ABSTRACT

Temperature is the key factor influencing the spontaneous combustion of coal, but it is difficult to obtain accurate temperature data because of the complex physical environment of the mining area. A mathematical model relating coal temperature to CO concentration was derived from data collected from a low-temperature oxidation experiment. Subsequently, a model is established that uses a genetic algorithm to select and optimize penalty factor C and kernel function parameter g of a support-vector regression model (GA-SVR). Taking O₂, CO₂ and C₂H₆ as independent variables, the GA-SVR model is then employed to calculate CO concentration. This predicted CO concentration is then used to calculate coal temperatures and assess the risk of spontaneous combustion. The performance of the GA-SVR model is compared with standard SVR, random forest and back propagation neural network models. The results demonstrate that the GA-SVR model has superior accuracy and generalization capabilities. This model can be used to predict coal temperatures within mines and provide an early warning for spontaneous combustion.

摘要

温度是影响煤自燃的关键因素，但由于矿区物理环境复杂，难以获得准确的温度数据。从低温氧化实验收集的数据推导出煤温度与 CO 浓度相关的数学模型。随后，建立了一个模型，该模型使用遗传算法来选择和优化支持向量回归模型（GA-SVR）的惩罚因子C和核函数参数g。取O ₂、CO ₂和C ₂ H ₆作为自变量，然后采用GA-SVR模型计算CO浓度。然后使用该预测的 CO 浓度来计算煤的温度并评估自燃的风险。将 GA-SVR 模型的性能与标准 SVR、随机森林和反向传播神经网络模型进行比较。结果表明，GA-SVR 模型具有优越的准确性和泛化能力。该模型可用于预测煤矿内的煤温，并为自燃提供早期预警。