Installation

Download Conda.

setup conda path in .vscode/settings.json

change the following lines if conda isnt recognised "python.defaultInterpreterPath": "C:\user\anaconda3\python.exe", // here "python.condaPath": "C:\user\anaconda3\Scripts\conda.exe" // here

Run conda env create -f SOML.yaml at root folder.

File structure

directories

/data: has source and cleaned data /alpg-master: code used to generate synthetic load data /diagrams: some of the graphs and diagrams used in reports /graphs: some of the graphs generated during exploration process /saved_model: saved trained ml models

python classes

• ml_models.py : loads pre-trained ml models and performs online training
• online_batches.py : loads pre-made sliding window batches (for ease of use in ml_model.py)

jupyter notebooks

main jupyter notebook to run the project

• optimization.ipynb : Main notebook with optimization, MPC simulation, and printing of the graphs

Optimisation

• optimization(bidirectional_pricing).ipynb
• optimization(unidirectional_pricing).ipynb
• optimization_with_penalty.ipynb : Copy of optimization.ipynb used to compare the effect of the penalty functions.
• optimization_null.ipynb : Copy of optimization but adjusted for Case 0 (no prediction).

ML training and exploration

• [Austin]solar_temp_predictions.ipynb : Data cleaning and DNN models experiments for solar insolation and temp predictions
• [BOM]solar_predictions.ipynb : Data cleaning and DNN models experiments for BOM-72 solar insolation predictions.
• [BOM]Outdoor_temperature_predictions.ipynb : Data cleaning and classical ML models experiments for BOM dataset.
• complete_temp.ipynb : Data cleaning and DNN models for BOM-72 temperature predictions.
• load_dnn.ipynb : DNN models experiments for load predictions
• load_prediction.ipynb : Classical ML models for load predictions
• mol_model_testin.ipynb : To test teh ml_model.py and online_batches.py classes
• load_data.ipynb