Learning to Arbitrage Congestion in Electricity Market with Virtual Bids

In this paper, we introduce a machine learning-based risk-constrained portfolio optimization framework to arbitrage congestion with virtual bids in wholesale electricity markets. The proposed trading strategy aims to maximize the profit from the perspective of a proprietary trading company. A deep neural network is designed to estimate the difference between congestion spreads in day-ahead (DA) and real-time (RT) markets. A clustering algorithm is adopted to separate pricing nodes into a few groups, between which the congestion spreads can be exploited. We validate the proposed algorithmic trading strategy using publicly available data from California Independent System Operator (CAISO). The empirical results indicate that our proposed algorithm attains significant profit with modest portfolio budget and low risk tolerance level.