by Grigore Stamatescu, Department of Automatic Control and Industrial Informatics, University “Politehnica” of Bucharest

Abstract: The talk will provide a brief overview of current methods for energy time series forecasting using data mining and computational intelligence techniques with particular focus on building energy traces. This will include: the Matrix Profile, an efficient technique for time series exploratory analysis and data mining, typical regression models with domain-specific feature engineering and new end-to-end deep learning approaches by means of recurrent neural networks. An hierarchical energy system architecture with embedded control will be discussed which pushes prediction models to edge devices to address the challenge of managing demand-side response locally. We employ a two-step approach: At an upper level of hierarchy, we adopt a conventional machine learning pipeline to build load prediction models using automated feature extraction and selection. On a lower level of hierarchy, computed labels are used to train impact models realized by LSTM networks running on edge devices to infer the probability that the power consumption of the player contributes to the upper level prediction failure event. The system is evaluated on individual and aggregated public energy traces of academic buildings.

Speaker Bio: Grigore Stamatescu received the Dipl.-Ing. degree in automation and industrial informatics in September 2009 and the Ph.D. degree in systems engineering in January 2013, both from University Politehnica of Bucharest. He is an Associate Professor in the Department of Automatic Control and Industrial Informatics, University Politehnica of Bucharest. He is IEEE Senior Member and currently acts as the IEEE Romania Robotics and Automation chapter chair.

This event is technically supported by the IEEE Romania RAS Chapter within the seminar of Automation and Information Systems of the Department of Automatic Control and Computers, University “Politehnica” of Bucharest.