The proliferation of distributed energy resources (DER) over the past decade has shifted the paradigm from centralized power generation to decentralized, localized generation. This shift aims to enhance the stability and sustainability of electrical grids. Demand response (DR) aggregators, which manage the provision of balancing services (BS) to system operators (SO) using clients’ DER, face the challenge of determining the optimal generation capacity to contract with the SO.
Traditionally, heuristic or deterministic methods have been used to set these capacities. However, these methods often fail to incorporate risk assessments and do not optimize capacity allocation, leaving potential profits unrealized. This study aims to address these shortcomings by developing a stochastic optimization model that considers uncertainties in site electricity demand and the frequency of short-term operating reserve (STOR) calls by the SO.
Methodology
The study introduces a novel two-step approach:
-
Demand Bin Characterization:
- The site electricity demand is characterized into different bins using historical demand data. This helps in understanding and accounting for the uncertainty in site demand.
-
Stochastic Model Using Mixed Integer Nonlinear Programming (MINLP):
- The model is developed using the General Algebraic Modeling System (GAMS) and incorporates five years of historical data on STOR calls. This stochastic approach is designed to handle the uncertainty in both the frequency and timing of STOR calls and the site electricity demand.
- The model’s objective is to maximize site profitability by determining the optimal generation capacity a DER can offer in a STOR contract. It considers the site’s electricity demand, the cost of fuel, availability and utilization prices, and penalties for under-delivery or non-delivery during STOR calls.