基于DBO-SVM的采空区煤自燃危险性预测

Prediction of Spontaneous Combustion Fire in Goaf Based on DBO-SVM

为实现采空区煤自燃危险的准确预警,避免自燃火灾的发生,引入蜣螂优化(dung beetle optimizer,DBO)算法与支持向量机(support vector machine,SVM)算法相结合的煤自燃预测模型。选取O_(2)、N_(2)浓度等十种指标作为自燃预测输入指标,自燃危险性等级作为输出指标,对所建模型进行训练,以四种分类性能评价指标检验模型的预测性能和精度,同时将DBO-SVM模型分别与DBO优化反向传播神经网络(BPNN)模型、粒子群优化算法(PSO)、优化BPNN神经网络模型以及SVM模型的预测结果进行对比分析。结果表明:DBO-SVM模型准确率相较于DBO-BPNN、PSO-BPNN、SVM模型分别提高了1333%、20%、3333%。将DBO-SVM模型应用于山西晋牛煤矿工作面煤自燃预测,该模型能快速准确地对不同矿井采空区煤自燃危险性进行预测,表明DBO-SVM模型相较于其他模型更具普适性和稳定性,更适合钻孔自燃预测。

To achieve accurate warning of spontaneous combustion fire in goaf and avoid it,a coal spontaneous combustion prediction model combined with dung beetle optimizer(DBO)and support vector machine(SVM)algorithm is introduced.Ten indexes such as O_(2) and N_(2) are selected as the input indexes of spontaneous combustion prediction,and the spontaneous combustion risk level is used as the output index to train the established model.Four classification performance evaluation indexes are used to test the prediction performance and accuracy of the model.Meanwhile,the pre-diction results of DBO-SVM model,back propagation neural network(BPNN)model optimized by DBO,BPNN neural network model optimized by particle swarm optimization algorithm(PSO)and SVM model are compared and analyzed.The results show that the accuracy of DBO-SVM model is increased by 1333%,20%and 3333%respectively compared with DBO-BPNN、PSO-BPNN and SVM models.Finally,the model is applied to the prediction of coal spontaneous combustion in the working face of Jinniu Coal Mine in Shanxi Province.The prediction results show that the DBO-SVM model can quickly and accurately predict the risk of spontaneous combustion fire in goaf of different mines,indicating that the DBO-SVM model is more universal and stable than other mod-els,and that it is more suitable for the prediction of spontaneous combustion of boreholes.