Accurately predicting the evolution of microseismic energy is of great significance for identifying the potential operational risks and ensuring safe operation of underground gas storage (UGS) facilities. However, microseismic energy sequences usually have significant nonlinear and non-stationary characteristics, making it challenging for traditional prediction methods to effectively describe their complex fluctuation patterns. To address this issue, this study proposes a microseismic energy prediction method for UGS that integrates multi-scale decomposition with Deep Learning. Specifically, considering the nonlinear and non-stationary nature of the microseismic energy sequences, the Grey Wolf Optimizer (GWO) is employed to automatically tune the key parameters of Variational Mode Decomposition (VMD), namely the number of modes ( k ) and the penalty factor ( α ). The original energy sequence is thereby decomposed into multiple relatively stable sub-sequences. For each decomposed multi-scale component, a Deep Learning module based on a Bidirectional Long Short-Term Memory (BiLSTM) network with an Attention Mechanism is constructed to perform prediction, and the final energy forecast is obtained accordingly. Evaluated by mean absolute error ( MAE ), mean squared error ( MSE ), mean absolute percentage error ( MAPE ), and coefficient of determination ( R 2 ), the corresponding metrics are 0.968, 1.417, 3.982% and 0.968, respectively. The proposed model demonstrates superior predictive accuracy compared to existing models. Furthermore, its stability in terms of predictive performance has been validated through multiple experimental trials. • A microseismic energy forecasting model is proposed for underground gas storage. • Nonstationary energy sequences are decomposed by GWO-optimized VMD. • BiLSTM-Attention is employed to capture bidirectional and key temporal features. • The proposed method outperforms comparative models in multiple evaluation metrics.
Liu et al. (Fri,) studied this question.