Integrated Optimal Control for Electrolyte Temperature With Temporal Causal Network and Reinforcement Learning.

Area of Science:

• Metallurgy
• Chemical Engineering
• Control Systems

Background:

• Electrowinning is crucial for nonferrous metals, demanding significant energy.
• Maintaining optimal electrolyte temperature is vital for high current efficiency and reduced power consumption.
• Current control methods struggle with dynamic variable fluctuations and complex electrowinning mechanisms.

Purpose of the Study:

• To develop an advanced control strategy for optimizing electrolyte temperature in electrowinning processes.
• To address challenges in accurately estimating current efficiency and managing temperature fluctuations without process modeling.
• To improve energy efficiency in hydrometallurgical operations.

Main Methods:

• An integrated optimal control method combining temporal causal networks and reinforcement learning (RL) was proposed.
• Temporal causal networks were employed to accurately estimate current efficiency under various working conditions.
• RL controllers were developed for each condition, incorporating optimal temperature into reward functions for control strategy learning.

Main Results:

• The proposed method effectively estimates current efficiency and determines optimal electrolyte temperatures across different working conditions.
• Reinforcement learning controllers successfully learned control strategies to maintain electrolyte temperature within the optimal range.
• Experimental validation on a zinc electrowinning process confirmed the method's ability to stabilize temperature without process modeling.

Conclusions:

• The integrated approach offers a robust solution for optimal electrolyte temperature control in electrowinning.
• This method enhances current efficiency and reduces power consumption in hydrometallurgical processes.
• The technique provides a viable alternative to traditional modeling-based control strategies for complex industrial operations.