摘要
为提升煤与瓦斯突出预测准确度,减小数据缺失对煤与瓦斯突出预测的不利影响,提出1种基于链式多重填补马尔科夫链蒙特卡罗(MCMC)的麻雀搜索算法(SSA)优化支持向量机(SVM)预测模型。根据突出影响因素选取模型参数,运用MCMC对突出事故缺失值进行数据填补,采用SSA优化SVM,建立MCMC-SSA-SVM模型对填补后数据集进行预测,验证MCMC填补有效性和SSA优化性能;分别构建SVM、SSA-SVM、PSO-SVM、GAM-SVM、CMC-SVM、MCMC-PSO-SVM和MCMC-GA-SVM这7种模型进行突出预测,对比预测准确度,分析MCMC-SSA-SVM、MCMC-PSO-SVM和MCMC-GA-SVM的适应度。研究结果表明:MCMC填补后准确度均提升7.89个百分点以上,SSA的优化性能强于PSO和GA,MCMC-SSA-SVM预测准确度最高,为97.37%,泛化能力优于对比模型。研究结果可为煤与瓦斯突出预测研究提供借鉴和参考。
To improve the accuracy of coal and gas outburst prediction and reduce the adverse effect of missing data on coal and gas outburst prediction,a sparrow search algorithm(SSA)optimized support vector machine(SVM)prediction model based on chain multiple filling Markov chain Monte Carlo(MCMC)was proposed.The model parameters were selected according to the influencing factors of outburst,and the MCMC algorithm was applied to fill in the missing values of outburst accidents,then SSA was used to optimize SVM.A MCMC-SSA-SVM model was established to predict the filled data set,and the effectiveness of MCMC filling and the optimization performance of SSA were verified.Seven models,namely SVM,SSA-SVM,PSO-SVM,GA-SVM,MCMC-SVM,MCMC-PSO-SVM and MCMC-GA-SVM,were constructed respectively for outburst prediction to compare the accuracy,and the adaptability of MCMC-SSA-SVM,MCMC-PSO-SVM and MCMC-GA-SVM were analyzed.The results show that all the accuracies of MCMC after filling increase by more than 7.89 percentage points,the optimization performance of SSA is stronger than those of PSO and GA,MCMC-SSA-SVM has the highest prediction accuracy of 97.37%,and the generalization ability is better than the comparison models.The results can provide reference for the research on coal and gas outburst prediction.
作者
邵良杉
高英超
SHAO Liangshan;GAO Yingchao(Institute of Systems Engineering,Liaoning Technical University,Huludao Liaoning 125105,China;Liaoning Institute of Science and Engineering,Jinzhou Liaoning 121000,China)
出处
《中国安全生产科学技术》
CAS
CSCD
北大核心
2023年第8期94-99,共6页
Journal of Safety Science and Technology
基金
国家自然科学基金项目(71771111)。