基于数据挖掘的煤矿安全风险评价体系研究

The accuracy of the coal mine safety risk evaluation is important to the coal mine production management. Now, the Chinese coal mine safety risk evaluation system has wide examining scope and many indexes, so it is very difficult which is executed, furthermore, it has many qualitative indexes lead to limitations and subjectivity, and then affect the accuracy of evaluation. The coal mine safety risk originates from coal and gas, mine water, discharge and mine fire, furthermore, the coal mine enterprise has monitoring systems of them, and the mining based on the history monitoring data can direct the coal mine safety evaluation, and then we can build the coal mine safety risk evaluation system. In this paper we research the coal mine safety risk evaluation system based on coal and gas, mine water, discharge and mine fire, and build the coal mine safety risk evaluation model by using the multiple algorithms of data mining and improved algorithms ,and develop the software based on the algorithms, which it can direct the coal mine safety evaluation.

煤矿安全风险评价的准确性对于煤矿生产管理具有重要意义。我国现有的煤矿安全风险评价体系考查范围广，评价指标多，往往在煤矿企业全面实施起来比较困难，并且体系中包含很多定性指标，具有一定的局限性和主观性，影响评价的准确性。煤矿安全风险主要来自煤与瓦斯、矿井涌水和矿井火灾，而且现有煤矿企业都已具有三者的监测系统，通过对历史监测数据的挖掘，将为煤矿安全风险评价提供决策依据，进而建立基于煤与瓦斯、矿井涌水和矿井火灾的煤矿安全风险评价体系。本课题将系统地研究基于煤与瓦斯突出、矿井涌水和矿井火灾的煤矿安全风险评价体系及相关理论，综合运用多种数据挖掘算法及其改进算法建立煤矿安全风险评价模型，并开发基于这些算法和方法的软件包模型，支持煤矿安全风险评价。

本项目按照预定研究计划分析研究了国内外煤矿安全风险评价体系与评价方法，建立了煤与瓦斯突出、矿井涌水和矿井火灾3个主题的评价体系。针对每个评价体系的因素特征及数据的分布特点，分别应用基于几何轮廓相似度的多变量决策树、互学习神经网络和C45算法建立预测模型；针对煤与瓦斯采集数据的缺失、噪声等问题，研究设计了高效的k质心组合聚类数据预处理模型。研究发现煤与瓦斯突出、矿井涌水和矿井火灾在每次煤矿安全风险评价基本事件中的影响程度是不同的，因此研究了基于3个主题的动态集成学习模型，并针对不同地质条件的煤矿企业进行煤矿安全风险评价模型的应用研究。