Dynamic control of deep-level mine cooling systems using artificial intelligence

Abstract

Deep-level mines in South Africa face rising electricity costs, with cooling systems contributing up to 28% of a mine's total energy. Existing energy-saving approaches lack adaptability. The mining industry has rarely embraced artificial intelligence (AI)-based energy efficiency strategies. This study proposes a dynamic AI-based strategy to improve energy efficiency in deep-level gold mine cooling systems. Long-short-term memory recurrent neural network (LSTM-RNN) models are integrated into dynamic control strategies for cooling components. The strategy acquires real-time data from the mine's SCADA, forecasts temperatures, and adjusts cooling component operations, potentially saving ZAR 1.5 million annually. Advantages brought to light include reduced human error and no infrastructure changes. The results serve to motivate the struggling industry to incorporate more Industry 4.0 technologies. This study contributes to mining by providing energy savings and operational improvements, as well as to AI by indicating the effective use of AI techniques in complex industrial environments.