Research On Fuzzy Cognitive Maps For Mine Strata Pressure Behavior Prediction

Coal resource is the largest share of China’s energy structure,which provides a steady stream of power for the rapid development of China’s economy and society.Therefore,the safe mining of coal resources is particularly important.In the process of coal mining,the abnormal appearance of mine strata pressure seriously threatens the personal safety of operators,so it is of great significance to carry out the research on the prediction of mine strata pressure behavior and establish the early warning mechanism,which is an important technical guarantee for coal mine safety production.In this paper,primarily,based on the actual demand of mining prediction in mining activities,a practical prediction model based on fuzzy cognitive maps was proposed.After that,the Real Coded Genetic Algorithm,which was used in the prediction model,was improved,weight regularization and dropout regularization were introduced,and the Improved Genetic Algorithm was proposed to improve the ability to find optimal solutions.At last,taking field monitoring data as an example,Real Coded Genetic Algorithm,L2-Regularization Genetic Algorithm and Improved Regularization Genetic Algorithm were adopted to train the MPP model.MSE(Mean Square Error)and Fitness Function were used to evaluate the applicability of FCM model with different optimization algorithms,and verifying Improved Genetic Algorithm is better than the other two algorithms.Finally,the performance of the FCM model based on the improved genetic algorithm is compared with two different neural network architecture models(the dense connection network and the recurrent neural network)to further verify the applicability of the proposed prediction model and prove that it can meet the demand of mine pressure behavior prediction in the production site of coal mine.The innovative achievements of this paper mainly include the following aspects:1.The fuzzy cognitive map is applied to the prediction research of mine pressure behavior.According to the requirements of actual work,the input and output attributes of the model are determined by feature engineering,and the prediction model is constructed.2.An improved genetic algorithm is proposed to improve the ability of the algorithm to find the optimal solution in the predefined solution space,so as to improve the prediction performance of the model;3.By comparing the FCM model in this paper with the two neural network architecture models mainly used in the regression task,the preliminary study on the introduction of deep learning into the prediction of mine pressure behavior was completed.