Is there still merit in the merit order stack? The impact of dynamic constraints on optimal plant mix

Dr Iain Staffell, Imperial College London

The UK faces ever more challenging questions in terms of its energy policy:

• What are the consequences of adding 30 GW of wind to a 60 GW system?

• What value should we place on increased interconnection with our neighbours?

• How will capital-intensive new build fare in a more volatile marketplace?

• Can a market that pays only for energy deliver sufficient peaking plant?

A growing literature examines the impact of wind generation on electricity markets.  The first papers considered the impact on prices, given the existing mix of generating capacity: intermittent generation acts as negative demand and raises the volatility of prices, while the so-called merit order effect showed that adding wind capacity would tend to reduce the average level of prices.  A second generation of studies considered the long-run impact of renewable generation on the optimal capacity mix and what this would do for prices.

Almost all of these studies, however, use the so-called merit order stack approach, in which power plants are ordered by their variable costs, and selected in the quantities that minimise the sum of variable and fixed costs.  This approach ignores the growing role that dynamic constraints will have on determining the optimal capacity mix.  Plant start-ups and shut-downs, limits on ramp-rates and energy storage are three issues that cannot be incorporated into a simple merit order stack, and require a dynamic dispatch model with high temporal resolution.

To test the impact of these constraints, we demonstrate a non-linear dispatch optimiser coupled with a long-run investment model.  This finds the mix of plants such that each type covers its costs, leaving no further opportunities for profitable entry to the market.  A representation of the 2020 GB electricity system is used to explore the capabilities of this model relative to a merit order stack.

We test the impact of individual constraints on marginal prices and carbon emissions, dispatch profiles and generator profits.  Deviations between the two methods of modelling are explored along with tipping points (such as maximum wind penetration and maximum ramp rates).  We identify which constraints have a significant impact on our results, and hence need to be included in future modelling work.

Furthermore, we propose a relatively simple extension to the merit order stack model, modifying the input cost parameters to account for the expected number of plant starts per year.  The impact of this extension on results is quantified, and limitations to its applicability are explored.  The paper will show whether this heuristic can be used to simplify future modelling efforts.

Models based on a merit-order stack show that changes to the capacity mix almost exactly offset the impact of wind generation on the pattern of prices – each type of generator would receive the same average income in a system that has been optimised for wind as in an optimised system with conventional generators alone.  This paper shows whether that result holds once dynamic constraints are taken into account.