• {{searchSuggestions.title}}

Exploring electricity consumer behavior using big data


Madrid, 11th June 2018

Enel Foundation and Endesa Energia present an innovative study with Politecnico di Torino to learn more about residential consumer behavior, taking advantage of anonymized data collected by Smart meters with the implementation of Spain’s Real-Time Electricity Pricing Policy.  

This project represents a milestone on the use of big data to understand consumption behavior of about one million Spanish residential customers over one year.

The analysis of residential consumers is particularly challenging because unlike industrial clusters, families –even with similar composition and lifestyle- generally have very different consumption patterns. To address this complexity, a new methodology – called CONDUCTS - based on duration curves was developed to reduce the volume of hourly data analyzed. With the CONDUCTS methodology, the storyboard of all individual electricity profiles is reconstructed with good accuracy focusing on selected data points identified as representative of the consumer behavior.

As Giuseppe Montesano, Deputy Director of Enel Foundation explained, “The innovative methodology used allowed us to focus on selected cut points taken from the hourly data consumption profile. Building duration curves focused on only 9 specific consumption moments proved to be highly representative of the differences in the electricity usage of residential consumers”.

In order to achieve a meaningful analysis of hourly energy consumption over different timeframes, the research factored numerous parameters such as temperatures, type of market (free market vs regulated market) and hourly price variations, geographical location or level of contractual power.

The value of this research lies in the possibility to use such methodology to better group consumers on the basis of their dynamic behavior rather than on the aggregate level of consumption. In the future, this kind of capability can provide useful consumption prediction models and enable the design of customized commercial offers.

Academic and business oriented approach

This study represents a step forward for Endesa Energia Demand & Prices Department where a dedicated team is in charge of developing sophisticated mathematical models for analyzing customers’ behavior according to different habits. In particular, among other responsibilities, they apply machine learning techniques to manage relevant information from huge amounts of data, in order to optimize trading activities. To keep up to date with the latest developments and state of the art techniques, the team establishes collaborations with different research centers and universities and the current research project with Enel Foundation represents another landmark in this direction.

The methodology presented in this study could also be useful for other businesses at Enel, such us distribution companies in their grid management activities.

The results of the research have been presented in academic and scientific conferences, such as the 14th International Conference on Development and Application Systems in Romania (DAS2018) and the International Conference on Smart Energy Systems and Technologies in Spain (SEST2018).  The innovative nature of the research has earned the recognition of the peer-reviewers panel and the paper “Clustering-Based Assessment of Residential Consumers from Hourly-Metered Data” was distinguished with the SEST2018 Best Paper Award for the highest quality contribution.

Prof. Gianfranco Chicco from Politecnico di Torino, who attended the DAS 2018 and SEST 2018 conferences, commented "It has been a great opportunity for us to work together with Enel Foundation and Endesa Energia in this exciting project that allowed us to obtain an excellent result recognized by the international scientific research community." 

Clustering-Based Assessment of Residential Consumers from Hourly-Metered Data

PDF (1.23MB) Download

Discovering Electricity Consumption Over Time for Residential Consumers Through Cluster Analysis

PDF (1.2MB) Download