Title: Forecasting power load curves from spatial and temporal mobile data

Authors: Frederico Coelho; Murilo Menezes; Lourenço Ribeiro; André Barbosa; Vinicius Silva; Antônio P. Braga; Carlos Natalino; Paolo Monti

Addresses: Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, CEP: 31270-901, Brazil ' Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, CEP: 31270-901, Brazil ' Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, CEP: 31270-901, Brazil ' Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, CEP: 31270-901, Brazil ' Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, CEP: 31270-901, Brazil ' Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, CEP: 31270-901, Brazil ' Department of Electrical Engineering, Chalmers University of Technology, Gothenburg, Sweden ' Department of Electrical Engineering, Chalmers University of Technology, Gothenburg, Sweden

Abstract: This work aims at applying computational intelligence approaches to telecommunication data, in order to associate mobile data to energy consumption load curves. Clustering methods are applied in order to allow the telecommunication network to infer about its topology and consumption load forecasting. Through an extensive analysis of Telecom Italia dataset and power distribution lines data available for the city of Trento, it was possible to confirm the high correlation between them, mainly when voice data is considered. To a great extent, this correlation can be explained by the fact that cellular communication devices are physically present in the service area of the distribution lines and when people are communicating, they are also consuming energy. Based on the aforementioned dataset, load curves for the city of Trento were constructed having as inputs data from telecommunication transactions. Results show that it is possible to use the telecommunication load as the input to predict the energy load, with the proposed model performing better than the naive predictor in 82% of the tested distribution lines.

Keywords: mobile data; smartphones; clustering; energy forecasting; learning.

DOI: 10.1504/WRSTSD.2020.105586

World Review of Science, Technology and Sustainable Development, 2020 Vol.16 No.1, pp.4 - 21

Accepted: 28 Aug 2019
Published online: 28 Feb 2020 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article