A Data-Driven Cost Budget Satisficing Model for Unit Commitment Under Solar Power Uncertainty

Motivated by the risk of overruns, this paper proposes a two-stage cost budget satisficing (CBS) model for solving the unit commitment (UC) problem under solar power uncertainty. To interpret the level of conservativeness inherent in the budget, we define a cost overrun risk measure (CORM) that restricts the expected cost violation from the budget over the possible distribution of uncertain solar output limits. Based on CORM, operators can set their satisfactory budget and accordingly obtain the most robust first-stage UC decisions with out-of-sample cost bound coherent with the budget. For computational tractability, solar outcome data is leveraged via the L1-norm Wasserstein metric representing distribution deviation. The linear decision rule is employed to approximate the adaptation between second-stage economic dispatch (ED) decisions and solar prediction errors, which is lifted via the L1-norm representing accumulative deviation. In the case study, we employ a stochastic optimization model as a baseline to obtain a predictable budget, together with loss of optimality as a tradeoff to withstand greater uncertainty, providing an instance of implementation. With three real-life datasets, we simulate solar penetrations into IEEE 14-bus, 30-bus, 57-bus, and 118-bus systems. The results validate CBS manages the cost overrun well while achieving superior out-of-sample performance.