Autonomous pH control by reinforcement learning for electroplating industry wastewater

Abstract

The electroplating industry, due to steps such as pickling, generates acid pH wastewater. Its treatment is important for environmental preservation and the future recovery of metals. Therefore, the main objective of this work was the development of an autonomous pH controller for electroplating industry liquid effluents, based on fully automated Reinforcement Learning (RL). In order to do that, a Continuous Stirred-Tank Reactor (CSTR) neutralization simulator, and an adapted Particle Swarm Optimization (PSO) algorithm to automate the choice of RL hyperparameters were developed. The controller was developed and validated when it stabilized the effluent's pH in a neutral range in different scenarios during the regulatory and servo operations better than a Proportional Integral Derivative (PID) controller. The development of autonomous wastewater pH control systems in coated surface treatment units is a significant advancement, as it reduces human intervention and allows the monitoring of variability associated with the electroplating industry.