基于智能算法的煤矿瓦斯浓度随钻预测模型对比研究

Comparative study on methane concentration prediction models in coal mines based on intelligent algorithms

瓦斯超前预测对于煤矿的安全生产至关重要,然而何种模型更加适宜井下随钻瓦斯预测目前仍无相关研究。该文在研究常用智能预测模型建模的基础上,通过曹家滩矿区某掘进工作面的实测数据,对比研究了灰色预测模型、LSTM模型和BP神经网络模型的随钻瓦斯预测性能,实验结果表明,3种模型在理想的煤矿井下实验条件下均具有良好的预测准确性,但LSTM的预测效果最佳且理想条件下的预测误差小于0.15%。该研究结果可用于高精度的随钻瓦斯浓度预测,并可为煤矿井下瓦斯抽采孔轨迹的动态调控提供必要的数据支撑。

Gas advanced prediction is crucial for the safety production of coal mines.However,there is still no consensus on which model is more suitable for underground drilling gas prediction.Based on this,this paper introduces the common prediction models,and through the comparison of the prediction performance of the grey prediction model,the LSTM model,and the BP neural network model using the actual measured data from a driving face in the Caojiatan mining area.The experimental results show that all three models have good prediction accuracy under the actual underground mining conditions,but the LSTM model has the best prediction performance,with an ideal prediction error of less than 0.15%.The research results of this paper can realize high-precision detection of gas concentration during drilling,and provide necessary data support for the dynamic control of gas drainage hole trajectories in underground coal mines.