|Title:||Bridging deep learning to power system state estimation with PMUs|
|Advisors:||Xu, Zhao (EE)|
|Subject:||Deep learning (Machine learning)|
Smart power grids
Hong Kong Polytechnic University -- Dissertations
|Department:||Department of Electrical Engineering|
|Pages:||xii, 130 pages : color illustrations|
|Abstract:||In smart grids, the periodical magnitude-based measurement by remote terminal units (RTUs) will be gradually replaced by real-time phasor measurement unit (PMU)-based measurements in the future. The PMU measurements supply synchronous and precise measurements of voltage/current phasors, i.e., synchrophasors at a rate of up to 60 samples per second. These measurements are widely used for the state estimation (SE), power system stability analysis and fault location, thereby improving the situational awareness of the grid operators.|
SE plays a vital role in contemporary energy management systems, where sufficient measurements must be provided to make the system observable. However, in a real power grid with thousands of buses, it is impossible to install PMUs wholly as they are costly. Therefore, it is necessary to investigate more economical and effective PMU placement frameworks to mitigate voltage estimation uncertainty. In addition, various incidents can result in the system unobservability, and SE cannot be implemented appropriately via conventional estimators. On the other hand, another critical issue is that classical SE methods are inapplicable in distribution systems without an ascertained network topology due to frequent reconfiguration actions and limited topology measurements. To cope with these challenges, the generative adversarial network (GAN)-based deep learning frameworks are proposed for the SE tasks of both transmission and distribution systems. The proposed framework is data-driven, model-free, and has a strong capability of handling missing data, which can result in the unobservability of the system per the classical SE method.
In this thesis, a comprehensive study is carried out to mainly investigate PMU-based SE where optimal PMU placement as well as deep learning-based SE algorithm are considered. The research background and purpose of this thesis are presented in Chapter 1. Chapter 2 proposes a reliability-based probabilistic optimal PMU placement approach to ensure minimal voltage magnitude estimation uncertainty under various operating scenarios, with supplementary PMUs installed in the power grid equipped with the SCADA system. Chapter 3 proposes a data-driven deep learning approach for power system static SE based on conditional GAN. Compared with classical SE methods, the proposed method does not require any prior knowledge of the system model. Without knowing the specific model, GAN can learn the inherent physics of underlying state variables purely with historical samples. Once the GAN model has been trained, it can estimate the corresponding system state accurately given the system raw measurements even with incompletions and corruptions. Chapter 4 proposes a novel data-driven deep learning approach for distribution system SE based on the topology-aware GAN (TAGAN). Compared to conventional methods, the new method can effectively estimate system states given contaminated or even missing measurements under varying network topology, representing the first effort of applying one integral deep learning framework for SE to address the uncertainties involved in both measured states and distribution grid topology simultaneously. Chapter 5 summarizes the whole thesis with some valuable conclusions drawn.
|Rights:||All rights reserved|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: