Research On Prediction And Early Warning Of Gas Diaster Based On ISSA-LSSVM

With the deepening of coal mining,the danger of coal and gas outburst and other gas disasters also tends to become more serious,seriously restricting the sustainable development of coal mine safety,efficient and reliable prediction and early warning technology is the key and foundation to enhance the ability of coal mines to prevent and control gas disasters.Using artificial intelligence algorithms such as intelligent optimization algorithms and machine learning algorithms,an in-depth study of coal mine gas disaster prediction and early warning methods is carried out.In view of the complexity,non-linearity,and uncertainty of coal and gas outburst disasters,a coal and gas outburst prediction model with nuclear principal component analysis and ISSA-LSSVM is proposed.Based on the analysis of coal and gas outburst occurrence conditions,processes,and the correlation between multiple factors causing protrusion,an index system is established to guide the protrusion prediction in accordance with the scientific principles in the relevant regulations.A kernel principal component analysis method was introduced to reduce the dimensionality of the multidimensional index data while retaining a valid amount of information on the predicted indexes.The least-squares support vector machine with excellent performance on nonlinear,small-sample problems is selected as the base prediction algorithm,and the sparrow search algorithm is introduced to optimize its key parameters.To address the shortcomings of the sparrow search algorithm,which is prone to local optimality and slow convergence,an improved sparrow search algorithm is proposed that incorporates an elite chaotic backward learning strategy,a random following strategy,and a Corsi-Gaussian variational strategy.The test results and the Wilcoxon signed-rank test results validate the excellent search capability of the improved algorithm for 10 benchmark test functions with different characteristics.The outburst prediction model is validated by an engineering example,which show that it has higher prediction accuracy than other models and can achieve accurate prediction of coal and gas outburst hazards.In order to improve gas concentration prediction accuracy and warning efficiency,a gas concentration prediction and warning method with variable modal decomposition and ISSA-LSSVM is proposed.The gas concentration time series are subjected to variable modal decomposition to minimize the influence of other noise on the monitoring data.The decomposed gas concentration multicomponent is used as input to the ISSA-LSSVM model,and the multicomponent is regressed and reconstructed separately to obtain the predicted gas concentration values.The early warning method is based on the predicted gas concentration values and the statistical analysis of the actual mine monitoring data,and the early warning indicators are selected and classified into different warning levels according to the different levels of risk.The results show that the gas concentration prediction model has the ability to fit the non-linear data well,which provides a solid foundation for the subsequent warning;it verifies the effectiveness and practicality of the gas concentration warning method and provides theoretical and methodological support and decision-making assistance for gas disasters prevention and control.The paper has 31 pictures,23 tables and 126 references.