Optimizing Non-Cooperative Bidding Strategies in High-Renewable Electricity Markets Using Multi-Agent Deep Deterministic Policy Gradient

The rapid rise of renewable energy penetration in electricity markets is fundamentally altering market dynamics and the clearing process. Traditional bidding strategies struggle to accommodate the variability and uncertainty introduced by high shares of wind and solar generation. This paper proposes a non-cooperative bidding strategy optimization framework based on Multi-Agent Deep Deterministic Policy Gradient (MADDPG) that explicitly addresses the challenges of market clearing under a high proportion of renewable energy. By modeling multiple heterogeneous market participants—including conventional thermal units, renewable energy units, and energy storage devices—in a non-cooperative game setting, we employ a MADDPG-based deep reinforcement learning approach to simulate and optimize their bidding decision-making processes. Renewable output uncertainty is incorporated into the environment model, allowing agents to learn robust strategies that anticipate and respond to fluctuations in wind and solar generation. Experimental simulations under a high-renewable penetration scenario demonstrate that the proposed MADDPG approach not only enhances the profitability of individual market participants but also improves overall market stability and clearing efficiency.