| Units: | - |
| Default Value: | 0 |
| Validation Rule: | In (0,1,2,3) |
| Description: | Stochastic optimization method for MT Schedule |
MT Schedule Stochastic Method sets the algorithm used in MT Schedule to resolve multi-sample data (defined by Variable objects), where the number of samples is set by the Stochastic Stochastic Risk Sample Count. The attribute can take the following values:
- Deterministic (value = 0)
- The expected value is used for sample data. For variables using endogenous sampling this means that the Profile value is used, and for variables that read their sample values from multi-band input, the first band is used (the assumption is that the first band is the expected value).
- Sequential Monte Carlo (value = 1)
- MT Schedule runs S times, one time for each sample, choosing the appropriate values for each. This yields S complete production simulation solutions. These independent samples are executed in sequence.
- Parallel Monte Carlo (value = 3)
- As above but the independent samples are executed in parallel i.e. all samples are executed at the same time on separate threads.
- Stochastic (value = 2)
- MT Schedule incorporates all S samples into a stochastic optimization. This optimization can be either two-stage or multi-stage depending on whether or not a scenario tree is defined via the Global settings. Stochastic optimization applies to many classes of object for example Storage objects identified by the Trajectory Non-anticipativity property, but also Generator, Gas Storage, etc. When a scenario tree is defined, the solution method employed is controlled by the Stochastic Algorithm setting.
When MT Schedule runs in either of the multi-sample modes the decomposition of constraints and other elements performed by the simulation phase yields S solutions and so the corresponding samples in ST Schedule read the decomposition for the appropriate sample. This improves the co-ordination between the simulation phases.
See also: