Storage Decomposition Method
| Units: | - |
| Mode: | Input Only |
| Multi-band: | False |
| Default Value: | 1 |
| Validation Rule: | In (0,1,2) |
| Key Property: | No |
| Description: | Method used to pass the optimal storage trajectory from one simulation phase to the next. |
| Detail: |
Storage Decomposition Method controls the way the simulator decomposes the optimal LT Plan or MT Schedule storage trajectory (the sequence of End Volume values in time) down to ST Schedule. Decomposition is generally required only for storage that are large enough that they do not attain their bounds within a single step of ST Schedule. This switch takes the following values:
- None (value = 0)
- No trajectory or shadow prices information is passed to ST Schedule, thus ST Schedule is free to run the storage without reference to the mid-term solution.
- Targets (value = 1)
- The sequence of End Volume values out of LT Plan or MT Schedule is used to set target end storage volumes for each ST Schedule step. Thus ST Schedule will follow the same storage trajectory as the preceding simulation phase. To maintain feasibility, a penalty function is automatically created that allows the target to be violated at a pseudo-cost. The form of this penalty function are described below.
- Shadow Prices (value = 2)
- The Shadow Price from LT Plan or MT Schedule is used to set the value of step End Volume in ST Schedule. Thus ST Schedule will account for this future value of water in storage when optimizing the storage release decisions.
More information is available in the Hydro Reservoirs article.
Notes:
- For any method other than "None" and when using partial chronology, it is preferred, but not required, that the LT Plan LDC Type or MT Schedule LDC Type is at least the resolution of the Horizon Chrono Step Type e.g. if ST Schedule is running with day-sized steps (look-ahead included) then the preceding phase should use a day resolution. If the preceding phase runs a coarser resolution LDC such as week then the targets created could violate storage bounds. To compensate for this resolution mismatch penalties are introduced on the storage bounds according to the Decomposition Bound Penalty property. This does not apply to chronological LT Plan or MT Schedule.
- Decomposition Method = "Shadow Prices" has a special function when MT Schedule is run with Stochastic Method = "Stochastic". In that mode the Storage Shadow Price is combined with the objective function value in each sample to produce a multi-dimensional shadow prices surface called the Future Cost Function (FCF). The components of this function can be access with the Diagnostic Future Cost Function.
- When ST Schedule is run without MT Schedule you may enter a Future Cost Function (FCF) via the Global class.
- Decomposition Method = "Targets" creates a penalty function around the decomposed trajectory so that variations are allowed in ST Schedule without creating infeasibilities in the target constraints. The penalty function is a piece-wise linear function that penalizes under-reaching of the target but not over-reaching. The function has 10 bands and the price in each band is based on a multiple of the Shadow Price calculated in MT Schedule thus:
where:
i is the band numberThe size of the bands is controlled by the following power function:
(Target - Min Volume)) / 100.0) × a × e ^ (b × i)where:
- a is Storage Decomposition Penalty a
- b is Storage Decomposition Penalty b
- c is Storage Decomposition Penalty c
- x is Storage Decomposition Penalty x