| Units: | - |
| Default Value: | -1 |
| Validation Rule: | Any Value |
| Description: | Number of stages rolled forward each iteration where -1 means this is automatically determined based on the scenario tree |
This option applies to models with a scenario tree solved using Stochastic Algorithm = "Rolling Horizon". Please refer to Multi-stage Stochastic Optimization for more information on solving models with scenarios trees. The Rolling Stage Increment sets how many stages (weeks or months) of each iteration of the Rolling Horizon contain branching. By default this is automatically determined based on the probabilities in the scenario tree (see Minimum Sample Weight), but you can override the default and force the Rolling Horizon method to branch over more or less stages in each iteration.
Example:
Consider the progress messages from this Rolling Horizon simulation with default settings:
MT Schedule
Started Multi-stage Stochastic Optimization
Started Roll 0 of 51. Branching Stages 1-5 of 209
Completed Roll 0 of 51
Started Roll 1 of 51. Branching Stages 5-9 of 209
Completed Roll 1 of 51
Each iteration considers branching over five stages ("Branching Stages 1-5"). If we set Rolling Stage Increment to 10, then the number of iterations will be halved and the length of each iteration in terms of branching will double.
Note that increasing the number of branching stages inevitably leads to numerical instability as the probabilities expressed in the optimization objective function become tiny.
See also: