Assessing Dynamic Reserves vs. Stochastic Optimization for Effective Integration of Operating Probabilistic Forecasts

Abstract

Probabilistic forecasting is becoming pivotal in utilities' decision-making processes, offering an accurate portrayal of plausible forecast deviations as opposed to deterministic forecasting which only focuses on the expected forecasted variables. Two methods, dynamic reserve and stochastic optimization, have been used to integrate probabilistic forecasts into power system operational planning. Dynamic reserve predicts system reserve requirements based on observed (from historical observations) or expected (from probabilistic forecasts) uncertainty spreads. This approach has a low computational burden, but it is commitment and dispatch agnostic. Stochastic optimization, on the other hand, considers multiple scenarios simultaneously (from probabilistic forecasts), allocating recourse across the commitment and dispatch variables, but demanding high computational resources and time. The selection between these methods depends on utility requirements and specific situations. This paper conducts a comprehensive evaluation of both methods using a calibrated real-size system representing the Southern Company for medium and high solar penetration levels. Additionally, it proposes a hybrid dynamic reserve and stochastic optimization approach with a risk evaluation pre-scheduling procedure to enhance decision-making.