Prediction of Gas Emission in Mining Face Based on GRA-ISA-LSSVM

Abstract

Coal, as a crucial primary energy source in China, provides a stable energy supply and security for the nation's economic and sustainable high-quality development. However, gas disasters can lead to severe casualties and significant economic losses. Therefore, accurately and efficiently predicting gas emission is of paramount importance and holds significant practical value for the safe production of high-gas mines. This paper employs the Least Squares Support Vector Machine (LSSVM) model, which offers better regression prediction accuracy for nonlinear problems with small datasets. To enhance the model's prediction accuracy and effectiveness, the Grey Relational Analysis (GRA) method is used to analyze the influencing factors of gas emission. Factors with high correlation are selected as model inputs. Experimental results show that coal seam gas content, coal seam depth, coal seam thickness, adjacent seam gas content, adjacent seam spacing, mining face length, and adjacent seam thickness have high correlation. To validate the results, Kendall's tau-b, Pearson, and Spearman correlation analysis methods are used, and the experimental results align with GRA. Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Improved Simulated Annealing Algorithm (ISA) are employed to find the global optimal solution for the LSSVM model parameters. Finally, the GRA-GA-LSSVM, GRA-PSO-LSSVM, and GRA-ISA-LSSVM models are constructed and tested. Experimental results show that after applying GRA, the RMSE and MAE metrics decrease, while the R² metric increases, proving that the GRA method optimizes the model. The GRA-ISA-LSSVM model has the lowest RMSE and MAE values and the highest R² value, indicating superior performance in static gas emission prediction.