Mining is the backbone of industrial development in many countries. During mining, groundwater levels generally cut across the mining and has led to flooding of the mine floors hindering the progress of mining and jeopardising the safety of the working force. Most of the mines adopt depressurisation schemes to pump out the excess water that causes mine flooding. The interaction between different layers in a multi-layered aquifer and the effect of discontinuity in the mineral deposit was studied to understand its impact on depressurisation schemes. The discontinuity in the lignite deposit lying above a confined aquifer was modeled using Visual MODFLOW and simulation results showed an increased inflow flux by 2.64 times to the confined aquifer because of the presence of discontinuities, which in turn raised the piezometric levels in the confined aquifer resulting in more pumping. Particle tracking analysis was also performed using MODPATH to identify the source of additional flux in the zone of discontinuity. A rise of 3.3 m in water level was observed in the region of discontinuity, which needs to be pumped out for safe mining. The presence of discontinuity resulted in 30% additional pumping near the discontinuous zone for carrying out safe mining operations. The process of depressurization involves extraction of huge amount of groundwater. A simulation - optimization model was developed on a real-time basis using Genetic Algorithm to reduce the excess groundwater pumping by maintaining the ground water levels at safe operating limits for mining. Real-time operations were performed with and without Data Assimilation to improve the predictions of the optimization model. The real-time operations were able to conserve 0.36 - 2.76 Mm3 of groundwater in a year compared to the historical operation of the mines. The real-time model with data assimilation was able to predict 13 - 18% lesser pumping compared to the model without data assimilation. Therefore, it is essential for mining operations that involve huge groundwater pumping to adopt optimization models based on real-time data update for a model more closer to reality which can reduce the total amount of groundwater extraction.