MultiStageStochasticBlock

test/MultiStageStochasticBlock

A tester which provides initial tests for MultiStageStochasticBlock, the SMS++ Block that aggregates one TwoStageStochasticBlock per outer-stage scenario and ties the first-stage variables across them with an outer set of non-anticipativity constraints, as well as for quite a lot of the mechanics of the "core" SMS++ library.

This executable, given the filename of a netCDF file containing the description of a MultiStageStochasticBlock, solves its deterministic equivalent (the extensive form of the whole multi-stage scenario tree) with a :MILPSolver, comparing the result against an optional reference objective value passed on the command line. The running time is printed. The relative tolerance for the comparison is fixed at 1e-5.

The usage of the executable is the following:

   ./MSSB_test MSSB-file [BSC-file ws ref]
   BSC-file: BlockSolverConfig description [BSPar-MS.txt]
   ws:       0 = LagrangianDualSolver, 1 = reserved [0]
   ref:      reference objective value to compare against [none]

The inner Block of each TwoStageStochasticBlock can be any SMS++ Block: the tester is problem-agnostic. The instances shipped under batches/ happen to embed a UCBlock (for energy-community-type applications), but nothing in the executable assumes a specific inner Block type.

For the instances produced by csv2nc4.jl --multistage, the multi-stage extensive form coincides with the flattened TwoStageStochasticBlock one, so the reference objective values under batches/batch-ec are exactly those of the corresponding TSSB_EC_* instances of tests/TwoStageStochasticBlock.

The instances of batches/batch-pypsa are the multi-stage counterpart of the PyPSA family: the same PyPSA-Eur network the two-stage ones are drawn from, with its single axis of uncertainty split in two, the climate year in the outer stage and the demand, drawn conditional on it, in the inner one. Each of them says in its name what the climate acts upon, the availability of the renewables, the hydro inflow or both, i.e., they are the counterparts of the maxpower, hydroinflow and complete instances of the two-stage family; the demand one has no counterpart here, the demand being the inner stage of all of them. Their reference objective values are the optimum of the equivalent flat network solved by PyPSA, which coincides with the tree one as long as the only here-and-now Variable are the design ones. Unlike the two-stage batch, this one has no LagrangianDualSolver to cross-check the :MILPSolver against: relaxing the outer non-anticipativity constraints leaves one TwoStageStochasticBlock per LagBFunction, and a LagBFunction requires its inner Block to carry an FRealObjective of its own, which a TwoStageStochasticBlock has not, its objective being the scaled sum of the objectives of its sub-Blocks.

A makefile is also provided that builds the executable including the MultiStageStochasticBlock, TwoStageStochasticBlock, LagrangianDualSolver, BundleSolver, MILPSolver modules and the core SMS++ library, together with the inner-Block module needed by the instances in batches/ (currently UCBlock).

Configuration files

• BSPar-MS.txt — outer BlockSolverConfig registering a :MILPSolver (default GRBMILPSolver) on the deterministic equivalent.
• MILPCfg.txt — ComputeConfig of the :MILPSolver (continuous relaxation, used to match the EnergyCommunity.jl linear reference).
• InnerBCfg.txt — "meta"-BlockConfig mapping each inner-Block classname (ThermalUnitBlock, DCNetworkBlock, ...) to its own BlockConfig (TUBCfg.txt, DCNBCfg.txt, ACBCfg.txt, PFBCfg.txt), applied recursively over the whole Block tree.

Authors

• Antonio Frangioni
  Dipartimento di Informatica
  Università di Pisa

• Donato Meoli
  Dipartimento di Informatica
  Università di Pisa

License

This code is provided free of charge under the GNU Lesser General Public License version 3.0 - see the LICENSE file for details.