基于IDBO-HKELM-Adaboost的煤与瓦斯突出危险性预测方法  

A Method for Predicting the Risk of Coal and Gas Outburst Based on IDBO-HKELM-Adaboost

作　　者：李曼 徐耀松[1] 王雨虹[1] 王丹丹 LI Man;XU Yaosong;WANG Yuhong;WANG Dandan(School of Electrical and Control Engineering,Liaoning Technical University,Huludao Liaoning 125105,China;School of Mechanical Engineering,Liaoning Technical University,Fuxin Liaoning 123000,China)

机构地区：[1]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125105 [2]辽宁工程技术大学机械工程学院,辽宁阜新123000

出　　处：《传感技术学报》2025年第3期477-486,共10页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(51974151);辽宁省教育厅重点实验室项目(LJZS003);辽宁省教育厅辽宁省高等学校基本科研项目(LJ2017QL012);辽宁省教育厅科技项目(LJ2019QL015)。

摘　　要：为实现更加高效准确地完成煤与瓦斯突出危险性预测,提出了一种采用Adaboost算法增强的改进蜣螂算法(IDBO)优化混合核极限学习机(HKELM)的预测模型。首先,在数据降维时采用核主成分分析(KPCA)对影响因素进行处理并提取有效的特征量,得到预处理样本数据。将PWLCM混沌映射、非线性递减策略以及邻域学习机制融入到蜣螂算法中,之后,利用IDBO对HKELM的关键参数进行寻优,构建IDBO-HKELM煤与瓦斯突出危险性分类预测模型。最后,使用Adaboost算法对IDBO-HKELM模型进行增强。结合工程实际数据进行验证,验证结果表明:相较于其他模型,基于IDBO-HKELM-Adaboost的预测方法具有更高的预测精度,在提高运算效率的同时满足煤与瓦斯突出预测的精度和可靠性要求,准确率达到97.44%。To achieve a more efficient and accurate completion of coal and gas prominence hazard prediction,a prediction model using the improved dung beetle optimizer(IDBO)enhanced by Adaboost algorithm to optimize the hybrid kernel extreme learning machine(HKELM)is proposed.First,kernel principal component analysis(KPCA)is used to process the influencing factors and extract effective feature quantities during data dimensionality reduction to obtain pre-processed sample data.The PWLCM chaotic mapping,nonlinear decreasing strategy,and neighborhood learning mechanism are incorporated into the dung beetle algorithm,and the pre-processed sample data are trained and tested for IDBO performance.The key parameters of HKELM are optimized by using IDBO,and the IDBO-HKELM coal and gas prominence hazard classification prediction model is constructed.The validation results show that the prediction method based on IDBO-HKELM-Adaboost has higher prediction accuracy than other models,and meets the requirements of accuracy and relia-bility of coal and gas prominence prediction with an accuracy rate of 97.44%while improving the computing efficiency.

关 键 词：煤与瓦斯突出 突出预测 改进蜣螂算法 混合核极限学习机 核主成分分析 预测模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP212[自动化与计算机技术—控制科学与工程]