Indicator analysis of lake water quality based on fuzzy neural network and spectrophotometry

This paper proposes a method of using fuzzy neural network modelling to predict water quality indicators to handle the difficult problems of detecting lake water quality. The multi-parameter water quality tester uses the principle of spectrophotometry to measure water quality parameters. It and a portable temperature and pH meter are used to measure 7 indicators of lake water quality. A total of 89 sets of data are obtained. The correlation coefficient method is used to screen the water quality indicators. Randomly select 73 sets of training data and 16 sets of prediction data to establish a network model. Using the model to predict the chemical oxygen demand (COD) of the training data, it shows that the correlation coefficient is above 0.9, and the relative error is controlled within 6%. Using the model to predict the COD of the prediction set data multiple times, the correlation coefficient is between 0.8-0.95, and the relative error is within 15%. The research result shows that the fuzzy neural network has high prediction accuracy and generalization ability for water quality indicators with a short prediction time. This is a method that can effectively predict lake water quality.