Research And Implementation Of Prediction System For The Time,Space,and Energy Grade Of Microseismic Events Incorporating 3BNet

In preventing induced destructive mine earthquakes and rockburst disasters,predicting the occurrence time,space,and energy grade of microseismic events is a prerequisite and basis for prevention and control.Especially,accurate quantitative prediction of time,space,and energy grade of microseismic events is the prerequisite for studying disaster occurrence mechanisms,taking preventive measures in advance,reducing property losses and avoiding casualties.Coal is the main energy source in China.Even under the"dual carbon"strategy,coal will still account for 46%by 2030.Shallow coal resources are gradually being depleted,and deep mining is increasing with harsher environmental conditions and greater risk of rockburst,which threatens coal mine safety.The main research content of this article is as follows:Firstly,this article focuses on the research of predicting the occurrence time,space location,and energy grade of microseismic events.Addressing the problem of insufficient generalization and accuracy of traditional prediction models,a prediction network based on 3BNet(Butterworth Filter,Bidirectional Gated Recurrent Unit,and Bayesian Optimization)is proposed for quantitative prediction of the time,space,and energy grade of microseismic events.This method starts from a data-driven perspective,first using Butterworth filtering to extract the time,space,and energy grade of microseismic events’sequence data features.Secondly,bidirectional gated recurrent units are used to extract implicit data features.Finally,the hyperparameters of the entire3BNet are optimized using the Bayesian optimization algorithm.This article evaluates the prediction performance of this method through comparative experiments in the 501working face of a coal mine.The results show that because 3BNet can automatically update the model structure based on new knowledge,the prediction accuracy and stability of the proposed method are better than traditional time series prediction models,avoiding the reduction of model accuracy over time.Secondly,this article conducted a demand analysis,detailed design,system implementation,and system testing of prediction system for the time,space,and energy grade of microseismic events.The system is implemented using a client/server architecture:the client is used to implement the prediction algorithm,including the hyperparameter optimization module,data analysis module,model training module,and time,space,and energy grade prediction module;the server is used to implement data management,including microseismic data and user information data.Test results show that the system has implemented all functional requirements and operates stably.Finally,this article applied the designed 3BNet to 5 working faces in 2 coal mines on site,and proposed evaluation metrics suitable for predicting the time,space,and energy grade of microseismic events.The experimental results show that 3BNet is suitable for predicting the time,space,and energy grade of microseismic events,and the model performance has significant effects on the evaluation metrics of Median AE,RMSE,MAE,and R~2.In terms of time and space location prediction,this article calculates the model accuracy under different errors and defines that the accuracy increase within 1%is approaching stability,and the error value that first starts to approach stability is the accuracy of 3BNet.In the prediction of time,the average accuracy rate reached 86.76%(precision of±1.1h);in the prediction of space,the average accuracy rate reached 94.74%(precision of±60m,direction x),95.03%(precision of±70m,direction y)and 90.99%(precision of±9m,direction z)respectively.In the prediction of energy grade,the levels were divided into 1-7 levels,and the average accuracy rate reached 95.39%,the precision rate reached 95.74%,the recall rate reached 95.01%,and the F1 value reached 95.36%after statistics.Finally,through precise prediction in practical applications,two large energy microseismic events with levels of 6.39 and 6.86 were accurately predicted.The above experimental results verify the feasibility and advancedness of 3BNet for predicting the time,space,and energy grade of microseismic events.