The orthogonal basis NN based prediction modeling for river water quality

The East River is the source of drinking water for residents of Hong Kong and Shenzhen of China. The water quality of Huizhou-Dongan reach has a direct effect on freshwater quality provided to Hong Kong and Shenzhen. In accordance with the location of automonitors, two adaptive neural networks based predicting models of water quality for the river reach are put forward in this paper. One is that of anticipating the lower course water quality by measuring the upriver water quality. Another is estimating the future state with current water quality in a same position. The learning algorithms with orthogonal basis transfer function for static and dynamic neural networks are given. Both the neuron numbers and orthogonal basis transfer function can be established automatically in training process. The local extremum problem does not exist in the method. Simulation results prove that the proposed approaches have high precision, good adaptability and extensive applicability