IEEE Power & Energy Magazine - November/December 2017 - 89

complex for systems that are realistic in terms of size. To solve
this problem, different simplifications have been proposed.
EDF's R&D group has developed in-house stochastic optimization tools to support investment decisions for large interconnected systems-including both capacity adequacy and
flexibility sufficiency and considering both the variability and
uncertainty of generation and demand-for time scales ranging
from years to an hour. The approach used is hierarchical, based
on employing fewer operational flexibility constraints at the planning stage, complemented by a verification of the performance
of the proposed generation mix when additional flexibility constraints and operational uncertainty are added. This provides a
good compromise between tractability and solution robustness.
To address both capacity adequacy and the flexibility sufficiency of a large interconnected system, input data requirements include
✔ technical and economic data regarding the generation
units and key transmission corridors for each country
as well as cross-border interconnection capacities
✔ demand and VG data using time-synchronized demand,
wind and PV generation, and associated forecast errors
with hourly (or shorter) resolution for each BA based on
weather observations over the last 50 years.
This approach is illustrated in Figure 3.
The approach builds on a multi-area PCS model, called
the continental model (CM), that includes hydro generation
optimization, unit commitment (UC), and dispatch. The
model is adapted to perform coordinated, multi-area generation planning using what is called the continental investment
loop (CIL). Once the investment problem has converged, the
generation schedules are modified to be used as inputs in
the flexibility assessment model.
The CM is a market-equilibrium model based on the fundamentals of load-generation balancing. The model simulates
the hydro-thermal dispatch of a multi-area system for every
hour of the year, given the interconnection constraints between
the areas. It also uses dynamic programming to define a set of
strategies that optimize use of hydro reservoirs to minimize the
global production cost. The results are the water values for
each time step and each scenario. These are used within
the CM as inputs to the UC and dispatch problem that minimize thermal and hydro generation costs using mixed-integer linear programming (MILP). The constraints of the MILP
problem include primary, secondary, and tertiary reserve and
generation-dynamic ratings such as minimum stable generation,
minimum up and down times, and ramp rates.
The problem is solved for many annual scenarios of demand,
wind and PV generation, water inflows, fuel costs, and thermal
unit availabilities. Each scenario represents a full year with
hourly (or shorter) resolution. To render the problem tractable
(the system has seven interconnected BAs and more than
800 thermal units), individual units are clustered into groups
of similar units. The CM outputs include hourly commitment
status, generation output, and scheduled reserve participation
for all groups of units, as well as marginal prices for all zones.
november/december 2017

The CIL computes the thermal generation mix, using an
iterative process. The objective is to obtain the thermal generation mix which ensures that, for every new unit, the expected
market revenue equals its fixed plus variable costs on an annualized basis. The renewable generation mix (hydro, wind, and
PV) is an input to the model.
The overall CIL-CM optimization includes two iterative steps:
1) a heuristic to propose a candidate solution generation mix
solution to satisfy an adequacy criterion
2) validation of the heuristic solution solving the hourly
load-generation dispatch using the CM to create a price
signal to verify the unit's profitability criteria.
Finally, the flexibility sufficiency of the generation planning solutions is analyzed using a probabilistic approach,
called FlexAssessement, with three main features:
✔ Simulation of unplanned generation outages and forecast errors are handled operationally, to quantify flexibility needs.
✔ For each hourly dispatch solution, the flexibility available is quantified and assessed relative to needs.
✔ The results are a set of flexibility indicators characterizing the flexibility sufficiency of the generation mix.

VG Data
Installed Capacity and
Location Load Factors
(Hourly or Lower),
Forecast Errors

Other Data
Hourly (or Lower) Demand and
Forecast Error, Generation
Investment and Fuel Costs,
Storage, DSM, Dynamic
Constraints, FOR, Ps,
CO2 Price

Iterative Process
Continental Investment Loop
Generation Mix

System
Marginal Costs

Continental Model
Generation Dispatch,
Network Flows
FlexAssessment

System Costs, Generation, and Grid Infrastructure
and Utilization
Generation Mix, Dispatch and Plant Load Factors,
Networks, Storage and DSM Use, RES Curtailment,
System Costs, Plant Revenues, Operation Margins
and Reserves, Flexibility Indicators

figure 3. The structure of the generation investment approach. DSM: demand-side management; FOR: forced
outage rate; Ps: probability of a unit's failure to start.
ieee power & energy magazine

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2017