pH stablizing control of a neutral leaching process based on proximal policy optimization

Neutral leaching is the key process of zinc hydrometallurgy. The reasonable flow of waste acid added to make the terminal pH stable in a reasonable range is very important to improve the leaching rate and the removal rate of impurity ions. However, due to the fluctuation of inlet flow rate, feeding amount and composition change, there is great uncertainty in the process. And the stabling control of pH value is very challenging. In this paper, a pH optimization control algorithm based on proximal policy optimization (PPO) for leaching process is proposed. First, a nonlinear autoregressive (NARX) model is established based on process data for interacting with the agent. Then, the state space, action space and reward function are defined for the leaching process, and the strategies of action boundary setting and state normalization are adopted to ensure its learning performance. Finally, the experiments show that PPO controller can realize stabling control of pH value compared with manual experience. The optimal control framework based on deep reinforcement learning brings stability and intelligence to the zinc leaching process.