Panel sessions

At the ISGT2024 conference, 3 panels, listed below, will be presented:

1. ML/AI applications in power systems: Drivers and barriers
2. Sharing Experiences on Workforce Challenges and Solutions for the Electric Power Industry
3. Effective dissemination of research results: Discussion with the Editor-in-Chiefs

ML/AI applications in power systems: Drivers and barriers

Time: TBD

Abstract: Application of Machine learning (ML) and Artificial Intelligence (AI) in power systems holds promise based on initial demonstrations. Many journal and conference papers, and panels at different conferences have been promoting future uses across monitoring, control and protection applications. However, most of the proposed applications are tested on synthetic data or ideal data coming from either small or simplified power system models. This did not encounter dealing with bad data and imprecise labels, and large-scale power systems complex dynamics found in the field environment. Furthermore, simplified assumptions also facilitated simplified feature engineering, and algorithm assumptions that proliferated wide use of the off-the -shelf ML/AI solutions. While the results looked promising, many of such applications did not reflect difficulties when implementing solutions to be deployed in a production environment where of-the-shelf algorithms require significant tuning. Also, most of the past solutions where not concerned with the sufficient evaluation that can justify the use of ML/AI over traditional solutions.
This Panel is focusing on experiences from recent developments dealing with real-life problems and applications facing both the opportunities and challenges of scaling such solutions to the cost-effective applications with clear return on investment. The speakers will illustrate the drivers and barriers stemming from their developments aimed at future deployments of ML/AI in the utility environment

Moderated by:  Mladen Kezunović, Texas A&M University, USA

Bio: Dr. Mladen Kezunovic (Life Fellow, IEEE) has been with Texas A&M University, College Station, TX, USA, for over 35 years, where he is currently a University Distinguished Professor, Regents Professor, Eugene E. Webb Professor, and the Site Director of the ‘‘Power Engineering Research Center’’ Consortium. He served for over 30 years as the Principal Consultant of XpertPower AssociatesTM, a consulting firm specializing in power systems data analytics. His expertise is in protective relaying, automated power system disturbance analysis, computational intelligence, data analytics, and smart grids. He is CIGRE Fellow, Honorary and Distinguished Member, registered Professional Engineer in Texas, and member of the US National Academy of Engineering 

Panelists:

1. Ricardo Bessa, INESC TEC, Portugal

Bio: Ricardo Bessa earned his Licenciado degree in Electrical and Computer Engineering from the University of Porto (UP) in 2006, followed by an M.Sc. in Data Analysis and Decision Support Systems from UP in 2008. In 2013, he completed his Ph.D. in the Doctoral Program in Sustainable Energy Systems (MIT Portugal) at UP. He is the Coordinator of the Center for Power and Energy Systems at INESC TEC. His research interests include renewable energy forecasting, computational intelligence applied to energy systems, decision-making under risk, and smart grids. He worked on several international projects, such as the European Projects FP6 ANEMOS.plus, FP7 SuSTAINABLE, FP7 evolvDSO, Horizon 2020 InteGrid, H2020 Smart4RES and coordinates the AI4REALNET project. IEEE Senior Member, Associate Editor of Journal of Modern Power Systems and Clean Energy, received the Energy Systems Integration Group Excellence Award in 2022.

Title:  “Human-AI frameworks and knowledge representation for AI in control room tasks”

Summary: This talk will showcase various AI use cases in system operations, highlighting the main benefits driving AI adoption, and it will discuss the need to establish an interdisciplinary framework for AI-based decision systems in critical infrastructures to overcome barriers like algorithm aversion. Additionally, we will emphasize the importance of effective knowledge representation—integrating diverse data sources and human expertise—to fully exploit available information. This approach enhances the interpretability and transparency of models and data for human users and decision-makers.

2. Jochen Cremer, Delft University of Technology, The Netherlands

Bio: Dr. Jochen Cremer works as Co-Director of the TU Delft AI Energy Lab, as Assistant Professor at the Faculty of Electrical Engineering, Mathematics, and Computer Science Delft University of Technology and as Principal Scientist at Austrian Institute of Technology. The Delft AI Energy Lab focuses on applying machine learning and data analytics to energy systems operation and control. He holds the PhD from Imperial College London.

Title:  “Status quo: Probabilistic Reliability Assessment with Deep Learning”

Summary: Preparing power systems for large-scale inverter-based resources requires future reliability tools to anticipate uncertainties in monitoring and controlling energy supply. Driven by the first successes in the energy domain, such as renewable or load forecasting, Deep Learning (DL) is promising for other applications. However, applying these DL-based approaches beyond forecasting is challenging as it involves considering multiple existing operating tools. The barrier to inter-operably implementing DL-based methods with other conventional tools requires domain adaptation beyond ‘plug-and-play’. Given this barrier, this talk introduces the concept of constraint-driven DL for probabilistic reliability assessment implemented in a security-constrained optimal power flow problem (SCOPF). Recently, a proposed DL approach modelled the loss function as a polynomial, consecutive matrix multiplication of the post-fault power flows with line outage distribution factors improving the scaling with k-outages. The implicit function theorem guarantees the pre-fault power flow is physically feasible, and the DL model learns pre-fault power settings that are N-k secure. Lastly, embedding DL in existing dynamic security assessment tools boosts the computational times to near real-time capabilities.

3. Panagiotis Papadopoulos, The University of Manchester, UK

Bio: Panagiotis Papadopoulos is currently a Reader (Associate Prof.) in the Department of Electrical and Electronic Engineering at the University of Manchester and a UK Research and Innovation Future Leaders Fellow working on “Addressing the complexity of future power system dynamic behaviour”. He received the Dipl. Eng. and Ph.D. degrees from the Department of Electrical and Computer Engineering at Aristotle University of Thessaloniki, in 2007 and 2014, respectively. From 2014-2017, he was a post-doctoral Research Associate at the University of Manchester and in 2017, he joined the University of Strathclyde as a Lecturer. He is currently the Technical Committee Program Chair for Power System Dynamic Performance committee of IEEE Power and Energy Society. His research interests are in the area of power system stability and dynamics under increased uncertainty, introduced due to the integration of new technologies. He is also interested in power system applications of machine learning to tackle complex problems related to power system stability.

Title:  “How can machine learning help with power system security assessment?”

Summary: The presentation will introduce key challenges linked to increased complexity and uncertainty related to dynamics and stability of modern power systems, and discuss how can machine learning help addressing those. Recent advances going beyond the notion that machine learning models are just powerful black box predictors will be presented, related to explainability/interpretability (understanding and gaining insights from ML models), physics informed (embed known physics in the training of ML models) and graph based (deal with changing topology) methods will be discussed. Remaining challenges related to implementation of machine learning based methods for dynamic security assessment will also be discussed.

4. Robert Eriksson, Uppsala University and Svenska Kraftnät, Sweeden

Bio: Robert Eriksson received his M.Sc. and Ph.D. degrees in Electrical Engineering from the KTH Royal Institute of Technology, Stockholm, Sweden, in 2005 and 2011, respectively. He was an Associate Professor at the Center for Electric Power and Energy at the Technical University of Denmark (DTU) from 2013 to 2015. He has several years of experience in operations and system stability development at the Swedish National Grid (Svenska kraftnät), where he has been employed since 2015 and is currently working part-time. He served as an Adjunct Professor at the KTH Royal Institute of Technology from 2020 to 2023. He is currently a Full Professor in the Division of Electricity at Uppsala University. His current research interests include power system dynamics and stability and control; data-driven, AI and machine learning approaches in power system stability; HVDC systems; and wide-area monitoring and control.

Title:  “Opportunities and Challenges with AI/ML in the Swedish Power System Operation”

Summary: The integration of AI and machine learning (AI/ML) technologies in the Swedish power system presents both opportunities and challenges. This presentation will explore how AI/ML can advance load forecasting, enhance frequency and voltage stability, and improve overall system operation. Despite these benefits, implementing AI/ML technologies presents several challenges, which will also be discussed. Key issues include the need for high-quality, sufficient data that is accessible and easily integrated for effective model training, as well as the integration with existing legacy systems. Additionally, the acceptance, transparency, and trust in AI/ML solutions within operational environments impact their effective deployment.

5. Balthazar Donon, RTE Research & Development, France

Bio: Balthazar Donon is a researcher at RTE R&D (Réseau de Transport d’Électricité). He is working on designing an AI algorithm that will provide advice to power grid operators on the best actions to take. He completed his studies at École polytechnique and Stanford University, where he obtained a MSc in Civil & Environmental Engineering. Later, he earned a PhD in Computer Science at Université Paris-Saclay and RTE R&D under the guidance of Isabelle Guyon, Marc Schoenauer, and Rémy Clément. Following this, he pursued postdoctoral research at the Université de Liège, collaborating with Prof. Louis Wehenkel on the development of an AI methodology for tertiary voltage control. Balthazar’s research interests lie in the energy domain and deep learning, specifically focusing on how graph neural networks can be used to create a true AI assistant for real-time operation. His goal is to create innovative artificial neural network algorithms tailored for real-life and real-time Power Systems applications.

Title:  “Topology-Aware Reinforcement Learning for Tertiary Voltage Control”

Summary: In recent years, transmission systems have seen an increase in the frequency and intensity of high voltage events. Traditional methods for optimal power flow are not well-suited for real-life systems, so there is a pressing need to develop new approaches to help operators improve tertiary voltage control. Fast neural networks could help address this challenge by offloading most of the computational burden to an offline training phase, resulting in extremely fast inference during operation. However, real-life power systems undergo significant topological variations such as asset disconnections and bus-splitting, making it difficult to represent them simply as vectors, which are typically required by feedforward neural networks. In this presentation, we will discuss the concept of hyper-heterogeneous multigraphs for data representation, along with a corresponding graph neural network architecture designed to seamlessly handle real system data. This architecture learns to operate voltages in an unsupervised manner by interacting with an industrial simulator.

Sharing Experiences on Workforce Challenges and Solutions for the Electric Power Industry

Time: Monday, October 14, 2024, 2:00-4:00 pm

Abstract: The Power industry is undergoing a remarkable transformation driven mainly by decarbonization and the resulting intensified electrification and grid modernization. The need for a qualified workforce for tackling these challenges has been pointed out as top priority by major power industry professionals. This Workforce panel is comprised of academic and industry leaders from Europe and the United States who will discuss the challenges, opportunities, unique solutions, and ideas for collaboration between academia, industry, and IEEE. Each person will give a 10 minute presentation, and this will be followed by audience Q&A and discussion. The later will be documented and made part of the formal report that will be issued by the PES Workforce Initiative Committee

Moderated by:  Wayne Bishop Jr. Chair of Industry Committee for IEEE PES Workforce Taskforce. Vice President, Quanta Technology

Panelists:

1. Dr. Luka Strezoski, University of Novi Sad, Serbia
2. Ambra Sannino, Vice President R&D Vattenfall, Sweden
3. Dr. Martha Symko-Davies, NREL, United States
4. Annika Moman, WSP, Gothenburg, Sweden
5. Pär Lundström, Senior Policy Advisor, The Swedish Installation Federation
6. Brittany McKannay, Director, Electric Engineering, Pacific Gas and Electric Company

Effective dissemination of research results: Discussion with the Editor-in-Chiefs

Time: TBD

The objective of the Pane/Round table is to present the main Power Engineering journals and discuss the challenges and opportunities of publication of original, theoretical and applied research results with prospective authors. In addition to brief introduction about individual journals, the focus of the round table will be on practices and approaches to presentation of research results and the advantages of effective communication of those to wider audience.

Moderated by: TBD

Panelists:

1. Prof Jovica V Milanovic, FIEEE, IEEE PES Vice President, Publications & immediate past Editor in Chief of IEEE Transactions on Power Systems, The University of Manchester, UK
2. Prof Claudio Cañizares, FIEEE, Editor in Chief of IEEE Transactions on Smart Grids, University of Waterloo, Canada
3. Prof Gianfranco Chicco, FIEEE, Editor in Chief of ELSEVIER’S International Journal of Sustainable Energy Grids and Networks, Politecnico di Torino, Italy
4. Prof. Antonio Gomez-Exposito, FIEEE, Vice Editor in Chief of Modern Power Systems and Clean Energy Journal, University of Seville, Spain.
5. Prof Vladimir Terzija, FIEEE, Editor in Chief of ELSEVIER’S International Journal of Electrical Power and Energy Systems, University of Newcastle, UK