基于计算智能的煤矿自然发火预警研究

发布时间：2018-04-04 14:22

本文选题：粒子群　切入点：支持向量机　出处：《中国矿业大学》2015年博士论文

【摘要】：自然发火预警(包括煤层自然发火危险性评价)是煤矿安全领域的研究热点之一。本论文发挥多学科交叉的优势,从自然发火预警研究和应用的实际需要出发,引入计算智能领域的新理论、新方法(例如支持向量机、粒子群算法、粗糙集算法等),开展了以下几方面的研究工作。1)粒子群(PSO)算法的改进。针对标准PSO算法难以收敛于全局最优解及易于发生早熟收敛等问题,利用混沌映射和一种基于粒距的权值动态改变策略,提出了一种改进的粒子群优化算法(IPSO)。利用3种标准函数分别对IPSO算法和另外4种PSO算法进行了测试比较,结果表明:IPSO算法具有较高的全局寻优能力,其精度优于另外4种比较算法。2)基于IPSO的支持向量机(SVM)参数优化方法。SVM参数对SVM预测精度有重要影响。为此,本文研究了利用IPSO进行SVM参数优化的方法,并分别提出了IPSO-v-SVR算法和IPSO-SVC算法。在UCI基准数据集上分别对这2种算法及多种回归或分类算法进行了测试比较,结果表明:IPSO-v-SVR和IPSO-SVC的精度均优于相应的比较算法。3)基于IPSO-v-SVR的煤层瓦斯含量预测和基于IPSO-SVC的煤层自然发火危险性预评价。在煤层瓦斯含量数据集上的测试结果表明:在小样本条件下,利用IPSO-v-SVR算法进行煤层瓦斯含量预测可以取得较高的预测精度。在煤层自然发火危险性数据集上的测试结果表明:利用IPSO-SVC算法进行煤层自然发火危险性预评价可取得较高的分类精度。此外,还对数据集之外的2个开采煤层进行了自然发火危险性的预评价,评价结果与矿井实际发火情况相符。4)粗糙集(RS)理论与IPSO-SVC算法的融合及其在采空区自然发火预警中的应用。建立了基于瓦斯抽采钻孔气体分析的采空区自然发火预警指标体系,以RS作为IPSO-SVC模型的预处理器提出了RS-IPSO-SVC算法,该算法通过对指标体系进行属性约减,去掉冗余特征属性,从而降低了样本维度。在采空区自然发火状态数据集上进行了算法测试,结果表明:RS-IPSO-SVC算法的分类准确率优于比较算法,能够较好地适用于采空区自然发火预警。5)《煤矿自然发火信息管理及预警系统》的开发。针对煤矿企业对自然发火信息管理及预警的实际需求,以及智能算法在应用过程中对数据更新的需要,基于.NET技术开发了B/S框架的“煤矿自然发火信息管理及预警系统”。该系统已在峰峰集团羊东煤矿部署应用。通过该系统的实施,进一步丰富了已有的相关数据集,并多次成功识别出与自然发火隐患相关的异常现象,取得了良好的应用效果。通过上述研究得出了以下几点主要结论:⑴改进粒子群算法(IPSO)具有较强的寻优能力。⑵利用IPSO可显著提高SVM参数的寻优效果。⑶IPSO-v-SVR、IPSO-SVC以及RS-IPSO-SVC算法能够分别适用于煤层瓦斯含量预测、煤层自然发火危险性预评价以及采空区自然发火预警,且精度较高。
[Abstract]:The early warning of spontaneous combustion (including coal seam spontaneous combustion risk assessment) is one of the research hotspots in the field of coal mine safety.Based on the practical needs of the research and application of spontaneous combustion early warning, this paper introduces new theories and methods in the field of computational intelligence (such as support vector machine, particle swarm optimization algorithm).Rough set algorithm has been studied in the following aspects: 1) the improvement of particle swarm optimization (PSO) algorithm.Aiming at the problem that the standard PSO algorithm is difficult to converge to the global optimal solution and is prone to premature convergence, a modified particle swarm optimization algorithm is proposed by using chaotic mapping and a dynamic change strategy of weights based on particle spacing.The IPSO algorithm and the other four PSO algorithms are tested and compared with each other by using three standard functions. The results show that the IPSO algorithm has a high global optimization ability.Its precision is better than the other four comparison algorithms. 2) the parameter optimization method of support vector machine (SVM) based on IPSO. The parameters of SVM have important influence on the prediction accuracy of SVM.Therefore, this paper studies the optimization method of SVM parameters using IPSO, and proposes IPSO-v-SVR algorithm and IPSO-SVC algorithm respectively.The two algorithms and various regression or classification algorithms are tested and compared on the UCI datum data set.The results show that the accuracy of% IPSO-v-SVR and IPSO-SVC is better than the corresponding comparison algorithm .3) the prediction of coal seam gas content based on IPSO-v-SVR and the pre-evaluation of coal seam spontaneous combustion risk based on IPSO-SVC.The test results on the data set of coal seam gas content show that the prediction accuracy of coal seam gas content can be obtained by using IPSO-v-SVR algorithm under the condition of small sample.The test results on the data set of coal seam spontaneous combustion risk show that high classification accuracy can be obtained by using IPSO-SVC algorithm to pre-evaluate the coal seam spontaneous combustion risk.In addition, two coal seams outside the data set have been pre-evaluated for the risk of spontaneous combustion.The evaluation results are in accordance with the actual mine fire. 4) the integration of rough set theory and IPSO-SVC algorithm and its application in the spontaneous combustion early warning in goaf.The index system of spontaneous combustion early warning in goaf based on gas analysis of gas drainage borehole is established. The RS-IPSO-SVC algorithm is proposed by using RS as the preprocessor of IPSO-SVC model. The algorithm reduces the attributes of the index system and removes the redundant feature attributes by using RS as the pre-processor of the IPSO-SVC model.Thus, the sample dimension is reduced.The algorithm is tested on the data set of spontaneous combustion in goaf. The result shows that the classification accuracy of the RS-IPSO-SVC algorithm is better than that of the comparison algorithm.It can be applied to the development of coal mine spontaneous combustion information management and early warning system.In view of the actual demand of coal mine enterprises for the management and early warning of spontaneous combustion information and the need for data updating in the process of application of intelligent algorithms, a "coal mine spontaneous combustion information management and early warning system" based on .NET framework is developed.The system has been deployed in Yangdong Coal Mine of Fengfeng Group.Through the implementation of the system, the existing related data sets are further enriched, and the abnormal phenomena related to the hidden trouble of spontaneous combustion are successfully identified many times, and good application results have been obtained.Through the above research, the following main conclusions are drawn: the improved particle swarm optimization algorithm of 1: 1 has strong optimization ability .2 using IPSO can significantly improve the optimization effect of SVM parameters. 3IPSO-v-SVRV IPSO-SVC and RS-IPSO-SVC algorithm can be applied to coal, respectively.Bed gas content prediction,The coal seam spontaneous combustion risk pre-evaluation and goaf spontaneous combustion warning, and the accuracy is high.
【学位授予单位】：中国矿业大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TD752.2

【相似文献】