Demand Biasing and Virtual Bidding Can Lead to Suboptimal Unit Commitment Under Uncertainty

Power system operators currently use deterministic optimization for unit commitment (UC). However, increased uncertainty caused by, e.g., intermittent renewable energy can reduce the social welfare from these UC solutions. To improve deterministic solutions given uncertainty, operators can bias forecasts of uncertain quantities, e.g., wind power production or load. Another option is to use virtual bidders, who are financial participants that arbitrage prices between forward and real-time markets. Alternatively, operators can directly address uncertainty with stochastic optimization, which will provide better solutions if the uncertainty distributions are known. However, stochastic UC is generally not used. The objective of this paper is to compare the social welfare obtained by stochastic UC, deterministic UC with demand biasing, and markets with virtual bidders. We provide examples where deterministic UC cannot recover the stochastic commitment independent of the bias value. We also show an example where virtual bidders neither recover the stochastic optimum nor the deterministic optimum with biasing. Finally, we provide computational examples where we fail to recover the stochastic solution on larger systems. From these results, we conclude that current deterministic methods likely suffer from a performance gap compared to stochastic methods, even under demand biasing and virtual bidding.