PMU- and AI-based analysis for a resilient operation of future power systems

Date: Tuesday, May 24                              Time: 1:30 pm – 3:00 pm (CEST)

Click here to download the presentations

Name of the organizer: Prof. Marjan Popov

Organization: Delft University of Technology, Netherldands

Email: M.Popov@tudelft.nl

Short biography of the chair: Marjan Popov (M’95–SM’03-F’22) obtained his Ph.D. degree in electrical power engineering from Delft University of Technology, Delft, The Netherlands, in 2002. In 1997, he was an Academic Visitor at the University of Liverpool, Liverpool, U.K., working in the arc research group on modeling SF6 circuit breakers. His major fields of interest are future power systems, large-scale power system transients, intelligent protection for future power systems, and wide-area monitoring and protection. He has been given invited lectures at many universities and keynote speeches at several conferences. Prof. Popov is a member of CIGRE and actively participated in WG C4.502 and WG A2/C4.39. In 2010 he received Hidde Nijland Prize for extraordinary research achievements. He is IEEE PES Prize Paper Award and IEEE Switchgear Committee Award recipient for 2011 and associate editor of the Elsevier’s international journal of electric power and energy systems.

Panel Abstract: This panel deals with several essentials related to resilient operation of future power systems. The first one is intentional controlled islanding of power systems, which is important for blackout prevention and based on the principle of isolating a part of the power system, which is affected by a severe disturbance allowing the rest of the system to safely operate. One of the ways to implement this is to determine the islands by identifying slow coherent groups of generating units. This information serves as a preventive-step of several emergency control schemes to identify power system control areas and improve transient stability. For this purpose, it will be elaborated on an efficient method (open-source available algorithm), capable of online and near real-time tracking of grouping changes of the slow coherent generators in a power system. Here, we will also look into the algorithm’s performance during various grid conditions. The panel also deals with is Artificial Intelligence (AI) application. Future power systems are subjected to faster dynamics, more fluctuations, more possible contingencies, more data, and as such more uncertainties. Therefore, the operating safety margins (and the economic costs) may increase as we lack future-ready operating tools. Here, the development of future methods by using operating (and synthetic) data to train AI models for real-time Dynamic Security Assessment (DSA) will be presented.
It will be shown: how AI methods correlate with physics resulting in reduced errors for DSA; how to train interpretable AI models, considering changing topologies, selecting the best model with minimal training data, and how to quantify risks when using them for real-time DSA. The limits and possible promising future works will also be addressed. Furthermore, we are keen to put attention on protection and stability phenomena caused by high penetration of converter interfaced generation (CIG), more specifically short-term voltage stability.
The phenomenon as such is rather complex and requires deep understanding of EMT processes, as well as high fidelity modelling of the system components involved into the nature of this new type of instability. Next to representative EMTP examples, we are putting our attention to data driven approaches to short-term voltage instability detection and monitoring. In this context we introduce Maximum Lyapunov Exponent-Based approach supported by wide area voltage measurements, traditionally obtained using Phasor Measurement Units. The entire approach will be presented and discussed from the perspective of its applicability and efficacy. Finally, increased penetration of distributed generation on distribution level results in a gradual change of distribution grids from passive to active. Hence there is increased necessity for monitoring and control of the distribution networks to ensure secure and reliable operation of the distribution grid. For this purpose, State Estimation (SE) is a crucial tool to provide the best system estimate for many downstream applications, such as optimal power flows, contingency analysis and dynamic stabilities.

The next part of the panel also deals with the implementation of distribution system state estimation (DSSE) algorithms, which takes into account the anomaly detection discrimination and identification (ADDI), in a distribution network. Both normal and abnormal operation scenarios occurring in a power system are simulated to validate the algorithms. Here, a Forecasting-aided State Estimator (FASE) using extended Kalman filter technique will be presented on the real-life 50 kV ring distribution grid, which makes use of PMUs. The developed platform can accurately and successfully estimate the system states under both normal and abnormal operations.

Panelist 1:

Name: Dr. Ir. Jochen Cremer

Organization: Delft University of Technology, the Netherlands

Email: J.L.Cremer@tudelft.nl

Short biography: Jochen Cremer is Co-Director of the Delft AI Energy Lab and Assistant Professor Intelligent Electrical power Grids at the Technical University of Delft. His expertise is on AI and ML technology for use cases in energy systems, ranging from demand response, distributed real-time control over centralised coordinated operations in real-time. His novel algorithms can process very large amounts of data and advance energy systems operations from societal, sustainable, and economic perspectives. Before he worked on Machine Learning technology at Imperial College London, control theory at Carnegie Mellon and MIT. He worked in the chemical and energy industry, in China and Germany. He holds an M.Sc. in Chemical Engineering, a B.Sc. in Electrical Engineering, and a B.Sc. in Mechanical Engineering from RWTH Aachen University, Germany. He is member of the IEEE PES Taskforces for Big data processing, members of CIGRE C2 working groups.

Title of presentation: AI Methods For Realtime Dynamic Security Assessment

Abstract: “The future has faster dynamics, more fluctuations, more possible contingencies, more data and more uncertainty. As we lack future-ready operating tools for the future, larger operating safety margins may be needed which raises economic costs. This talk will summarize the development of future methods by using operating (and synthetic) data to train AI models for real-time Dynamic Security Assessment (DSA). In more detail, this talk will show how AI methods can function together with conventional offline DSA. The talk then touches on how to train interpretable AI models, considering changing topologies, selecting the best AI model with minimal training data, and how to quantify risks when using them for real-time DSA. The talk will close by discussing the limits and possible promising future works.”

Panelist 2:

Name: Dr. Matija Naglic

Organization: TSO TenneT, the Netherlands

Email: Matija.Naglic@tennet.eu

Short biography: Matija Naglic received Uni. Dipl. Ing. degree in Electrical Engineering, study filed Telecommunications from the Faculty of Electrical Engineering, University of Ljubljana, Slovenia in 2011. In 2020 he received Ph.D. degree in Power System Engineering from the Faculty of Electrical Engineering, Mathematics and Computers Science, Delft University of Technology, The Netherlands. He is experienced with the implementation of IEEE std. C37.118 specifications, both the measurement and the telecommunication parts. He developed a cyber-physical platform for closed-loop control testing of WAMPAC applications, and two algorithms for online detection of disturbances and slow-coherent generators in a power system. He is a member of CIGRE WG C2.18. Currently, he works as an advisor for TenneT TSO, with a focus on the Control Room of the Future.

Title of presentation: Online Identification Of Slow Coherent Generators

Abstract: In a power system, slow coherency can be applied to identify groups of the generating units, the rotors of which are swinging together against each other at approximately the same oscillatory frequencies of inter-area modes. This information serves as a prerequisite step for several emergency control schemes to identify power system control areas and improve transient stability. First I will discuss the challenges related to model and measurement-based coherency identification of generators. Next, I will elaborate on a recently developed and open-source available algorithm, capable of online and near real-time tracking of grouping changes of the slow coherent generators in a power system. Here, we will also look into the algorithm’s performance during various grid conditions. The talk will be concluded with possible further research directions.

Panelist 3:

Name: Prof. Dr. Vladimir Terzija

Organization: Skoltech, Russian Federation

Email: v.terzija@scoltech.ru

Short biography: Vladimir Terzija (M’95–SM’00-F’16) was born in Donji Baraci (former Yugoslavia). He received the Dipl-Ing., M.Sc., and Ph.D. degrees in electrical engineering from the University of Belgrade, Belgrade, Serbia, in 1988, 1993, and 1997, respectively. He is the Engineering and Physical Science Research Council Chair Professor in Power System Engineering with the School of Electrical and Electronic Engineering, The University of Manchester, Manchester, U.K., where he has been since 2006. From 1997 to 1999, he was an Assistant Professor at the University of Belgrade, Belgrade, Serbia. From 2000 to 2006, he was a senior specialist for switchgear and distribution automation with ABB AG Inc., Ratingen, Germany. His current research interests include smart grid application of intelligent methods to power system monitoring, control, and protection; wide-area monitoring, protection, and control; switchgear and fast transient processes; and digital signal processing applications in power systems.

Title of presentation: Data-driven and PMU-based solutions for short-term voltage instability detection and monitorin

Abstract: Protection and stability phenomena caused by high penetration of converter interfaced generation (CIG) are today in the focus of a number of studies and practical, industry driven, activities. One of phenomena, identified as a recognizable challenge is short-term voltage stability. The phenomenon as such is rather complex and requires deep understanding of EMT processes, requiring high fidelity modelling of the system components determining the nature of this new type of instability. Next to representative EMTP examples, we are putting our attention to data driven approaches for short-term voltage instability detection and monitoring. In this context we introduce Maximum Lyapunov Exponent-Based approach supported by wide area voltage measurements, traditionally obtained using Phasor Measurement Units. The entire approach will be presented and discussed from the perspective of its applicability and efficacy

Panelist 4:

Name: Prof. Marjan Popov

Organization: Delft University of Technology, Netherlands

Email: M.Popov@tudelft.nl

Short bio and photo: included above in the Panel Chair section

Title of presentation: Anomaly Detection by Applying using Real-time State Estimation

Abstract: Increased penetration of distributed generation on the distribution level results in a gradual change of distribution grids from passive to active. Hence there is an increased necessity for monitoring and control of the distribution networks to ensure secure and reliable operation of the distribution grid. For this purpose, State Estimation (SE) is a crucial tool to provide the best system estimate for many downstream applications, such as optimal power flows, contingency analysis, and dynamic stabilities. The next part of the panel also deals with the implementation of distribution system state estimation (DSSE) algorithms, which take into account the anomaly detection discrimination and identification (ADDI), in a distribution network. Both normal and abnormal operation scenarios occurring in a power system are simulated to validate the algorithms. Here, a Forecasting-aided State Estimator (FASE) using an extended Kalman filter technique will be presented on the real-life 50 kV ring distribution grid, which makes use of PMUs. The developed platform can accurately and successfully estimate the system states under both normal and abnormal operations.