GA model can be effectively used to assess the water quality.