基于PSO-SVM的煤层底板突水危险性预测研究

发布时间：2018-03-08 21:25

本文选题：底板突水　切入点：支持向量机　出处：《山东科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：我国华北型煤田底板突水问题普遍严重,事故的发生会造成重大的人员伤亡和财产损失,而风险预测和评价是矿井水害防治的一个重要环节,也是实现安全开采的基本前提和重要基础。本文通过对煤矿突水预测现状的研究,结合煤层底板突水的非线性特征,选用支持向量机进行煤矿突水危险性预测。支持向量机是基于统计学习理论发展起来的一种新型机器学习算法,具有较强的泛化能力,适用于解决突水预测这样的非线性、小样本问题。但是,支持向量机的泛化能力和预测精度受到惩罚参数和核函数参数等相关参数的影响,针对支持向量机预测模型参数难以确定的问题,通过对种群的初始化设置、适应度函数和终止条件的设置不断更新粒子速度和位置,从而对影响支持向量机性能的相关参数进行寻优,得到基于粒子群算法的改进支持向量机预测模型。在PSO-SVM突水预测模型的应用过程中,首先分析研究区域矿井的地质、水文地质条件,选取影响煤层底板突水的主控因素(即隔水层厚度、水压、底板破坏深度、含水层和断层落差),然后搜集典型突水工作面的历史数据资料,并将这些数据分为训练集和测试集两部分;以MATLAB2014a为实验平台,结合Microsoft Visual C++编译器在MATLAB软件中添加Libsvm工具箱,通过代码编程对训练集数据进行仿真训练与测试,得出支持向量机的最优惩罚因子C和核函数参数σ分别为694.8591和317.1063;将测试集数据代入训练好的支持向量机模型,对工作面突水危险性进行预测,并将PSO-SVM模型的预测结果与突水系数法的预测结果、实际情况进行对比分析。实验结果表明,PSO-SVM模型在突水预测中具有较高的精度与工程应用推广价值,对保证煤矿安全生产具有重要的意义。
[Abstract]:North China coal mine water inrush accident occurred serious problems, will cause serious casualties and property losses, and prediction and risk evaluation is an important link of water disasters, but also an important basis for safe mining. This paper studies the status quo of water inrush prediction, nonlinear characteristics combined with the water inrush from coal seam floor, support vector machine is used to predict the risk of water inrush in mine. The support vector machine is based on statistical learning theory developed a kind of new machine learning algorithm, has strong generalization ability, suitable for solving such nonlinear prediction of water inrush, the small sample problem. But the effect of support vector machine the generalization ability and prediction accuracy by the penalty parameter and kernel function parameters, model parameters for support vector machine is difficult to determine, through to the population Initialization, fitness function and termination condition set update the particle velocity and position, so as to optimize the relevant parameters affecting the performance of support vector machine, improved particle swarm optimization algorithm based on support vector machine prediction model. In the process of PSO-SVM application of water inrush prediction model, the first analysis of the regional geology research mine, hydrogeological conditions, main control effect of selection of coal seam floor water inrush factors (i.e. aquifuge thickness, pressure, depth of destroyed floor, aquifer and fault throw), and then collect the typical water inrush in working face of historical data, and these data are divided into training set and test set of two parts based on MATLAB2014a experimental platform; Visual C++, with Microsoft compiler to add the Libsvm toolbox in MATLAB software, the programming code of the training set data for training and testing of simulation, the support vector machine is the most Optimal penalty factor C and kernel function parameters were 694.8591 and 317.1063; the test set data into the trained support vector machine model to predict the risk of water inrush in working face, and the prediction results the prediction results of PSO-SVM model and water inrush coefficient method, comparatively analyzed the actual situation. The experimental results show that the precision, and engineering application value of PSO-SVM model is higher in water inrush prediction, which is important to ensure the safety production of coal mine.

【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD745.2

【参考文献】