Work

Joint Prediction of Goaf Temperature and CO Concentration Using Multi-Source Monitoring Feature Fusion and CA-WOA-Optimized Models

Year: 2026

Type: article

Abstract: Variations in goaf temperature and CO concentration during the low-temperature oxidation stage are affected by residual-coal oxidation, air-leakage oxygen supply, gas generation, migration and dilution, and ventilation disturbance, leading to multi-source coupling and nonlinear response characteristics. To continuously characterize this process, this study proposes a joint prediction method for goaf temperature and CO concentration based on multi-source monitoring feature fusion and Covariance-Adaptive Whale Optimization Algorithm (CA-WOA)-optimized models. Field monitoring data, including goaf-pipe gas, working-face-side gas, upper-corner gas, and return-air-side gas, were integrated to construct a daily-scale multi-source monitoring dataset. Pearson correlation analysis and random forest-based feature importance ranking were combined to identify dominant variables and construct top-k feature subsets. CA-WOA was then developed from the original Whale Optimization Algorithm by incorporating rank-weighted elite-center reconstruction, covariance-adaptive direction learning, WOA random-search injection, and geometric step-size decay. A composite normalized error of temperature and CO concentration was used as the fitness function to optimize the hyperparameters of RF, XGBoost, LightGBM, CatBoost, LSSVM, and TABM under a unified evaluation criterion. The results show that CA-WOA improved the overall predictive performance of all six models. Compared with the corresponding unoptimized baselines, the maximum increase in Mean R² was 0.076, and the maximum reduction in the composite fitness F was 22.8%. CA-WOA-TABM achieved the best performance, with a five-fold average Mean R² of 0.924 and F of 0.273. Its out-of-fold R² values for temperature and CO concentration were 0.928 and 0.931, respectively, demonstrating stable internal validation performance. SHAP and PDP analyses identified GoafPipe_CH4, GoafPipe_CO2, GoafPipe_C2H6, and GoafPipe_O2 as key variables, indicating nonlinear response relationships associated with gas generation, oxygen supply, and gas migration in the goaf during continuous monitoring.

Source: Applied Sciences

DOI: 10.3390/app16157422

Authors: Gang Yuan, Li Ma, Pengyu Zhang, Longcheng Zhang, Zhuoyang Lu, Yue Cao

Institutions: Xi'an University of Science and Technology, Jiangxi University of Science and Technology

Language: English

Topic: Geoscience and Mining Technology

Subfield: Safety, Risk, Reliability and Quality

Field: Engineering

Domain: Physical Sciences

SDG: Climate action

Open Access status: gold

Estimated APC paid: $2,896