Methodology (graphical abstract)

ISAC algorithm

Example of a model that characterizes the consumption profiles of a city in Spain

About the project

URSUS_CONSUMPTION_PROFILES is a script that implements a methodology to characterize the domestic electricity consumption profiles of any city using clustering algorithms and a new algorithm (k_ISAC_TLP) that allows determining the optimal number of clusters that best represents the consumption profiles.

For a more detailed description of the methodology and K_ISAC_TLP algorithm you can read the paper.... (añadir enlace al artículo)

...

Technologies

• Python
• Pyspark
• Jupyter Notebook
• MLIB

Datasets

Any dataset where each row represents the hourly consumption (24 hours) of the day for any consumer. The datasets that have been used for the publication of the article are private, so they cannot be uploaded to the repository.

Script content

The script implements the following functionalities:

• Generation clustering models using big data k-means algorithm for kMin and kMax values for the number of clusters
• Calculation the MAE of each model
• Calculation the number of small cluster of each model (less than 1% of the total observations)
• Calculation of optimal k-value (best model) analyzing the MAE for each model curve, and the curve with the number of small clusters for each model aplying k-ISAC algorithm.

Requirements

In order to run the script without problem, it is necessaryis necessary to have installed:

• Jupyter Notebook (Environment to run script cells). Script has comments on the main functionalities that each code cell implements.
• Python (main script language)
• Pyspark (spark with python)
• MLIB (machine Learning with Spark Library)

Instalation

1. Clone repo

git clone  https://github.com/ursusdm/consumptionprofiles.git

2. Open script with jupyter notebook and run cells

Contact

José del Campo Ávila - jcampo@uma.es
Llanos Mora Lopez - llanos@uma.es
Francisco Rodríguez Gómez - francisco.rdg.gmz@uma.es

ACKNOWLEDGEMENTS

• This work has been supported by the project RTI2018-095097-BI00 in the call for projects I+D+i 2018 from Ministerio de Ciencia, Innovación 𝑦 Universidades, Spain.