Volume 45: Energy Transitions toward Carbon Neutrality: Part VIII

Optimizing Participation in The Local Energy Market Using a Deep Reinforcement Learning Approach Helder Pereira, Luis Gomes and Zita Vale

Abstract

The actors involved in the energy network have benefited much from its expansion in recent decades, but the network’s management has become more difficult as a result of the sources’ variability and unpredictability. Thus, it is essential to create models that can manage the current energy resources, which are becoming more and more dispersed. This study provides a new optimization model for participating in local energy markets based on peer-to-peer energy trading, using the twin-delayed deep deterministic policy gradient method and the double-auction trading mechanism. The model is integrated into an ecosystem based on agents, which enables the modeling of energy communities to produce a more plausible implementation scenario. The concept was used in a case study with 30 players in an energy community, and the findings revealed that each member saved an average of 1.54 EUR per week.

Keywords deep reinforcement learning, local energy markets, multi-agent systems, peer-to-peer energy trading, twin-delayed deep deterministic policy gradient