Automatic risk adjustment for profit maximization in renewable dominated short-term electricity markets

State-of-the-art trading strategies in short-term electricity markets use risk awareness for reducing, inter alia, their exposure to the volatility of electricity prices. To ensure an optimal balance between risk and profit, risk-aversion parameters are traditionally fine-tuned via an offline out-of-sample analysis. Such a computationally-intensive analysis is typically run once, which yields time-invariant risk policies. Instead, this paper proposes the use of machine learning to select, in an online fashion, optimal risk-aversion parameters. This novel automatic risk-tuning approach offers the benefit of continuously adjusting the risk policy based on the dynamically changing market operating conditions. Our approach is tested on two risk-aversion parameters, that is, the confidence level of the conditional value-at-risk and the budget of uncertainty, respectively considering scenario-based and robust optimization frameworks. A set of performed case studies-focusing on the very short-term dispatch of a market actor participating in electricity markets-using real-world market data from the Belgian power system demonstrate the ability of the proposed methodology to outperform traditional offline risk policies.