基于粒子群优化随机森林的煤矿涌(突)水水源机器学习识别技术

Machine learning recognition technology of coal mine water inflow(outburst)source based on particle swarm optimization random forest

随着华北型煤田不断向深部开采,突水水源的威胁日益严重,涌(突)水事故不时发生。从宿南矿区的3种含水层中提取了120个水样,基于6个地下水常规离子数据,划分了100个训练集和20个测试集,采用粒子群算法对随机森林的参数进行优化,并结合随机森林建立了水源判别模型。研究结果表明,单一随机森林的测试集综合准确率为85%,粒子群优化的随机森林模型(PSO-RF)测试集水源判别综合精确度为95%。PSO-RF判别模型远高于传统的单一RF模型。因此,提出PSO-RF水源判别模型以提高识别精度有助于矿井水害精准预防与控制。

With the continuous deep mining of North China type coalfields,the threat of water inrush sources is becoming increasingly serious,and water inrush accidents occur from time to time.120 water samples were extracted from three aquifers in Sunan mining area.Based on six conventional ion data of groundwater,100 training sets and 20 test sets were divided.Particle swarm optimization algorithm was used to optimize the parameters of random forest,and a water source discrimination model was established in combination with random forest.The results show that the comprehensive accuracy of the test set of single random forest is 85%,and the comprehensive accuracy of water source discrimination of the test set of particle swarm optimization random forest model(PSO RF)is 95%.The PSO-RF discriminant model is much higher than the traditional single RF model.Therefore,the PSO RF water source discrimination model is proposed to improve the recognition accuracy,which is conducive to the precise prevention and control of mine water disasters.