Seminar Announcement

12 Sep 2017 [Seminar Announcement] "Seeing events in 2D pictures with deep learning applied to visual cortex neural architectures: a tool of the future for power system control centres" 

Date:    14 Sep 2017 (Thursday)
Time:    18:30 – 20:00 
Venue: Room V302, Main Campus, PolyU
Speaker: Professor Vladimiro Miranda (IEEE Fellow)
No pre-registration is required. Certificate of attendance will be issued.

Abstract

PMUs are being installed and then control centers are becoming flooded with a tsunami of data. The information contained in such big data environment is largely left unused, but the PMU data contain signatures of dynamic events and the prompt recognition of such events could lead to better alarm processing, more accurate event diagnosis and possibly to preventive problem detection. The talk will reveal the most recent advances in dealing with this big data problem through deep learning with CNNs (Convolutional Neural Networks), whose architecture mimics the visual cortex of the human brain, and computing with GPUs (Graphic Processing Units). PMU real data from Brazil will show how the transformation of PMU time domain measurements of system frequency into 2D colour pictures allows a 100% success, with a CNN,  in recognizing and classifying dynamic events such as, for example, generator or line tripping, load disconnection of inter-area oscillation.
 
Speaker
Professor Vladimiro Miranda holds a Ph.D. degree (1982) from FEUP, the Faculty of Engineering of the University of Porto, Portugal. He has been a lecturer at FEUP since 1980, holding presently the position of Full Professor (Professor Catedrático) in the area of Power Systems, and President of INESC P&D Brasil, São Paulo, Brazil. 

Note 
The seminar is jointly organized by the IEEE-HK Joint Chapter of PES/IAS/PELS/IES and EE Dept of PolyU.  
For enquiries, please contact Professor XU Zhao at eezhaoxu@polyu.edu.hk 

[Read More]