CIGRE Kyoto 2022

Advanced distribution grid applications typically rely on grid models including precise phase connection information, which may suffer from inaccuracies due to human errors during the asset commissioning process as well as the potential lack of model updates when changes are made in the field. Traditional improvement of the phase labels model involves sending personnel into the field to do manual verification and labelling, which can be a tedious and expensive task. On the other hand, with the introduction of advanced metering infrastructure, data-driven methods could be used to improve the traditional grid models. 

This paper presents a novel hybrid clustering method for phase label identification that relies only on voltage magnitude measurements from customers’ smart meters. The method is applicable to a range of distribution grids and contains automatic hyperparameter tuning, making it easily transferable to a range of grids without tedious manual parameter tuning.

Accurate time synchronization plays a pivotal role in the effective operation of a modern power grid. As the industry is evolving toward a wider use of digital applications, real-time measurements are paramount for the operational continuity and stability of the electrical system. This fundamentally changes the role of time information, from pure historical correlation of events for fault analysis to a need for real-time situational awareness of correct grid operation.

Traditionally, GNSS-based solutions have been utilized in each substation to provide timing signals. However, concerns about the reliability of this approach are increasing since GNSS availability can be degraded accidentally or maliciously, resulting in erroneous satellite timing and, ultimately, in vital applications failing to operate correctly. Consequently, utilities are looking at ways to mitigate outages and disruptions related to GNSS reception.

With recent advancements in packet switching technology, time synchronization over packet networks has become a very attractive solution. IEEE1588v2 PTP can enable a highly secure and resilient network-based timing architecture with a 1us precision, while removing the risk of relying merely on satellite-based synchronization. 

This paper demonstrates the suitability of a PTP network clock approach versus a classical GNSS-based solution in providing accurate time of day information with high availability from a centralized grandmaster infrastructure to individual IEDs installed in substations. 

Oral session 6
Session chair : Prof. Hiroshi Asano, Tokyo Insitute of Technology
Technical expert : Alexandre Oudalov

When substation primary and/or secondary systems approach the end of their life cycle, the owners and operators are confronted with a series of questions on type and approach of the substation retrofit. Depending on network conditions, asset lifecycle status, operational requirements and constraints, as well as available technology, different approaches may be chosen. In this situation, an industrial customer in Japan decided to replace the existing switchgear and install a new, digital substation automation, protection and control system with an IEC 61850 process bus. The prime reason for choosing a digital substation solution, was the requirement to reduce the commissioning time and keep the interruption of the power supply to an absolute minimum during the retrofit process. 

During the retrofit, the existing GIS and indoor AIS switchgear were replaced with a new gas insulated system (GIS) and a new mixed technology system (MTS) switchgear, allowing to reduce footprint requirements and increase reliability and operational safety. And for the first time, also an IEC 61850 substation automation, protection, and control system (PACS) with IEC 61850 process bus was introduced in the Japanese industrial market. 

Secondary solutions of digital substation underly a changed or extended set of requirements, compared to conventional IEC 61850 systems, which has an impact how the system is designed to provide the required availability, reliability, performance, and maintainability. The employed solution is based on a process bus networks that are segregated per protection systems and bays, to enable bay-wise installation and also to simplify maintenance activities on individual system parts. 

Utilities are facing issues that makes critical activities more challenging . Assets are being used beyond their original design lifetime while demands on transmission equipment are rapidly changing as renewables cause frequent load fluctuations, yet the demand for reliability are higher than ever.  

Additionally, we can see increasing requirements on security and efficiency of power system operations as well as a grid code compliance for all connecting to the grid.  

Additionally, many new sensors have been installed in the substation to effectively maintain the substation assets, to reduce the operational cost and investments, to extend the lifetime of the assets and to better support the resources maintaining the assets. The integration of sensors in the substation is necessary and has become a major challenge for the utilities. Utilities still invest a lot of time and resources to collect the data as a basis for asset management and analytic performance model. Often the systems are deployed in silos and data are converted multiple time, which is error prone and results often in loss of information.  

The implementation of sensors is also rapidly increasing the volume of data available on the enterprise level as sensors and monitors generate more than ever before. Still too many data conversions are necessary to get useful data for the decision-making process.  

The challenge, however, is to use that data to improve decision making. What maintenance is needed? What time-based maintenance can be avoided? When should we replace an asset? And last but by far from least important, Do I comply to the latest Cybersecurity standards and requirements for my digital assets?  

A holistic solution is needed to support asset owners to better use the collected data form the substation. Digitization and Digitalization as major technology trends to address these challenges provide a unique opportunity to make data actionable. 

Scheduling of electricity generation, also known as unit commitment (UC), is a mission-critical task in grid operation. It is significantly increasing in complexity due to the growing share of renewable generation sources being added to grids as a result of the energy transition. This trend is driven by increasing numbers of (distributed) generation units involved in the energy markets, generation becoming more volatile, the necessity to solve UC problems on a finer time resolution, and evolution of the grid topology and possible contingency. 

Under this trend, the established optimization-based approaches using sequences of mixed integer programming (MIP) for solving the UC problem are expected to hit their computational limits very soon. Machine learning (ML) has recently been explored to mitigate this trend and enable faster and more robust MIP solutions. 

Most of the ML successes reported so far use ML models trained on the UC setup to be solved. This poses a restriction, in that the models need some training data, typically in the form of solved instances, before they can be used. In order to enable successful real-world applications, it would be ideal to have pre-trained ML models that generalize with little or no specific training data to a new UC setup. 

This paper focuses on improving the generalization capabilities of ML models to predict generation decisions. To this end, we propose graph neural networks (GNN) that use the solved LP relaxation of the MIP as an input. We show that these models have promising generalization capabilities across UC setups, both in terms of raw predictive performance and speeding-up the worst-case solving time of the UC MIP problems.

Power system resilience is becoming increasingly crucial as climate change drives more extreme weather events. During the CIGRE Virtual Centennial Session 2021, participants discussed the increasing importance of resiliency and how regulators adopt their framework to it. 

This paper suggests innovative solutions for sustainable and resilient infrastructure to reduce future risks and discusses the importance of long-term climate resilience of the energy system.  

The context for promoting resilience building is provided through historical and future climate-driven implications of extreme precipitation events and flooding in Europe, and a qualitative climate adaptation assessment that highlights the impacts of a recent flood event. Different elements that increase resilience can then be identified and innovative solution concepts for electricity-based resilient energy infrastructure can be developed. These concepts include technical countermeasures to protect the infrastructure against extreme weather catastrophes through improved design, installation, operation, monitoring, risk simulation and warning schemes. To ensure long-term resilience in a changing climate, the viability of adaptation solutions needs to be considered in a range of future climate and grid development scenarios.