Accurate prediction of coal mine gas emission is crucial for disaster prevention, yet challenging due to complex, non-stationary data and traditional models’ tendency to converge to local optima. The present study proposes a novel SHSCOA-BiLSTM model, which integrates an enhanced chimpanzee optimisation algorithm to optimise a bidirectional long short-term memory network. The methodology employs data imputation, principal component analysis, and enhanced global search strategies to tune critical hyperparameters. The model has been validated on real-world data, and it has been demonstrated to significantly outperform existing benchmarks, with a reduction in mean absolute percentage error of 57.18 − 74.10% and mean squared error of 80.16 − 92.35%. The findings indicate that the SHSCOA-BiLSTM model offers a highly accurate and robust instrument for gas emission forecasting, providing a reliable scientific foundation for early warning systems that can significantly enhance proactive safety management and prevent gas-related disasters in coal mines.
Jinzhang et al. (Sun,) studied this question.