An intelligent control strategy based on improved LSTM prediction model for improving gas extraction efficiency in boreholes

In order to break through the technical bottleneck that current coal mines are unable to accurately and dynamically increase the gas concentration in essence. This paper carries out a systematically research on the intelligent control technology for enhancing the extraction concentration and efficiency in boreholes from various research perspectives, such as theoretical analysis, control strategy construction and engineering application. First, the “gas-air” migration model in coal seam fissures is constructed from the perspective of the air leakage mechanism in boreholes. And the numerical simulation of gas extraction effect under different extraction conditions is carried out. Subsequently, the CNN (Convolutional Neural Network) and Attention mechanism are introduced to optimize the LSTM (Long Short-Term Memory) model. The intelligent control strategy for enhancing gas concentration in boreholes is proposed. Finally, the proposed control strategy in this paper is applied in field practice to verify the strategy’s reliability. The study results indicate that there is an obvious correlation between concentration and negative pressure, and the extraction concentration’s attenuation rate is obviously different with varying negative pressure conditions. Different arrangements for the extraction data will lead to significant discrepancies in the model’s prediction results. For predicting the extraction negative pressure, “Pressure-Wave” is the optimal data arrangement method. Among the many intelligent prediction models, the LSTM prediction model has the highest match with the extraction data characteristics. The CNN-LSTM-Attention negative pressure prediction model obtained by introducing CNN and Attention can effectively overcome this limitation for LSTM model which is overly dependent on time series. In addition, it can also enhance the predictive model’s ability to capture the gas concentration attenuation law under different extraction parameters. After applying the intelligent control strategy for enhancing gas concentration in the field, the three test boreholes showed the value of concentration increase by more than 5% compared to the pre-control period and the relative increase reached as high as 25%. And the borehole’s effective extraction time has increased by approximately one month.