Prediction-Free, Real-Time Flexible Control of Tidal Lagoons through Proximal Policy Optimisation: A Case Study for the Swansea Lagoon

TL;DR

This paper introduces a Deep Reinforcement Learning method using Proximal Policy Optimization for real-time, prediction-free control of tidal lagoons, demonstrating comparable performance to traditional optimization methods in a Swansea Lagoon case study.

Contribution

The paper presents the first application of DRL with PPO for real-time tidal lagoon control, eliminating the need for tide predictions and repeated optimizations.

Findings

Achieved energy maximization comparable to existing methods.

Enabled real-time control using only measured ocean data.

Reduced operational costs by avoiding repeated optimization routines.

Abstract

Tidal Range Structures (TRS) have been considered for large-scale electricity generation for their potential ability to produce reasonably predictable energy without the emission of greenhouse gases. Once the main forcing components for driving the tides have deterministic dynamics, the available energy in a given TRS has been estimated, through analytical and numerical optimisation routines, as a mostly predictable event. This constraint imposes state-of-art flexible operation methods to rely on tidal predictions to infer best operational strategies for TRS, with the additional cost of requiring to run optimisation routines for every new tide. In this paper, a Deep Reinforcement Learning approach (Proximal Policy Optimisation through Unity ML-Agents) is introduced to perform automatic operation of TRS. For validation, the performance of the proposed method is compared with six different operation optimisation approaches devised from the literature, utilising the Swansea Bay Tidal Lagoon as a case study. We show that our approach is successful in maximising energy generation through an optimised operational policy of turbines and sluices, yielding competitive results with state-of-art optimisation strategies, with the clear advantages of requiring training once and performing real-time automatic control of TRS with measured ocean data only.