Curtailment simulation
Background
Curtailment has a significant impact on asset revenue, as assets that are curtailed cannot sell the electricity produced. It is therefore important to simulate the effect of curtailment. There are two main aspects to consider for simulating curtailment in the future: how much curtailment a plant will have, and when is this curtailment going to happen.
The first point is quite complex and depends on many different factors, such as amount of solar installed in a zone, power interconnection and distribution capacity, availability of storage, etc. To estimate the total amount, we use projections from the grid operators as baselines in our platform. These can be modified as part of the scenario assumptions.
For the second point, determining when the curtailment happen, we use a mix of heuristics and statistical model to select day and hours of curtailment for each plant, given a total for each year.
Calculation logic
Parametrization
We determine the (non-normalized) probability of a day having curtailment as follows:
where is the day-ahead price. When a day is curtailed, we determine the start and end hour with a simple statistical model, based on historical data. The probability of hour being a start hour at month is proportional to the number of curtailment events that started at hour during month in the past. Similarly, we calculate the . We use the full historical record of curtailment data up to the present for the parametrization. This is updated weekly to reflect new data.
Calculation
Assuming that total curtailment for the plant is estimated at :
for each year Y: # initialize curtailment for the plant this year while : let X be a random variable with probability for each extract a day from X extract the start and end hour from and