Merge pull request #892 from koen-vg/io-api-option

Add support for the linopy `io_api` option

config/config.default.yaml

@@ -687,6 +687,7 @@ solving:
     rolling_horizon: false
     seed: 123
     custom_extra_functionality: "../data/custom_extra_functionality.py"
+    # io_api: "direct"  # Increases performance but only supported for the highs and gurobi solvers
     # options that go into the optimize function
     track_iterations: false
     min_iterations: 4

doc/configtables/solving.csv

@@ -7,6 +7,7 @@ options,,,
 -- rolling_horizon,bool,"{'true','false'}","Whether to optimize the network in a rolling horizon manner, where the snapshot range is split into slices of size `horizon` which are solved consecutively."
 -- seed,--,int,Random seed for increased deterministic behaviour.
 -- custom_extra_functionality,--,str,Path to a Python file with custom extra functionality code to be injected into the solving rules of the workflow relative to ``rules`` directory.
+-- io_api,string,"{'lp','mps','direct'}",Passed to linopy and determines the API used to communicate with the solver. With the ``'lp'`` and ``'mps'`` options linopy passes a file to the solver; with the ``'direct'`` option (only supported for HIGHS and Gurobi) linopy uses an in-memory python API resulting in better performance.
 -- track_iterations,bool,"{'true','false'}",Flag whether to store the intermediate branch capacities and objective function values are recorded for each iteration in ``network.lines['s_nom_opt_X']`` (where ``X`` labels the iteration)
 -- min_iterations,--,int,Minimum number of solving iterations in between which resistance and reactence (``x/r``) are updated for branches according to ``s_nom_opt`` of the previous run.
 -- max_iterations,--,int,Maximum number of solving iterations in between which resistance and reactence (``x/r``) are updated for branches according to ``s_nom_opt`` of the previous run.

doc/release_notes.rst

@@ -33,6 +33,8 @@ Upcoming Release
 
 * Bugfix: Correctly read out number of solver threads from configuration file.
 
+* Add support for the linopy ``io_api`` option; set to ``"direct"`` to increase model reading and writing performance for the highs and gurobi solvers.
+
 
 PyPSA-Eur 0.9.0 (5th January 2024)
 ==================================

scripts/solve_network.py

@@ -890,6 +890,7 @@ def solve_network(n, config, solving, opts="", **kwargs):
         "linearized_unit_commitment", False
     )
     kwargs["assign_all_duals"] = cf_solving.get("assign_all_duals", False)
+    kwargs["io_api"] = cf_solving.get("io_api", None)
 
     if kwargs["solver_name"] == "gurobi":
         logging.getLogger("gurobipy").setLevel(logging.CRITICAL)