Groundwater Level Prediction Based On RBF Neural Network

China is a country with a serious shortage of water resources.With the over-exploitation of human beings,the groundwater level has been continuously declining,bringing with it geological disasters such as ground subsidence and road deformation.Therefore,groundwater level prediction plays a very important role in real life.The current prediction methods have the disadvantages of large prediction error and slow convergence speed.In this study,temperature,humidity,evaporation and rainfall were selected as input parameters.The BP neural network and RBF neural network model were used to make a comparison of groundwater level prediction.The QPSO algorithm was used to optimize the RBF neural network to predict groundwater level.The main research content is as follows:(1)The network structure of BP neural network is 4-8-1,training algorithm is momentum gradient descent and activation function is tansig and purelin.The data is normalized and the BP neural network model is established to predict the groundwater level.The average relative error is 9.21%,the mean square error is 0.1426,and the training time is 18.95 s.The network structure of the BP neural network is determined as 4-8-1,the training algorithm is the momentum gradient descent method,and the activation function is tansig and purelin.The data was normalized and a BP neural network model was established to predict the groundwater level.The average relative error was 9.21%,the mean square error was 0.1426,and the training time was 18.95 s.(2)For the BP neural network prediction model,the prediction error is larger.The RBF neural network prediction model is used to determine the radial basis function as a Gaussian function,the training algorithm is a least-squares product algorithm and the expansion function is 0.05,and the groundwater level is performed.The prediction shows that the average relative error is 4.42%,the mean square error is 0.0432,and the training time is 14.73 s.The results show that the prediction accuracy and training time of RBF neural network are better than those of BP neural network,but the prediction error of RBF network is larger in individual period.(3)The QPSO algorithm is used to optimize the thresholds of the hidden layer of the RBF neural network,the weights of the hidden layer to the output layer,and the threshold of the output layer.The groundwater level was predicted by the optimized RBF neural network model,and the average relative error was 1.35%.The mean square error was 0.0039.The results show that the prediction accuracy of the optimized RBF neural network is better than that of the RBF neural network and BP neural network.Accuracy.(4)A simulation interface based on QPSO algorithm to optimize RBF neural network to predict groundwater level was designed.Through the layout of the interface and the programming of the control program,a simulation interface is established.Through the text of the prompts to operate,make it easier to operate,more convenient to watch the prediction graphics,to achieve human-computer interaction.The RBF neural network optimized by QPSO algorithm can effectively predict the groundwater level with high precision,and can provide a basis for agricultural irrigation and groundwater exploitation.