Optimal Control of a Gold Cyanide Leaching Process Based on Reinforcement Learning

Leaching by cyanide solutions in aerated alkaline slurry is the main process for gold extraction from ores. In this study, the model-based reinforcement learning (MBRL) is used to design an optimal controller for a gold cyanidation leaching process (GCLP) by automatically learning a control policy. The simulation experiments are performed on a first principle model to verify the effectiveness of the proposed method. The results show that the MBRL method can learn an optimal control policy for the GCLP.