Study of rainwater quality assessment model based on radial basis function artificial neural network

In order to solve the problems existing in water quality comprehensive assessment model and method, the paper establishes K-means dynamic clustering algorithm applicable to rainwater quality assessment. By adding inferior class V, water quality standard based on the Surface Water Environmental Quality Standard(GB3838-2002) generates training samples and testing samples through random uniformly inserted values between every two assessment standards. Normalization processes the training samples and the testing samples, where the model have one network input layer with 6 nodes,one output layer with 1 nodes and one hidden layer whose nodes number can be automatically determined by network training. The model output is a continuous variation value, which not only satisfies the demand of water quality assessment but also has quantitative evaluation effect. It is used to assess the water quality of drinking water source based on rainwater harvesting in Xifeng District Qingyang City Gansu Province . The assessment result shows that the water quality of different underlying surface lies between III~V. Compared the assessment results obtained by principal component analysis method, we find that the RBF-ANN model output assessment result is scientific, reasonable and intuitive.