Abstract: The nonlinearity and uncertainty of water environmental pollution make the traditional water quality evaluation methods have limitations． In order to improve the accuracy of water quality evaluation，the paper put forward a water quality evaluation model based on improved wavelet neural network ( WNN) ． The initial weights of the wavelet neural network was optimized based on adaptive genetic algorithm ( AGA) ，and then training the network was trained by using the wavelet neural network algorithm，and finally，the trained network was tested． The simulation results showed that the combination of adaptive genetic algorithm and wavelet neural network improved the efficiency of network training， and this method could be used in for model of water quality evaluation，and the evaluation result would have higher precision and accuracy．