Solving Models

Every potpourri model (single-period and multi-period) exposes a solve() method that wraps Pyomo's SolverFactory and handles result mapping back to the pandapower network.

Basic usage

opf.solve(solver='ipopt', print_solver_output=False)

After a successful solve, net.res_bus, net.res_line, etc. are populated automatically (unless to_net=False).

Parameters

Parameter	Type	Default	Description
`solver`	`str`	`'ipopt'`	Solver name: `'ipopt'`, `'glpk'`, `'cbc'`, `'gurobi'`, `'mindtpy'`, or `'neos'`
`print_solver_output`	`bool`	`False`	Stream solver log to stdout
`to_net`	`bool`	`True`	Write solution back to `net.res_*` DataFrames
`load_solutions`	`bool`	`True`	Load variable values from solver into Pyomo model
`time_limit`	`int`	`600`	Wall-clock time limit in seconds
`max_iter`	`int`	`None`	Maximum solver iterations (passed as `max_iter` option)
`mip_solver`	`str`	`'gurobi'`	MIP sub-solver for `mindtpy`
`init_strategy`	`str`	`'rNLP'`	Initialisation strategy for `mindtpy`
`neos_opt`	`str`	`'ipopt'`	Solver requested from the NEOS server

Local solvers

Install one of the following before calling solve():

Solver	Problem type	Install
IPOPT	NLP (AC OPF)	`conda install -c conda-forge ipopt`
GLPK	LP / MIP (DC OPF)	`conda install -c conda-forge glpk`
CBC	LP / MIP	`conda install -c conda-forge coincbc`
Gurobi	LP / MIP / NLP	`pip install gurobipy` (licence required)

Continuous NLP (AC OPF)

opf.solve(solver='ipopt', time_limit=300)

Linear / MIP (DC OPF, hosting capacity binary relaxation)

dcopf.solve(solver='glpk')

Mixed-integer nonlinear (HC_ACOPF with binary variables)

Use the mindtpy decomposition algorithm, which alternates between an NLP and a MIP sub-solver:

hc.solve(
    solver='mindtpy',
    mip_solver='gurobi',   # or 'glpk', 'cbc'
    max_iter=50,
    init_strategy='rNLP',
)

NEOS — remote solver

NEOS is a free public optimisation server. It accepts Pyomo models over the network and returns results without requiring a local solver installation.

Requirements

NEOS requires a registered e-mail address, passed via the NEOS_EMAIL environment variable:

export NEOS_EMAIL="your@email.address"

Or set it in Python before calling solve():

import os
os.environ["NEOS_EMAIL"] = "your@email.address"

Usage

opf.solve(solver='neos', neos_opt='ipopt')

neos_opt selects the solver on the NEOS side. Common choices are 'ipopt' (NLP), 'knitro' (NLP), 'cplex' (MIP), and 'couenne' (MINLP).

!!! note NEOS solves run on shared public infrastructure. For large multi-period problems or time-sensitive work, a local solver installation is strongly preferred.

Checking the result

The raw Pyomo result object is stored on the model instance:

import pyomo.environ as pe

opf.solve(solver='ipopt')

print(opf.results.solver.termination_condition)   # TerminationCondition.optimal
print(pe.value(opf.model.obj))                     # objective value

If the solve is infeasible or the solver times out, to_net mapping is skipped and a warning is printed. Increase time_limit or relax constraints if this happens.