Classification Of Surface Movement Observation Stations In Huainan Mining Area And Calculation Of Regional Predicted Parameters

With the rapid development of economy,coal,as the main energy source,has been continuously developed in large quantities.For a planned coal mine,it is very meaningful to choose suitable prediction methods and prediction functions according to its known geological mining conditions before mining to predict in advance the degree of rock deformation and surface movement deformation after mining.In this paper,the BP neural network method is used to predict the parameters of the probability integration method to provide parameter support for the subsequent prediction by the probability integration method.The main purpose of this paper is firstly to study the feasibility of using BP neural network to calculate the parameters of probability integration method in thick unconsolidated mining area,and secondly to determine whether optimizing BP neural network with multi-population genetic algorithm can improve the prediction accuracy.This paper introduces in detail the mining factors,geological factors and topographic factors that affect the prediction parameters of probability integration method,and collects the measured data of surface observation stations in 40 typical mining areas in China as training samples and testing samples.The measured data are mainly divided into two parts:one part is the geological and mining conditions of typical observation stations,and the other part is the probability integration method parameters of typical observation stations.When using BP neural network for prediction,this paper selects the first 37 mining area data from typical observation stations as training samples and the last 3 mining area data as test samples.When the BP neural network is used to predict the parameters of the probability integration method,there are too many influencing factors on the parameters of the probability integration method.In order to reduce the complexity of the BP neural network and increase the prediction accuracy,the principal component analysis method is used to process the parameters in order to obtain the main influencing factors of the probability integration method.In this paper,the BP neural network model is established by MATLAB software to predict the parameters of probability integration method According to the relative error of model prediction,it is feasible to predict the parameters of probability integration method by BP neural network in thick unconsolidated mining area.Since the initial weights and thresholds of the BP neural network model are randomly generated,the BP neural network model has the disadvantages of long calculation time,slow convergence speed and easy falling into local optimization.In order to solve this problem,this paper uses multi-population genetic algorithm to optimize the weights and thresholds of BP neural network,and establishes MPGA-BP neural network.The calculation shows that it is feasible to use MPGA-BP neural network to predict the parameters of probability integration method in thick unconsolidated mining area.Comparing the parameters of probability integration method predicted by principal component-BP neural network model and principal component-MPGA-BP neural network model respectively,it is found that both methods have better nonlinear generalization ability,so both methods are feasible in predicting the parameters of probability integration method.By comparing the relative error,it is found that the overall relative error of MPGA-BP neural network is small and the calculation result is relatively stable.Therefore,the use of multi-population genetic algorithm can indeed optimize the traditional BP neural network.The optimized model has significantly improved accuracy and stability.Figure[12]table[11]reference[65]...