A physics-based probabilistic forecasting methodology for hazardous microseismicity associated with longwall coal mining

Mining-induced microseismicity is widely considered as a result of slippage of pre-existing critically stressed fractures caused by stress perturbations around an advancing face. An in-depth analysis of the recorded microseismicity associated with longwall top coal caving mining at Coal Mine Velenje in Slovenia has been previously carried out and reported by the authors. It has been concluded that while microseismic event rate is affected by mining intensity (longwall face daily advance rate) as well as local abundance of pre-existing fractures, spatial and magnitude characteristics of microseismicity are predominantly influenced by the latter. Based upon this improved understanding of fracture-slip seismic-generation mechanism, the current work aimed at establishing a data-driven yet physics-based probabilistic forecasting methodology for hazardous microseismicity using microseismic monitoring data with concurrent face advance records. Through performing statistical analyses and probability distribution fitting for temporal, magnitude and spatial characteristics of microseismicity within a time window, a short-term forecasting model is developed to estimate the probability of potentially hazardous microseismicity over the next time interval in the form of a joint probability. The real time forecasting of hazardous microseismicity during longwall coal mining is realised through regularly updating the statistical model using the most recent microseismic sequence datasets and face advance records. This forecasting methodology is featured by the physical basis which provides a good explicability of forecasting results, and the probabilistic perspective which accounts for the stochastic nature of mining-induced microseismicity. This model has been employed to make time-varying forecasts of hazardous microseismicity around two longwall panels over a one-year coal production period at Coal Mine Velenje, and satisfactory results at both panels were achieved. In addition, the analysis suggested that the energy magnitude distribution of microseismicity is a dominant factor in contributing to the potential of hazardous microseismicity. This statistical model using microseismic monitoring data has important implications in the evaluation of mining-induced hazards and optimal control of longwall face advance in burst-prone deep-level mining sites