Application of GA-BP neural network in prediction of chl-a concentration in Wuliangsu Lake

Abstract

The concentration of chl-a is one of the important criteria for water quality evaluation. The spectral data of water is closely related to the composition of water. The spectral reflectance data reflects a large amount of water quality information. Therefore, the health of water quality can be judged by analyzing the spectral reflectance. Aiming at the problem of slow convergence speed of neural network, genetic algorithm (GA) is used to optimize the initial parameters of the network, establish the relationship model between spectral reflectance, month and chl-a concentration in water, and analyze and predict the chl-a concentration in water. The results show that the determination coefficient of the optimal model for predicting chl-a concentration by GA-BP combined with monthly characteristics is 0.9561 and the root mean square error is 1.4751$\mu$G/L, the modeling effect is the best compared with other methods. Therefore, this method can realize the function of analyzing and monitoring water quality and meet the application requirements.