BWDF

This repository contains the code for the Battle of Water Demand Forecast submission

Usage

Setup a Python virtual environment

First, initiate a Python virtual environment The code was developed with Python 3.10.11 The code was tested with Python 3.8.10 Other Python version above 3.6 should be fine but were not tested

Open a terminal window in the project directory Create new venv by running the command:
python -m venv <venv name>
Activate the venv by:
source venv/bin/activate (for mac) or venv\Scripts\activate.bat (for windows)

Install dependencies

Once the virtual environment is setup and activated, Install the dependencies by running the following command: pip install -r requirements_no_versions.txt
Due to the large number of dependencies, the installation might take 10-15 minutes

Run the code

To run the code and generate a forecast for all the DMAs main.py should be run
Example usage to run the first test period
python main.py --inflow_data_file Inflow_Data_1.xlsx --weather_data_file Weather_Data_1.xlsx --predict test --models_config models_config_w1.json --test_name w1 --plot true python main.py --inflow_data_file Inflow_Data_2.xlsx --weather_data_file Weather_Data_2.xlsx --predict test --models_config models_config_w2.json --test_name w2 --plot true python main.py --inflow_data_file Inflow_Data_3.xlsx --weather_data_file Weather_Data_3.xlsx --predict test --models_config models_config_w3.json --test_name w3 --plot true python main.py --inflow_data_file Inflow_Data_4.xlsx --weather_data_file Weather_Data_4.xlsx --predict test --models_config models_config_w4.json --test_name w4 --plot true

Alternative way to run predictions:
The repository also contains .bat files for quick run. These files assume the virtual environment was created, and it is named venv To run the first test (w1) double-click the file predict_w1.bat To run the second test (w2) double-click the file predict_w2.bat To run the third test (w3) double-click the file predict_w3.bat To run the third test (w4) double-click the file predict_w4.bat

The run time is expected to be 5-10 minutes During the run the code will print to console according to its progress:
Predicting DMA A
Predicting DMA B
...

Arguments

inflow_data_file - Historic inflow data file, name only, not file path. file is located in resources directory

weather_data_file - Weather data file, name only, not file path. file is located in resources directory
predict - Must be one of test or experiment
models_config - Path to a json file that defines the models configuration
test_name - Pass only if predict argument is test. The test period to predict, must be one of w1, w2, w3, w4
experiment_idx - Pass only if predict argument is experiment. The experiment index
plot - "true" for plot the forecast, false otherwise. default is true
export - "true" for export forecast to csv, false otherwise. default is true

Output

The code generates two outputs:

1. A CSV file with the forecast - should be manually copy and paste to the BWDF template.
   The file is written to the local folder as forecast-test-<test_name>.csv or forecast-experiment-<experiment_idx>.csv
2. Forecast plot - Written to the local folder as forecast-test-<test_name>.png or forecast-experiment-<experiment_idx>.png