An efficient deep reinforcement machine learning-based control reverse osmosis system for water desalination

Water scarcity is a permanent problem that faces all over the world. Artificial Intelligence (AI) has many machine learning methods used to solve many problems in all fields. This paper suggests a novel and efficient approach to finding a trans-membrane pressure using Deep Reinforcement Learning (DRL). Our system uses Deep Deterministic Policy Gradient (DDPG) agent to adjust the pressure across the membrane. This adjustment considers the Salt Rejection (SR) to be 99% to investigate the desired water flux. The system takes the maximum height of the water in the tank (hmax), the salt concentration of feed flow (C), the temperature of feed flow (T), the recovery ratio (R), and the salt rejection ratio (SR) as input, and returns the water flux Qp. The results show the effectiveness and the power of the DDPG agent in finding that pressure. The agent is trained in a small number of episodes (150), and the average reward value is high.