Emissions-Aware Energy Storage Decision Based on Deep Reinforcement Learning

This paper presents a deep reinforcement learning model to optimize energy arbitrage in energy storage systems while considering real-time electricity prices and carbon emissions. The DRL model uses a deep Q-network algorithm to learn optimal charging/discharging decisions based on historical price and carbon data, aiming to maximize profit and minimize emissions. The model's reward function includes both financial rewards and carbon costs, providing a comprehensive assessment of the impacts of carbon costs on system profitability. The study compares two sets of rewards: one without carbon cost and one with carbon emissions factored in. Results show that carbon costs significantly influence profitability and decision-making strategies. The DRL model adapts by charging less and discharging more during high demand to minimize carbon costs and maximize rewards. The research concludes that with an optimized strategy, it is possible to achieve carbon neutrality while maximizing profit in ESS operations.