ML for Electricity Market Pricing

Locational Marginal Prices (LMPs) set the price of electricity at each node of a power grid, and computing them means solving an optimal power flow problem, which is slow at scale. In my master thesis at the Institute of Computing, USI Lugano, supported by a MaRS scholarship from the Hasler Foundation, I studied whether machine learning models can predict LMPs in near real time.

• Generated ground-truth data with MATPOWER and MOSEK on PGLib-OPF grids.
• Reached a maximum error of 5% across a range of n-1 security-criterion cases.

Outputs

• Paper: AI Driven Near Real-time Locational Marginal Pricing Method: A Feasibility and Robustness Study, IEEE ISGT Europe 2023.
• Thesis: Feasibility and Robustness of Machine Learning algorithms for calculating Locational Marginal Pricing, 2023.