This repo provides a tutorial of wind turbine performance prediction based on weather parameters.
Data are based on a subset of the ENGIE open dataset.
- Python ^3.8
- Poetry
# Navigate to your local folder
cd /your/local/folder
# Clone the WindML repository
git clone https://github.com/marcodigennaro/windml
# Enter the folder
cd windml/
# Install the package
poetry install
# Activate the environment
source .venv/bin/activate
# Start Jupyter Lab
jupyter-lab
Run any of the jupyter notebooks to visualize data and perform ML algorithms.
Data are available at this URL.
Since this is not always functioning, a data
folder was included in this package.
-
Scalability
-
Tests the memory/speed performances of 4 python libraries The results are shown below (test results from Processor: 3,1 GHz Dual-Core Intel Core i5 with Memory: 8 GB 2133 MHz)
File Size (MB) R80736.csv 51.41 R80721.csv 51.20 R80790.csv 51.03 R80711.csv 51.68 Library Time (sec) Max memory usage (MB) pandas 6.13 735.23 MB dask 18.38 603.20 MB vaex 3.11 268.40 MB modin 11.60 245.38 MB
-
-
Time Series and Forecast: learning from the past:
-
Calculates and visualises 3 quantities as function of time (Average Energy, Produced Energy and Capacity Factor)
-
Performs Auto-Regression analysis and Regularizing gradient boosting
-
-
Data Analysis
- Extract, transform, and load (ETL)
- Exploratory Data Analysis (EDA)
- Feature Selection Analysis
-
Machine Learning
- Perform several regression algorithms: Linear, Polynomial, Kernel Ridge Regression
- Perform pipeline including grid search analysis for parameter optimization
- Compare learning by plotting Learning Curve
Marco Di Gennaro
This project is licensed under the GPL v3 License - see the LICENSE.md file for details