18
margins. These results allow us to compare these variable costs against the incremental capital
costs required to achieve the higher planning reserve margins.
41
The multi-area economic and reliability simulations in SERVM include an hourly chronological
economic dispatch that is subject to inter-regional transmission constraints. Each generation unit
is modeled individually, characterized by its own economic and physical characteristics.
Planned outages are scheduled in off-peak seasons to minimize the impact on reliability, while
unplanned outages and derates occur stochastically using historical distributions of time between
failures and time to repair. Load, hydro, wind, and solar conditions are modeled based on
profiles consistent with individual historical weather years. Dispatch limitations and limitations
on annual energy output are also imposed on certain types of resources such as demand response,
hydro generation, pumped storage, and environmentally-limited combustion turbines (CTs).
The model implements a weekly commitment algorithm in each study region that considers the
outlook for weather and planned generation outages. The model then conducts an hourly
economic dispatch of baseload, intermediate, and peaking resources, including an optimization
of transmission-constrained inter-regional power flows to minimize total cost. Pumped storage
resources are dispatched economically to capture differences in pricing between peak and off-
peak periods. During most hours, hourly prices in each region reflect marginal production costs,
with importing regions realizing higher prices when import constraints are binding. During
emergency and other peaking conditions, SERVM simulates scarcity prices that exceed
generators’ marginal production costs as explained further below.
To examine a full range of potential economic and reliability outcomes, SERVM conducts a
Monte Carlo analysis over a large number of scenarios varying with both demand and supply
conditions. Because reliability events occur only when system conditions that reflect unusually
high loads or limited supply, these simulations must capture wide distributions of possible
weather, load growth, and generation performance scenarios. In this study, we incorporate 32
weather years, 6 economic load forecast error points, and 50 draws of generating unit
performance for a total of 9,600 scenarios for each simulated reserve margin case, with each
scenario simulating all 8,760 hours of the year.
42
This large number of simulations is necessary
for accurately characterizing the reliability and economic implications of different planning
reserve margins because the vast majority of reliability-related costs are incurred within a small
41
Note that SERVM as a modeling tool does not endogenously estimate capital costs, which are reflected as
a fixed annual cost (in $/kW-year). Total capacity costs simply increase as a function of the planning
reserve margin being evaluated. The question of whether a particular reserve margin is actually
achievable or realistic under various market designs, including in regions with capacity markets, energy-
only markets, or cost-of-service regulation, depends on whether those markets have been constructed such
that investors are able to recover their fixed costs at that reserve margin. We examine these and other
market design implications of our modeling exercise in Section IV below.
42
Note that SERVM results incorporate this set of scenarios for each reserve margin level simulated for the
base case and each alternative case we present in this report. Depending on data availability and study
needs, SERVM can incorporate any number of weather, hydro, wind, solar, economic forecast, and other
scenarios. For the purpose of this study and to facilitate exploration of a larger number of alternative
cases, we streamlined the simulation effort by modeling only 50 draws of generation outage conditions. In
other studies, Astrape typically models several hundred generation outage draws.