智能化综采工作面煤层顶底板形状预测技术研究

Prediction Technology of Roof and Floor Shape of Coal Seam in Intelligent Fully Mechanized Mining Face

煤矿综采工作面顶板和底板的起伏状态预测是工作面智能化开采的关键技术之一。为解决煤层顶底板形状预测误差较大的问题,基于采煤机记忆截割技术的基本原理,构建了一种基于卷积长短时记忆网络(CONV-LSTM)的综采工作面顶底板形状预测模型。首先,利用已采取区域采煤机走向的空间信息和回采走向的时间信息,提取出顶底板的形状特征;其次,利用所提取的形状特征,构建了CONV-LSTM模型;最后,利用青龙寺煤矿5-20109工作面煤层形状数据对模型进行评估,现场测试表明,顶板形状的预测平均误差为3.5 cm,底板形状的预测平均误差为5.8 cm。结果表明,CONV-LSTM模型可实现顶底板形状的精准预测,满足工程中对采煤机前后滚筒调高的需求,对于实现综采工作面智能化开采具有重要意义。

The prediction of the undulation state of the roof and floor in fully mechanized mining face of coal mine is one of the key technologies for the intelligent mining.In order to solve the problem of large prediction error of roof and floor shape of coal seam,based on the basic principle of memory cutting technology of shearer,a prediction model of roof and floor shape of fully mechanized mining face based on Convolutional Long Short-Term Memory Network(CONV-LSTM)was constructed.Firstly,the shape features of the roof and floor were extracted using the spatial information of the strike of the regional shearer and the time information of the mining strike.Then,the CONV-LSTM model was constructed using the extracted shape features.Finally,the model was evaluated using the coal seam shape data of 5-20109 working face in Qinglongsi Coal Mine.The field test shows that the average prediction error of the roof shape is 3.5 cm,and the average prediction error of the floor shape is 5.8 cm.The results show that the CONV-LSTM model can realize the accurate prediction of the roof and floor shape,and meet the demand for the height adjustment of the front and rear drums of the shearer in the project,which is of great significance for realizing the intelligent mining of the fully mechanized mining face.