An adaptive hybrid algorithm with system participants classification for efficient convex hull pricing in electricity markets

Abstract

Due to the non-convexities in electricity market, system operators may need to provide side payments to incentivize participants to follow the production plans. Convex hull prices, derived from the Lagrange dual of the unit commitment problem (typically modeled as a mixed-integer programming problem), can minimize these side payments. We present an adaptive hybrid algorithm designed to efficiently compute convex hull prices by approaching the convex primal formulation of this Lagrange dual problem asymptotically. The algorithm classifies system participants into four groups based on the complexity of their convex hull descriptions and applies tailored convex hull formulations or column/row generation techniques to each group. By seamlessly integrating advanced models and algorithms within a unified primal framework, our approach enhances both computational efficiency and accuracy. We evaluated the algorithm on 40 instances and compared its performance against other methods, including column generation, row generation, and the Level Method. Results demonstrate that our adaptive hybrid algorithm reduces computation time by at least 90 % compared to the traditional Level Method. These findings confirm the algorithm’s computational feasibility for large-scale market clearing problems.