Integration of design and control for renewable energy systems with an application to anaerobic digestion: A deep deterministic policy gradient framework

Abstract

In recent years, the urgent need to develop sustainable processes to curb the effects of climate change has gained global attention and led to the transition into green technologies, such as Anaerobic Digestion Systems (AD). As these technologies present a complex dynamic behavior, there is a motivation to seek for new ways to optimize these systems. This study presents a Deep Deterministic Policy Gradient (DDPG) strategy for integration of process design and control. DDPG is a state-of-the-art reinforcement learning algorithm used to search for optimal solutions. The proposed approach considers stochastic disturbances and parametric uncertainty. Also, a penalty function included in the reward function is considered to account for process constraints. The proposed approach was tested in AD systems involving Tequila vinasses. Two reactor AD configurations were explored under multiple scenarios using the proposed DDPG strategy. While the two-stage AD system required a larger capital investment in exchange of higher amounts of biogas being produced, the single-stage AD system required less investment in capital costs in exchange of producing less biogas and therefore lower revenues than the two-stage system. The results showed that DDPG was able to identify optimal design and control profiles thus making it an attractive method for optimal process design and operations management of renewable systems.