| 1. |
- Li, Yufei, et al.
(författare)
-
Machine Learning at the Grid Edge : Data-Driven Impedance Models for Model-Free Inverters
- 2024
-
Ingår i: IEEE transactions on power electronics. - : Institute of Electrical and Electronics Engineers (IEEE). - 0885-8993 .- 1941-0107. ; 39:8, s. 10465-10481
-
Tidskriftsartikel (refereegranskat)abstract
- It is envisioned that the future electric grid will be underpinned by a vast number of smart inverters linking renewables at the grid edge. These inverters' dynamics are typically characterized as impedances, which are crucial for ensuring grid stability and resiliency. However, the physical implementation of these inverters may vary widely and may be kept confidential. Existing analytical impedance models require a complete and precise understanding of system parameters. They can hardly capture the complete electrical behavior when the inverters are performing complex functions. Online impedance measurements for many inverters across multiple operating points are impractical. To address these issues, we present the InvNet, a machine learning framework capable of characterizing inverter impedance patterns across a wide operation range, even with limited impedance data. Leveraging transfer learning, the InvNet can extrapolate from physics-based models to real-world ones and from one inverter to another with the same control framework but different control parameters with very limited data. This framework demonstrates machine learning as a powerful tool for modeling and analyzing black-box characteristics of grid-tied inverter systems that cannot be accurately described by traditional analytical methods, such as inverters under model-predictive control. Comprehensive evaluations were conducted to verify the effectiveness of the InvNet.
|
|
| 2. |
- Liao, Yicheng, et al.
(författare)
-
Neural Network Design for Impedance Modeling of Power Electronic Systems Based on Latent Features
- 2024
-
Ingår i: IEEE Transactions on Neural Networks and Learning Systems. - : Institute of Electrical and Electronics Engineers (IEEE). - 2162-237X .- 2162-2388. ; 35:5, s. 5968-5980
-
Tidskriftsartikel (refereegranskat)abstract
- Data-driven approaches are promising to address the modeling issues of modern power electronics-based power systems, due to the black-box feature. Frequency-domain analysis has been applied to address the emerging small-signal oscillation issues caused by converter control interactions. However, the frequency-domain model of a power electronic system is linearized around a specific operating condition. It thus requires measurement or identification of frequency-domain models repeatedly at many operating points (OPs) due to the wide operation range of the power systems, which brings significant computation and data burden. This article addresses this challenge by developing a deep learning approach using multilayer feedforward neural networks (FNNs) to train the frequency-domain impedance model of power electronic systems that is continuous of OP. Distinguished from the prior neural network designs relying on trial-and-error and sufficient data size, this article proposes to design the FNN based on latent features of power electronic systems, i.e., the number of system poles and zeros. To further investigate the impacts of data quantity and quality, learning procedures from a small dataset are developed, and K-medoids clustering based on dynamic time warping is used to reveal insights into multivariable sensitivity, which helps improve the data quality. The proposed approaches for the FNN design and learning have been proven simple, effective, and optimal based on case studies on a power electronic converter, and future prospects in its industrial applications are also discussed.
|
|
| 3. |
- Rolander, Arvid, et al.
(författare)
-
Real-time transient stability early warning system using Graph Attention Networks
- 2024
-
Ingår i: Electric power systems research. - : Elsevier BV. - 0378-7796 .- 1873-2046. ; 235
-
Tidskriftsartikel (refereegranskat)abstract
- In this paper, a classifier based early warning system is designed, trained and tested based on time-series of Phasor Measurement Unit (PMU) measurements at all buses in a power system. The classifier is based on a novel combination of Graph Attention Networks and Long Short-Term memories, and is trained to label power system data in the form of captured windows of PMU measurements. These labels are then used to provide early warning for transient instability. The classifier is trained and tested data from simulations of the Nordic44 test system, and includes extensive topological variations under two different load levels. It is found that accurate early warnings can be provided, but the quality of prediction is highly dependent on specific power system characteristics, such as how quickly the power system responds to transient disturbances.
|
|
| 4. |
- Cardias, Ramon, et al.
(författare)
-
Unraveling the connection between high-order magnetic interactions and local-to-global spin Hamiltonian in noncollinear magnetic dimers
- 2023
-
Ingår i: Physical Review B. - : American Physical Society (APS). - 2469-9950 .- 2469-9969. ; 108:22
-
Tidskriftsartikel (refereegranskat)abstract
- A spin Hamiltonian that characterizes interatomic interactions between spin moments is highly valuable in predicting and comprehending the magnetic properties of materials. Here, we explore a method for explicitly calculating interatomic exchange interactions in noncollinear configurations of magnetic materials considering only a bilinear spin Hamiltonian in a local scenario. Based on density-functional theory calculations of dimers adsorbed on metallic surfaces, and with a focus on the Dzyaloshinskii-Moriya interaction (DMI) which is essential for stabilizing chiral noncollinear magnetic states, we discuss the interpretation of the DMI when decomposed into microscopic electron and spin densities and currents. We clarify the distinct origins of spin currents induced in the system and their connection to the DMI. In addition, we reveal how noncollinearity affects the usual DMI, which is solely induced by spin-orbit coupling, and DMI-like interactions brought about by noncollinearity. We explain how the dependence of the DMI on the magnetic configuration establishes a connection between high-order magnetic interactions, enabling the transition from a local to a global spin Hamiltonian.
|
|
| 5. |
- Cheng, Li, et al.
(författare)
-
Neural-Network-Based Impedance Estimation for Transmission Cables Considering Aging Effect
- 2023
-
Ingår i: 2023 8th IEEE Workshop on the Electronic Grid, eGRID 2023. - : Institute of Electrical and Electronics Engineers (IEEE).
-
Konferensbidrag (refereegranskat)abstract
- In power-electronic-based power systems like wind farms, conducting stability analysis necessitates a comprehensive understanding of the system impedance across a wide frequency range, from sub-harmonic frequencies up to the Nyquist frequency of control systems of power converters. The cable aging effect can significantly impact the cable impedance, while accurately estimating the degree of aging proves challenging. To avoid the requirement for precise aging prognostic, this paper proposes an approach based on Artificial Neural Networks (ANN) that enables the estimation of AC cable impedance in a wind farm solely through fundamental frequency measurements. The data used for training the ANN is obtained from the cable model in PSCAD, incorporating physical and geometrical parameters, which accurately approximates real cables within power systems. The training results of the ANN validate the accuracy of the proposed identification approach. As a result, the proposed approach effectively eliminates the potential misjudgment of system stability caused by the aging effect of power cables.
|
|
| 6. |
- Cheng, Li, et al.
(författare)
-
Online Identification of Wind Farm Wide Frequency Admittance with Power Cables Using the Artificial Neural Network
- 2023
-
Ingår i: 2023 IEEE Energy Conversion Congress and Exposition, ECCE 2023. - : Institute of Electrical and Electronics Engineers (IEEE). ; , s. 1530-1535
-
Konferensbidrag (refereegranskat)abstract
- In power-electronic-based power systems like wind farms, stability analysis requires knowledge of system impedance across a wide frequency range, from sub-harmonic frequencies to the Nyquist frequency. Although it is feasible to take the fundamental frequency measurement during power system operation, obtaining a wide-frequency impedance curve in real time is very challenging. This paper proposed an ANN-based approach to estimate wide-frequency system admittance of wind farms with power cables, through fundamental frequency measurements. Real-life uncertainties are considered, including shunt capacitor injection, filter inductance variance, cable aging, errors in voltage and current measurements, and the variance of other system parameters. The generalization ability of the ANN is validated on a new dataset with different uncertainty distributions, and the error sensitivity to the potential system parameter variance is evaluated. These results can be referenced in the data acquisition step in future neural-network-based applications.
|
|
| 7. |
- Forsberg, Samuel
(författare)
-
Power Grid Resilience to High Impact Low Probability Events
- 2023
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The electrification of societies and the decarbonisation of electricity production are changing energy systems worldwide. A fast transition towards the replacement of fossil fuels by intermittent renewable energy sources is expected in the next decades to combat climate change. A significant share of the produced electricity is likely to be generated from offshore wind farms, due to the abundant wind resources in the offshore regions and the lack of available onshore sites. However, increased electricity dependence in combination with expanded offshore wind power generation introduce new vulnerabilities to the society. Specifically, the effects of high impact low probability (HILP) events are considered as potential threats to the power system, not least because of the increasing number of extreme weather events. Therefore, research on power grid vulnerability and power system resilience to HILP events are of significant interest.This thesis presents results of studies investigating power grid vulnerability from a topological perspective, and resilience to storm conditions of power systems with varying dependencies on offshore wind. To achieve this, methods based on complex network theory and AC power flow analysis have been developed, tested, and evaluated. Further, geospatial wind data from historical extreme storm events have been used to generate realistic power production profiles from hypothetical offshore wind farms.The results strengthen that complex network concepts can be used successfully in the context of power grid vulnerability analysis. Further, the results show that the resilience of power systems with large dependencies on offshore wind differ vastly depending on the grid properties and control strategies, which are further discussed in this thesis.
|
|
| 8. |
- Lu, Yizhou, et al.
(författare)
-
An Online Digital Twin based Health Monitoring Method for Boost Converter using Neural Network
- 2023
-
Ingår i: 2023 IEEE Energy Conversion Congress and Exposition, ECCE 2023. - : Institute of Electrical and Electronics Engineers (IEEE). ; , s. 3701-3706
-
Konferensbidrag (refereegranskat)abstract
- This paper proposes a neural network-based digital twin for online health monitoring of vulnerable components in converters. The proposed digital twin consists of a physics-informed model with uncertain parameters, and a neural network (NN) for real-time model updating and health monitoring of components. This method is noninvasive, without extra circuits, and can identify parameters in real-time with high efficiency. Simulation and experiment are conducted to validate the effectiveness of the proposed method in accurate parameter identification and degradation monitoring of capacitor and MOSFET.
|
|
| 9. |
- Natvig, Filip, et al.
(författare)
-
A Learning Testbed for False Data Injection Attacks
- 2023
-
Ingår i: Proceedings of 2023 IEEE PES Innovative Smart Grid Technologies Europe, ISGT EUROPE 2023. - : Institute of Electrical and Electronics Engineers (IEEE).
-
Konferensbidrag (refereegranskat)abstract
- This paper presents a user-friendly software-based power system cybersecurity testbed for educational purposes, focusing on false data injection attacks (FDIAs). While numerous testbeds are already available for experimenting with FDIAs, the high level of sophistication associated with these testbeds often requires users to possess extensive knowledge of the different hardware and software systems involved in power system control. This project aims to introduce cybersecurity to a broader crowd and, by extension, promote research in the field. Specifically, we propose an easy-to-use application with a graphical user interface (GUI) to allow non-computer-familiar users to utilize it. The current version of the application allows users to simulate the stationary behaviour of an IEEE test system, test it with customized FDIAs, and monitor the system’s response through the GUI. Currently, the application is in a proof-of-concept stage, and to encourage contributions from well-meaning developers, we have open-sourced the project for community involvement. We also suggest several potential application improvements to guide the development process in the right direction.
|
|
| 10. |
- Rolander, Arvid, et al.
(författare)
-
Real-Time Power System Stability Monitoring using Convolutional Neural Networks
- 2023
-
Ingår i: Proceedings of 2023 IEEE PES Innovative Smart Grid Technologies Europe, ISGT EUROPE 2023. - : Institute of Electrical and Electronics Engineers (IEEE).
-
Konferensbidrag (refereegranskat)abstract
- In this paper, we present a method for real-time transient stability prediction based on convolutional neural networks (CNN) using a novel CNN architecture compared to previous works on the topic. The method is based on monitoring voltage phasor and frequency measurements at the generator terminal buses, which are presented to the neural network in the form of three channel RGB images, taken as a sliding window. The sliding window consists of the most current set of measurements, as well as the four most recent historical measurements for a total of five time steps. When deployed, the neural network is continuously fed real-time measurements, thereby functioning as a real-time stability monitoring system. The neural network is trained and tested on simulated data of the Nordic32 system using five-fold cross-validation. The trained classifier is able to accurately predict instability in all but one case from the test set. In the successfully identified unstable cases, instability was predicted five cycles after fault clearance.
|
|
