Skip to content

Debugging

This page covers what to do when EPM doesn't run, doesn't converge, or produces unrealistic results.

First steps

Start by identifying the symptom:

Symptom Where to look
Python error before GAMS starts Terminal output, check input files and config.csv
GAMS compilation error .lst file, first ERROR line
Model infeasible PA.gdx for binding constraints, .lst for infeasibility message
Model runs but results look wrong PA.gdx variable levels, check input data units
Run times out or crashes .log file for memory/time, reduce problem size or use remote server

Enable verbose logging to get more detail:

python epm.py --folder_input your_data --config your_data/config.csv --debug

Or add --trace to log detailed input reading:

python epm.py --folder_input your_data --config your_data/config.csv --trace

Common errors

Infeasible model Usually caused by contradictory constraints: a renewable target that cannot be met, demand exceeding all possible supply, etc. Check:

  • Demand vs. installed + buildable capacity
  • Emission caps relative to the fuel mix
  • Reserve requirements vs. dispatchable capacity

Enable iis = 1 in your CPLEX option file to get an automatic infeasibility report; see Solver Options.

Unrealistic results (e.g. enormous unserved energy) Usually a data issue. Check:

  • Unit consistency (MW vs. GW, MWh vs. GWh)
  • Missing or zero capacity values
  • Costs set to zero unintentionally (allows infinite dispatch)

Python crashes before GAMS Check config.csv: a missing file reference or wrong path is the most common cause. Run with --trace to see exactly which input file fails to load.

GAMS compilation error Open the .lst file and search for ****; GAMS marks all errors with four asterisks. The line number points directly to the problem.

GAMS output files

PA.gdx — post-solve values

Generated after a solve. Open in GAMS Studio or with gdxdump. Each variable shows:

Field Meaning
Level Actual solution value
Marginal Dual value (shadow price or reduced cost)
Lower / Upper Variable bounds

What to look for:

  • All levels = 0: the variable may not be connected to any constraint, or has a cost preventing selection
  • Marginal = EPS: the bound is not active (usually normal)
  • Very large levels: check unit conversions and parameter bounds

.log — execution log

Contains timestamps, file inclusions, model generation time, and solver startup. Useful for diagnosing crashes and timeouts.

.lst — execution listing

Automatically generated. Contains compilation output, equation listings, and solver messages. Search for **** to find errors quickly.

Tips:

  • Set option limcol = 0; option limrow = 0; to reduce listing size
  • Use the "Variable Listing" and "Equation Listing" sections to inspect values in context

.ref — symbol reference

Shows where every set, parameter, variable, and equation is declared, defined, and referenced. Useful for finding variables that are declared but never used in a constraint.

To generate it in GAMS Studio: add rf=filename.ref to the command-line arguments in the Task Bar.

Still stuck?