| Units: | - |
| Default Value: | 0 |
| Validation Rule: | In (0,1) |
| Description: | Algorithm invoked by the Stochastic Method when a scenario tree is present |
This setting determines which solution method is employed to solve a
stochastic optimization in the presence of a scenario tree (as defined
by a Global object). The available
methods are:
- Rolling Horizon (value = 0, default): Solves using the Rolling
Horizon solution method as described in the article Multi-stage
Stochastic Optimization. This method is the most robust. It
can handle any level of complexity including integer decision
variables, non-linear constraints etc and guarantees a fully
converged solution. In many cases this method is also the fastest to
execute. However, if you require output of the Future
Cost Function diagnostic this method will only produce the
optimal cuts. Choose the SDDP method if you need the full set of
Benders cuts.
- Stochastic Dual Dynamic Programming (SDDP) (value = 1): Solves
using the 'traditional' SDDP algorithm. This method performs a
series of forward and backward passes across the stages of the
horizon. Benders cuts are developed at each stage representing the
future cost function and value of storage. Iterations occur until
convergence criteria are met. This method naturally results in a
detailed Future Cost
Function which can be used as input to another simulation e.g.
to a short-term ST Schedule with
the FCF defining the future value of storage.