Generation and Classification of Energy Load Curves Using a Distributed MapReduce Approach

Abstract

In nowadays energy markets, suppliers are encouraged to model the electricity consumption behavior of their customers in order to improve the quality of service and provide better products with lower investment and operating costs. New load models to support power system are required to mitigate scalability issues, especially considering the increasing penetration of distributed energy resources, varying load demands, and large volumes of data from smart meters. Smart metering allows obtaining detailed measures of the power consumption in the form of large time series that encode load curves. Clustering methods are applied to group costumers according to their similarity, by extracting characteristics of their behavior. Traditional computing approaches are not efficient to deal with the aforementioned problem, particularly when it must be solved in real time. This article proposes applying distributed computing and statistical learning methods to the problem of load curves classification of electricity consumers, applying the Map-Reduce model over the Hadoop framework. A case study, using real representative smart meter data from Uruguay is presented. The obtained results validate the stability and robustness of the approach. The main findings suggest that distributed computing can help electricity companies to deal with large volumes of data in order to improve energy management, provide services to consumers, and support modern smart grid technologies.