Study On The Model Of Dissolved Oxygen Prediction Based On RBF Neural Network

There are many factors that induce the outbreak of aquatic disease and even large quantities of death and cultured water deterioration is the most important factor. There are many kinds of water quality parameters with the character of nonlinear and uncertain, the accurate prediction of water quality has become an urgent problem to be solved in the aquaculture industry. Therefore, in order to achieve the accurate prediction of water quality and the healthy growth of aquatic products, exploring the suitable methods of cultured water quality prediction has theoretical value and practical significance.RBF(Basis Function Radial, radial basis function) neural network algorithm has been developed rapidly in the field of water quality prediction due to its excellent performance. The water quality prediction model based on RBF neural network algorithm can not only overcome the shortcomings of the traditional water quality prediction method, but also can improve the accuracy of the prediction of water quality. However, the RBF neural network algorithm itself has some defects, and aquaculture water quality is changeable. Thus, water quality prediction model based on optimized RBF neural network is particularly important.For the RBF neural network, the determination of the connection weights of the hidden layer to the output layer has a close relationship with the prediction accuracy of the network. The change of hidden layer to the output layer of RBF neural network is linear, therefore, the recursive least square algorithm is used to train the weights of hidden layer to the output layer. It can effectively improve the convergence speed and prediction accuracy of the network. However, if the training samples are too much, this method may cause large matrix and pathological operation, and then it can not get the accurate prediction results. In order to remedy this shortcoming, the training process of the weights of hidden layer to output layer is optimized by integrating an improved recursive least squares algorithm and RBF neural network. In this paper, the optimization of water quality prediction model based on RBF neural network will be analyzed in detail by taking the important water quality parameter dissolved oxygen as the research object.In this paper, the main contents are divided into the following aspects:(1)The research introduced the basic theory of RBF neural network, analyzing the current situation of water quality prediction and the application of RBF neural network algorithm in water quality prediction;(2)Formulating theoretical knowledge of improved recursive least square algorithm and analyzing its convergence and enlarging its rang of application;(3)Applying improved recursive least square algorithm to the optimization of the value of the hidden layer to the output layer of the RBF neural network. The water quality prediction of RBF neural network based on improved recursive least square algorithm is constructed by taking dissolved oxygen as an example.(4) The prediction accuracy and the convergence rate of RBF neural network in water quality prediction is improved by the comparison and analysis of the results of the water quality prediction with the help of MATLAB.