This project is an academic project carried out during the second semester of 2019-2020. I worked on this project alone and was overseen by a Phd student, Adrien Le Franc.
The goal of this project is to manage a battery in a microgrid fitted with a solar pannel.
This method uses the command predictive control method: at every step:
- we predict the net demand (difference between the demand and the production) in the microgrid for two weeks
- we set the trajectory battery load in order to minimize the total cost.
- only the first value of the prescription is retained
To have more details about this project, I invite you to read my report, in the .pdf file.
The main language used here is Python (with Numpy, Matplotlib, statsmodels, Pulp, Flask) but I also use HTML, JS and CSS for the dashboard
This method presents various file among which, what may be the most interesting are :
param.py, where is saved all the configuration of the fileforecast.py, containing the prediction methodsoptimisation.py, containing the optimisation problemcompute_result.pythat saves the interesting results for plot in .npy filesplot.pywhere we can draw the plot that appear in my reportapp.pythat launches the dashboardtools.pycontains tools. I have copied this file from an older project. Thus, some functions in this file are not used.
You'll have to execute either app.py, plot.py, compute_result.py
Once you've cloned the program, before your first execution, please set REFIT_MODEL to True in the file param.py. It will compute the fitted model for the prediction method and save it in a .pkl file in the save folder. Then, you can set REFIT_MODEL to False again.
The data comes from the website of the Australian compagny Ausgrid.com. Click here for the data