Research On Mine Outburst Prediction System Based On SFES-PSO-BP Algorithm

There are many factors influencing the mine's outburst.The mine's outburst prediction system presents nonlinearity,strong ambiguity,and the related models are difficult to determine and establish.It is difficult to meet the high-precision prediction requirements by the traditional BP neural network model.Aiming at the difficult problem of mine gas and mine gas outburst prediction in mine outburst prediction system,in order to further improve mine prediction and precision,a mine outburst prediction method based on SFES-PSO-BP(Similarity-based Fitness Estimation Strategy-Particle Swarm Optimization-Back Propagation)algorithm is proposed.Based on the PSO-BP algorithm,the SFES-PSO-BP algorithm uses the known particle information to obtain the fitness function value.In the valuation strategy,the similarity evaluation method between particles is introduced to increase the fitness function value.The accuracy of the valuation,the number of calculations of the fitness function value are reduced,and the performance of the algorithm is improved.A nonlinear dynamic prediction system model with fast convergence,high convergence precision and strong robustness is obtained.The improved algorithm model is applied to the outburst of coal and gas in the mine outburst prediction system and the outburst prediction of mine water.The performance of the prediction model based on SFES-PSO-BP algorithm is better than that based on PSO-BP and BP neural network prediction model.The main research contents of this paper include the following:Firstly,this paper systematically analyzes the operating mechanism of BP neural network.BP neural network can better simulate the complex nonlinear relationship of mine outburst prediction system,and can solve the inaccuracy of nonlinear problems predicted by traditional physical and mechanical models,but BP neural The network itself has a slow convergence rate,easy to fall into the local extremum,and can not get the global optimization and other defects.This paper analyzes the BP neural network optimization method and analyzes the process of particle swarm optimization(PSO)optimization BP neural network.Secondly,through the improvement and upgrade of the particle swarm optimization(PSO)algorithm for BP neural network,a mine outburst prediction model based on SFES-PSO-BP algorithm is proposed.The design idea is as follows:Because most particles are highly dispersed in the early stage of training.The state of motion,the parameter ? in the PSO algorithm is improved based on the adaptive weighting method and the learning factor method based on asynchronous change for c1 and c2,so that problems such as easy premature convergence and slow convergence in the early stage of the particle can be improved.It can ensure that the particles do not fly through the optimal solution in the early stage of training,and ensure that the particle group can be quickly reduced to a certain small range in the shortest time;for the PSO-BP algorithm model in the late iteration of the iterative period and the oscillation around the optimal value In this paper,we analyze it.In the method of PSO algorithm to optimize BP neural network parameters,in order to increase the diversity of particle swarm and prevent it from falling into local extremum,in order to reduce the complexity of particle swarm calculation time,this paper introduces particle-based similarity.Degree of fitness value evaluation strategy to enhance particle activity,improve its local search ability and The ability to optimize and improve the precision of convergence and global search capability,improved its prominent effect on the prediction of mine.Finally,the mine outburst prediction model based on SFES-PSO-BP algorithm is applied in mine water inrush prediction,and compared with PSO-BP algorithm prediction model and BP neural network prediction model.The results show that SFES-PSO-BP algorithm is based on SFES-PSO-BP algorithm.The mine outburst prediction model has higher nonlinear goodness of fit and prediction accuracy.The problem of over-fitting and oscillation near the optimal value of PSO-BP algorithm is overcome,and the convergence speed of the algorithm and the stability of the algorithm are greatly improved.