The purpose of this page is to serve as a permanent repository for the paper:
"Surrogate models for rural energy planning: Application to Bolivian lowlands isolated communities"
Inside this repository it is possible to find the scripts and data to reproduce the results of the paper. A brief explanation on how to do this is given below. It is important to note that the paper and this repository are meant to be read as one piece, in order to completely understand the theory and practical implementation of the work done. In order to have a complete description of the scripts, information and how to implement the methology, do not hesitate to contact any of the Authors with the emails provided or through github.
The first step to generate the surrogate models is the creation of a data base of optimal size microgrids. To do this, go to the main folder, open the file Micro-Grids_Surrogate in a development enviroment and run the script. A message regarding the status of each optimization should appear in the terminal. The results of each optimization are saved in the Results folder with a distinctive name. Once all optimizations are performed, the database can be created with the Data_Base_Creation script. This database is save in the Dabases folder. This database is also used for the creation of the surrogate moodels. Now, it is possible to analyze the results. This is done with the script call Results_Analysis. Once it is run, a message in the terminal should appear with a summary of the most important results. Also, Figure 10 and 11 of the paper are saved in the folder Plots with the names of BoxPlot_LCOE_NPC.png and LDC_Renewable_Penetration.png.
To do the crossvalidation test for each target variable (NPC, LCOE, PV installed capacity, Battery installed capacity) the files that begin with Crossvalidation must be run. The end of the file depends on the variable that we want to analyze. As the different analyses are performed, a message in the terminal will appear with the results. Finally, to create the surrogate models, the files that begin with Surroagate_Full_Data must be run. In this case also, the end of the file will depend on the variable that is needed. When the process finishes some information will appear in the terminal and the surrogate model file will be saved in the folder Surrogate_Models. To reproduce figure 12 of the paper, the file Plot_predicted_Computed must be run. Figure 13 can be reproduced by running the Plot_Predicted_Computed_Fix file. This script needs 4 additional databases (Database_Fix, Database_100, Database_300, Database_500). To create the first database, the file Micro-Grids_Surrogate_fix must be run. This file is in the folder Fix_Cost. Additionally, the Data_Base_Creation must be run to create the needed database. Finally, this data base must be place in the folder Databases. The other three data bases are created following similar steps in the Fix_Cost_Households (they are included already, for the sake of simplicity). The only difference is that the file to run the sizing process is call Micro-Grids_Surrogate_fix_Households.
In addition to the surrogate model creation and validation process, also in this repository is included the case studies with Onsset. these are located in the folder OnSSET_Scenario. The scenario onsset classic is saved inside the folder Onsset_Scenario/onssset_classic and can be run from the script Bolivia_runner. In the other hand, the scenario onsset surrogate models can be run from Onsset_Scenario/onssset_Surrogate with the help of Bolivia_runner file. The information needed to create Figure 14, can be extracted from each scenario if the file Plot_Data is run. The information is saved in two excel files called Plot_data_classic and Plot_data_surrogate. The coordinates are the X_deg (longitude) and Y_def (latitude) columns.
Sergio Balderrana University of Liege, Belgium - Universidad Mayor de San Simon, Bolivia, E-mail: [email protected] / [email protected]
Francesco Lombardi, Politecnico di milano, E-mail: [email protected]
Nicolo Stevanato, Politecnico di milano, E-mail: [email protected];
Gabriela Peña Royal Institute of Technology, E-mail: [email protected]
Emanuela Colombo, Politecnico di milano, E-mail: [email protected]
Sylvain Quoilin, University of Liege, Belgium, E-mail: [email protected]
This repository has been tested in Linux or windows and needs different programs and phyton libraries in order to work.
First of all Micro-Grids needs Python 3.7 install in the computer. The easiest way to obtain it, is download anaconda in order to have all the tools needed to run python scripts.
The most important python libraries needed to run this repository are the following:
- pyomo 5.7
- pandas 0.23.4
- pyDOE 0.3.8
- joblib 0.14.1
- scikit-learn 0.20.3
- numpy 1.18.1
- matplotlib 3.1.3
- requests 2.22.0
- pvlib 0.7.2
- scipy 1.4.1
- openpyxl 3.0.3
Any of the following solvents can be used during the optimization process:
- Gurobi
This is a free software licensed under the “European Union Public Licence" EUPL v1.1. It can be redistributed and/or modified under the terms of this license.