NSGA-II as Python Library for Calibration of Models

pynsga2 1.0

Released: 11-January-2017

Introduction

• Non-Dominated Sorting Genetic Algorithm II (NSGA-II) is a multi-objective optimization algorithm used as an automatic calibration tool in wide range of disciplines.
• pynsga2 is adapted from nsga2 for SWAT models. Thus, pynsga2 is tested on a Hydrologic Model.
• pynsga2 can be used with any model in any dicipline.

Setup

Install the Python module:

• Python setuptools are required for installation
• Open a command prompt and cd to "./pynsga2", then type Python setup.py install
• If the bullet above does not work, try pip install <./pynsga2>

Setup the Model:

• Use "ExampleModel" folder structure
• Setup Model directory in "pynsga2Example.py" file in the "ExampleModel" folder
• Setup NSGA-II parameters in "pynsga2.def" file in "NSGA2.IN" folder
• Setup calibration parameters for your model in "pynsga2_par.def" file in "NSGA2.IN" folder
• Edit "pynsga2userutilities.py" under C:\YourPythonDirectory\Lib\site-packages\pynsga2lib file to determine how to calculate objective functions

Run the Model:

• Once setup done, run "pynsga2Example.py" to start the calibration

pynsga2 Outputs

In this section, the output files will be explained to make users familiar with them. The output files are not in the example model but they will be created once "pynsga2Example.py" runs sucessfully.

• Output.out:

• This file contains all results starting from first generation to last generation (Last generation is the final result that included in next output file, "plot.out").
• This ouput is in same format with Prof. Deb's C code output.
• The first columns represent the calibration parameter values sperated with single space. The calibration paramers are in same order defined in "pynsga2_par.def" file. The next columns represent fitness values as much as defined in "pynsga2.def" file. Next two columns are related with NSGA-II methods. "|**|" sign seperates previous population results from current population results. After "|**|" sign, the order of columns are same as before the sign.

• Plot.out:

• This file contains fitness results from the population of last generation (pareto front).
• Each line, displaying objective function values, is a member of the pareto front. The order defined in "pynsga2userutilities.py" file is applied.

• g_rank_record.out:

• This file contains only results related with NSGA-II records and does not have any data related with model outputs.

Final Notes

• Read papers bellow to understand the process behing the scripts
• Visit my website for more information
• If you encounter any problems or have suggestions for the future development, please contact Mehmet B. Ercan at [email protected].

Important:

• "pynsga2userutilities.py" is the most important file to addapt pynsga2 into your model. It will be edited to calculate objective functions for the model.

Credit:

Please cite one of the bellow articles if you use this code:

Ercan, M. B. and J. L. Goodall(2016), Design and implementation of a general software library for using NSGA-II with SWAT for multi-objective model calibration., Environmental Modelling & Software, 84, 112-120. doi:10.1016/j.envsoft.2016.06.017.

Ercan, M. B. and J. L. Goodall (2014), A Python tool for multi-gage calibration of SWAT models using the NSGA-II algorithm., In: Ames, D.P., Quinn, N.W.T., Rizzoli, A.E. (Eds.), 2014. Proceedings of the 7th International Congress on Environmental Modelling and Software, June 15-19, San Diego, California, USA. (4):2325-2331, 2014. doi:10.13140/2.1.3865.4407.