瓦斯涌出量自适应预测模型研究
杨超¹，周文铮²，刘雨竹²
（1. 合肥工业大学，安徽 合肥 230000；2. 辽宁工程技术大学，辽宁 葫芦岛 125000）
摘要： 为有效预防和减少煤矿瓦斯灾害，提高对回采工作面瓦斯涌出量预测的精度，提出了耦合优化的自适应粒子
群算法与人工蜂群算法整合的瓦斯涌出量预测模型。通过数据预处理对原始瓦斯涌出因素进行维数约简，结合二次寻优
变为建立线性方程和损失函数进一步简化计算，代入耦合优化函数实现对瓦斯涌出数据特征向量的提取，并将它作为
LS-SVM 的输入。应用非线性调整惯性权重H的控制策略，更新特征解的最优位置及其适应度值，对LS-SVM 高斯核参
数σ和正则化参数γ寻优，建立自适应耦合优化算法瓦斯涌出量预测模型。结果表明，预测值的平均相对误差仅为
2.594%，相较于原优化算法和数据未处理的预测模型，实现全局搜索和局部搜寻性能的有效平衡，具备更好的泛化能力
和预测准确度。
关键词： 矿井瓦斯；预测模型；粒子群算法；人工蜂群算法
中图分类号： TD712.5；TP181     文献标志码： A 文章编号： 2095-0802-(2023)04-0011-06
Adaptive Prediction Model of Gas Emission
YANG Chao¹, ZHOU Wenzheng², LIU Yuzhu²
(1. Hefei University of Technology, Hefei 230000, Anhui, China; 2. Liaoning University of Engineering and Technology,
Huludao 125000, Liaoning, China)
Abstract: In order to effectively prevent and reduce coal mine gas disasters and improve the prediction accuracy of gas emission
in mining face, a gas emission prediction model integrating coupling optimization algorithms of adaptive particle swarm
optimization and artificial bee colony is proposed. Through data preprocessing, the dimension of the original gas emission factors is
reduced, combined with the method of quadratic optimization into the establishment of linear equation and loss function to further
simplify the calculation, and bring into the coupling optimization function to extract the feature vector of gas emission data, which
is used as the input of least squares support vector machine (LS-SVM). The control strategy of nonlinear adjusting inertia weight H
is applied to update the optimal position and fitness value of characteristic solution, optimize the Gaussian kernel parameter σ and
regularization parameter γ of LS -SVM, and establish the gas emission prediction model of adaptive coupling optimization
algorithm. The results show that the average relative error of the prediction value is only 2.594%. Compared with the original
optimization algorithm and the prediction model without data processing, it realizes the effective balance between global search
and local search performance, and has better generalization ability and prediction accuracy.

Key words: mine gas; prediction model; particle swarm optimization; artificial bee colony