Machine Learning Approach to Identify Important Parameters Influencing Pumping Load Shift in a Complex Dewatering System of a Deep-Level Mine

Abstract

This study investigated the application of machine learning to optimise the pumping load shift of a complex dewatering system in a deep-level mine, aiming to reduce energy costs associated with the dewatering process, which consumes an average of 14% of the mine’s electricity. Traditional practices, reliant on human control and simulations, often lead to inconsistent savings and occasional losses. The study employed multivariate linear regression (MLR) and extreme gradient boosting (XGBoost) on a mine dewatering system, to identify important parameters influencing the pumping load shift performance. Critical parameters significantly impacting the energy consumption of the dewatering system were identified by the best-performing model, XGBoost. Implementing a pumping schedule based on XGBoost insights resulted in consistent load shifting and enhanced energy cost savings. These findings highlight the potential of machine learning in comprehending and optimising complex systems in deep-level mines, with the case study approach proving effective in quantifying and validating real-world impacts. This approach could offer substantial energy savings through data-driven decision-making.