Prediction For Gas Emission Quantity Based On Compensation Algorithm Of Multiple Regression Analysis And RBF Neural Network

Gas disaster is one of the important factors that threaten our safe production of coal mining process.The potential impact of coal seam gas becomes more significant with the mining depth increases. Mining face gas emission quantity is an important basis for mine gas prevention and ventilation design, predicting the amount of gas emission accurately can be effective in preventing the occurrence of multiple gas disaster. Therefore, strengthening the research capacity of gas emission forecasting methods and techniques to improve the accuracy of prediction of gas emission has positive significance to improve the safe production of coal.Affected by many factors of gas emission rate, multiple linear regression analysis to determine the amount of gas emission of mathematical analysis and its influence factors, using the analytical formula can not only according to one or several variables predict or control the value of another variable and can influence factor analysis to determine which are the important factors, which are the secondary factors; however, the amount of gas emission is controlled by many factors, is a gray, nonlinear and complex dynamic system, a single linear regression analysis can be a good response in the nonlinear part, neural network has very strong nonlinear approximation ability, can forecast of gas mining working face good emission nonlinear part. RBF neural network is a feed-forward neural network is efficient, has the best approximation properties and global optimization characteristics of the network do not possess, and has the advantages of simple structure, fast training speed. RBF neural network can overcome the long convergence time of BP network learning, easy to fall into local minimum problem. In order to utilize the advantages of two kinds of model of volume, the prediction of gas emission is better, it presents a modified model of multiple regression residuals RBF neural network analysis.First, the sample data to establish the amount of gas emission of multiple linear regression model, the model is used for gas emission prediction residuals are preliminary, emission prediction results of calculation and actual gas; then the residual as the dependent variable, the amount of gas emission factors of raw data as independent variables, forecasting the residual using RBF neural network; finally the use of prediction results of RBF residual gas emission prediction results are compensation. In the SPSS 19 environment, predicted by the model, the predicted results accuracy has been greatly improved, the results showed that:the multiple regression analysis type is compensated by RBF neural network, the linear fitting algorithm and nonlinear fitting algorithm together for gas emission prediction is an excellent algorithm.