Dr. Àngela Nebot

Universitat Politècnica de Catalunya-BarcelonaTech, Spain


Talk Title
A smart approach to real-time energy forecasting
Talk Abstract

The development of the Smart Grid (SG) is growing fast, encouraged by the environmental targets and the long-term goals set out be the European Commission Energy Roadmap. When coupled with smart metering systems, the smart grids reach consumers and suppliers by providing information on real-time consumption. With smart meters, consumers can adapt, in time and volume, their energy usage to different energy prices throughout the day. They will save money on their energy bills by consuming more energy in lower price periods.

The integration of time-series forecasting into existing energy grids and new SGs brings up some major technical challenges that have to be addressed. This also holds for the potential storage capacity for both electrical and thermal energy within energy networks, which can be achieved by intelligent demand side management. A major requirement in the SG is to schedule the trading of energy between different consumers and producers (prosumers). According to the SGs European Technology Platform, a large fraction of the generation capacity in 2035 will be stochastic and/or intermittent.

Forecasting permits better and more efficient management of all options. It is essential to integrate forecasting models of consumptions and/or productions that dwellings, public buildings, industries and possibly other structures, are going to generate. In order to achieve this goal, several approaches have been studied in the last decades; from the most classical statistical models to dynamic regression techniques. On the other hand, a large variety of Artificial Intelligent (AI) techniques have been applied in the field of short-term electricity consumption forecasting, showing a better performance than classical techniques.

Our work is framed in the AI field and is centered in the development of a robust hybrid methodology based on fuzzy logic and instance-based learning that deals with the main forecasting challenges. In these keynote talk I want to pay special attention to two of them:

  • Real time predictions: In contrast to other approaches where offline modelling takes considerable computational time and resources, the hybrid methodology proposed in this work generates fast and reliable models, with low computational costs. These models could be embedded, for instance, in a second generation of smart meters where they could generate on-site forecasting of the consumption and/or production in the next hours, or even trade the excess energy with other smart meters.


  • Dealing with missing data: It is essential to know how to react to certain situations where missing values are present both during the model generation and the online forecasting. As an example, when a sensor fails in a production process, it might not be necessary to stop everything if sufficient information is implicitly contained in the remaining sensor data.

The hybrid methodology developed has been used in 8 different buildings with different profiles of usage and located in different cities, thus affecting different climatology, consumption patterns, schedules and working days. In this keynote talk I would like to present the developed methodology and some of the results obtained when applied to different buildings.

Short Biography

Dr. Àngela Nebot earned her Ph.D. degree in artificial intelligence from the Universitat Politècnica de Catalunya – Barcelonatech. She is the head of the Soft Computing (SOCO) research group and a board member of the Intelligent Data Science and Artificial Intelligence (IDEAI) research Centre at the UPC. She is currently a Associate Professor at the Department of Computer Science, giving lectures on Computational Intelligence and Advanced Topics in Computational Intelligence at the Artificial Intelligence Master’s degree. Her current research interests include fuzzy, neuro-fuzzy, genetic-fuzzy systems and other soft computing hybridization techniques, modelling for prediction and decision support. The application areas include energy, medicine, biology, atmospheric sciences, music, risk management and e-Learning. She has published more than 40 articles in international research journals, more than 20 book chapters and more than 120 papers in international congresses. Dr. Nebot is a member of the Editorial Board of the International Journal of General Systems and collaborates as a reviewer in multiple international journals, such as Artificial Intelligence Communication, Artificial Intelligence in Environmental Engineering, Neurocomputing, IEEE Transactions on Fuzzy System and Fuzzy Sets and Systems among others. She is also reviewer of research projects from the Ministry of Science, Innovation and Universities of Spain, reviewer of research projects of the European Regional Development Fund and also reviewer of research project of the Junta de Andalucía and Universidad Complutense Madrid (Spain).

Talk Keywords
Smart Grid, Soft Computing, Fuzzy Inductive Reasoning, Entropy-based Feature Selection, Prediction with Missing Values, Energy Modelling.
Target Audience
Students, Post doctoral, Industry, Doctors and professors
Speaker-intro video

Sponsors & Partners