Hedging Strategy and Electricity Contract Engineering - IFOR

116 Power portfolio optimization
6.2. Traditional power portfolio optimizations
Traditionally, optimization approaches in the electricity market have been
focused on the reliability of the whole power system and was developed for
central planning purposes, such as least cost planning [Sto89] and integrated
resource planning [SJR97, Gar00]. Prices were regarded to be deterministic
and the goal was typically cost-efficient and reliable supply of electricity
From an optimization point of view the power portfolio optimization started
with capacity planning, which was formulated as a least cost investment
problem. Solved with linear programming the total cost was minimized
subject to fuel availability, demand and capacity to mention the most important
constraints. The objective function typically summed up capital cost and
generation cost, like fuel cost, over the whole planning period. The constraints
as a rule included forecasted demand and plant availability. This deterministic
approach was pioneered by EdF in 1954 that developed a schematic linear programming
model with only 4 constraints and 5 variables. Described by Massé
and Gibrat [MG57] this was the first application of LP to electricity planning.
To account for the start-up times and shut-down costs of thermal units, like
coal plants, Schaeffer and Cherene [SC89] introduced integer variables and
solved their mixed-integer linear program by finding the optimal dispatch of an
existing array of generating plants. The generation cost was minimized subject
to meeting short-term demand. These so-called unit commitment problems
have been extended to allow for short-term transactions, i. e. entering of contracts.
Takriti et al. [TKW98] solve this problem using Lagrangian relaxation
and Bender’s decomposition. The goal is still to minimize the generation
costs while meeting the electric load. An optimal production schedule for a
longer period, typically a year, for hydro systems with reservoir capacities
with the possibility to enter contracts has also been studied. For example, Bart
et [BBCœ al. 98] examines a utility with both nuclear and hydro plants having
access to a spot market. The objective is to minimize the generation costs and
the problem is solved with linear programming. For an extensive survey of
optimization approaches in regulated power markets, see [Ku95].
Following the deregulation and the corresponding uncertainty on future
earnings, producers will however have to change their focus from reliable and