Title:Deep Learning for Energy Market Contracts: Dynkin Game with Doubly RBSDEs

Abstract:This paper examines a Contract for Difference (CfD) with early exit options, a widely used risk management instrument in electricity markets. The contract involves a producer and a regulatory entity, both of whom can strategically exit based on evolving market conditions. The problem is formulated as a two-player Dynkin game, where electricity prices follow a mean-reverting stochastic process, and contract terminations incur penalties. The strategic interaction is modeled using Doubly Reflected Backward Stochastic Differential Equations (DRBSDEs), which characterize the fair value of the contract and optimal exit strategies. To solve the DRBSDEs, we develop a deep learning-based numerical method, leveraging neural networks for efficient approximation. The results provide insights into the role of exit penalties, price volatility, and market dynamics in optimizing CfD agreements, offering practical implications for policymakers and market participants.