A stochastic unit commitment with Derand technique for ISO's Reserve Adequacy Assessment

After the day-ahead market is cleared and before the real-time market is started, Independent System Operator (ISO) uses a Reserve Adequacy Assessment (RAA) process to determine supplemental unit commitments to meet the hourly forecasted load and reserve requirements for the next operating day. This paper presents a stochastic programming model for the RAA process to manage the net load uncertainty. Due to the large size of ISO systems and the increasing net-load variability caused by increasing penetration of renewable resources, the problem is computationally challenging. We develop an effective scenario reduction technique, Derandomization (or Derand), to identify a small number of scenarios that extract key and unbiased information from the distributions of uncertain variables. Numerical testing results show that the stochastic model with only 3 or 5 scenarios outperforms its deterministic counterpart by a significant margin, yielding lower expected cost and fewer constraint violations. Results also show that the Derand method outperforms several conventional scenario reduction methods, and the solution quality is comparable to the cost based scenario-reduction technique but with less computational efforts.