摘要:
针对基于支持向量机（SVM）的煤与瓦斯突出预测算法存在预测精度和可靠性不高，选择核函数时未考虑非线性数据的分类，对非线性分布的煤与瓦斯突出影响因素提取效果较差的问题，提出了一种将改进的粒子群(IPSO)算法与Powell算法相结合（IPSO-Powell）优化SVM的煤与瓦斯突出预测算法。首先通过灰色关联分析提取出煤与瓦斯突出主控因素，即瓦斯放散初速度、瓦斯压力、开采深度、瓦斯含量和煤体破坏类型，作为算法的输入样本；然后运用IPSO算法改善粒子群算法（PSO）的早熟收敛性，结合Powell算法进行局部搜索得到最优解，对SVM算法的惩罚系数和高斯核函数参数进行寻优，得到SVM的最优参数组合；最后将煤与瓦斯突出的主控因素输入到SVM中进行分类，并将其与实际测试集分类结果进行对比，实现煤与瓦斯突出预测。仿真结果表明：与SVM算法、GA-SVM算法、PSO-SVM算法相比，利用IPSO-Powell优化SVM算法进行煤与瓦斯突出预测，具有更高的预测精度，同时提高了 SVM 求解过程的运算效率，能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95.9%。
Abstract:
In view of problems of coal and gas outburst prediction algorithm based on support vector machine(SVM) that prediction accuracy and reliability are not high, classification of nonlinear data is not considered when selecting kernel function, and extraction effect of influence factors of coal and gas outburst with nonlinear distribution is poor, a coal and gas outburst prediction algorithm which combines improved particle swarm optimization (IPSO) algorithm with Powell algorithm(IPSO-Powell) to optimize SVM was proposed. Firstly, main control factors of coal and gas outburst, namely initial velocity of gas emission, gas pressure, mining depth, gas content and failure type of coal body is extracted through grey correlation analysis and used as input samples of the algorithm. Then, IPSO algorithm is used to improve precocious convergence of particle swarm optimization (PSO), and Powell algorithm is used to search the local optimal solution, the penalty coefficient and Gaussian kernel function parameters of the SVM algorithm are optimized, the optimal parameter combination of SVM is obtained. Finally, the main control factors of coal and gas outburst are input to the SVM for classification , and compared with the actual test set classification results to achieve coal and gas outburst prediction. The simulation results show that compared with the SVM algorithm, GA-SVM algorithm and PSO-SVM algorithm, the application of IPSO-Powell optimized SVM algorithm for coal and gas outburst prediction has higher prediction accuracy, and improves the computational efficiency of the SVM solution process, which can meet the accuracy and reliability requirement of coal and gas outburst prediction with an accuracy rate of 95.9%.
百度学术
下载:
