earlyWARN

Partners: ETHZ (FEN), ZHAW-IEFE, Hitachi Energy, Swissgrid AG
Duration: 11/2023 - 11/2026
Funding: SFOE
Project Leader: ZHAW-IEFE
Project Team: Dr. C. Yaman Evrenosoglu, Dr. Alexander Fuchs, Dr. Turhan Demiray

external pageProf. Dr. Petr Korbacall_made, external pageDr.Miguel Ramírez-Gonzálezcall_made @ ZHAW-IEFE

• Identify a set of operation conditions under different energy target scenarios covering different pathways towards 2050, where different load and generation combinations are observed, covering different operation states.
• Identify different topologies considering (i) TYNDP actions, (ii) network configurations due to planned outages, and (iii) the location/type of critical disturbances.
• Perform dynamic stability simulations on the detailed ENTSO-E model with sufficiently accurate representations of DERs in the distribution networks for each operational state, topology, and scenario, to create a sufficiently large training dataset for the ML model.
• Identify the type of meaningful data to be used as the attributes to train the ML model.
• Train and test the ML model using the deep learning method, based on convolutional neural networks (CNN), to design the ''dynamic security assessor'', which will identify the critical sta-bility conditions online and autonomously provide recommendation for control actions.
• Assess the efficacy of the ML model in identifying precarious grid states and update the training data as required.

The TSOs have to plan for the operation of the system such that if a disturbance (e.g., short circuit, loss of a large load or a generation unit) occurs, the system is robust enough to contain the disturb-ance and the impacts of it on frequencies and voltages so that the electricity supply is uninterrupted. To achieve this, TSOs perform ''dynamic stability simulations'' to plan ahead of time for a limited set of operating conditions with selected disturbances so that the stability boundary conditions, and a list of mitigating actions can be identified. In this way, the TSOs can plan for preventive actions, if the studied system operation conditions are observed thanks to the time-stamped PMU measurements, to maintain the system robustness and to avoid cascading events triggered by disturbances and re-sulting in power outages. Dynamic stability simulations comprise the solution of linearized differential-algebraic systems of equations for large systems which are complex, requiring considerable time. This complexity has prevented the TSOs from adopting “online” dynamic security assessment simula-tion tools.

In combination with the electrification of e-mobility and heating, the increasing proliferation of converter-interfaced renewable generation is essential to reach the climate targets. The impacts of this energy transition on dynamic security are multifold and unprecedented, requiring a paradigm change in how TSOs perform dynamic security assessment due to new complexities in an increasingly stochastic system with a very large set of demand and generation combinations.

Research questions: 

• How can dynamic grid stability be assessed, and critical conditions be identified early on?
• And what are promising concepts for autonomous grid control?
   

• In this project, the analyses of multiple disturbances (generator outages, line outages, dif-ferent system splits) and multiple generator/load distributions in the continental ENTSO-E grid (capturing the typical variations throughout the year) are planned as opposed to a single dis-turbance and generator/load distribution in ACSICON.
• Simulations of various load flow scenarios over a full year horizon on the ETHZ-computer cluster are planned, as opposed to the variation of a single initial load flow in SCCER-FURIES. In addition, the proposed project investigates multiple security indicators and moni-toring parameters (ACSICON and SCCER-FURIES focused on one indicator only: the share of PV-production with and without grid support).
• Development of an online monitoring and security prediction tool is envisioned in this project. This was suggested as potential application in SCCER-FURIES, the proposed project is the realization of this recommendation for European TSOs. The increasing need for such monitoring tools was also stated by Swissgrid and other TSOs at the final dissemination presentation of the ACSICON project.
• ETHZ-FEN developed and used the high-performance power flow analysis and dynamic stability analysis tools in ACSICON and SCCER-FURIES. The proposed project combines these tools with the machine learning tools of ZHAW-IEFE, to develop an online dynamic security monitoring framework. The tools will be used to create the database with the offline training data, from which ML-based dynamic security indicators can be extracted and monitored based on online measurements.