Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Electrical and Electronic Engineering | en_US |
| dc.contributor.advisor | Chung, C. Y. (EEE) | en_US |
| dc.creator | Liang, Zipeng | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/14386 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Flexible operation of hybrid AC/DC microgrids : from smart meter analytics to uncertainty modeling | en_US |
| dcterms.abstract | The global energy system is undergoing a fundamental transformation driven by decarbonization, decentralization, and electrification. Declining costs of renewable energy sources (RES), rapid growth in electric vehicles (EVs), and increasing demands for flexibility and resilience are reshaping modern power networks. Microgrids, localized systems capable of operating in both grid-connected and islanded modes, have emerged as a key enabler of this transition. In particular, hybrid AC/DC microgrids (HMGs), which integrate AC and DC subsystems through bi-directional converters (BDCs), offer enhanced efficiency by reducing unnecessary power conversions while accommodating diverse energy resources and loads. | en_US |
| dcterms.abstract | Despite their promise, the reliable and flexible operation of HMGs faces three interrelated challenges: non-convex BDC modeling, limited observability of load behavior, and significant renewable generation uncertainty. At the same time, the widespread deployment of smart meters enables appliance-level load analysis through non-intrusive load monitoring (NILM), which is essential for demand response and operational flexibility. However, existing NILM methods often struggle with low-frequency real-world data and fail to exploit contextual information. Moreover, prevailing uncertainty modeling techniques in microgrid optimization, such as convex hulls or polyhedral sets, tend to be overly conservative, sensitive to outliers, or computationally prohibitive. These limitations motivate the development of new data-driven models, learning frameworks, and uncertainty-aware optimization methods. | en_US |
| dcterms.abstract | This thesis addresses these challenges through a unified framework that integrates smart meter analytics, converter modeling, and robust optimization. Five core contributions are made. First, data-driven convex BDC models are developed to approximate efficiency characteristics while preserving convexity. Sufficient conditions are derived to guarantee non-simultaneous rectification and inversion, eliminating the need for binary variables and enabling scalable optimization. Case studies demonstrate substantial reductions in computational burden with minimal loss of accuracy. | en_US |
| dcterms.abstract | Second, a heterogeneous multi-gate mixture-of-experts (HMMOE) framework is proposed for NILM. By combining convolutional, recurrent, and attention-based experts with adaptive task weighting, the method jointly improves appliance state detection and power estimation. Extensive experiments on public and real-world datasets confirm significant performance gains over state-of-the-art approaches. | en_US |
| dcterms.abstract | Third, NILM is extended to a multimodal, dual-task architecture that incorporates contextual information such as weather and indoor conditions. An analytical task-weighting solution is derived, and the resulting appliance-level insights are embedded into a robust hybrid building microgrid model to evaluate demand response potential more accurately. | en_US |
| dcterms.abstract | Fourth, novel renewable uncertainty sets are introduced, including pairwise convex hull, intersected convex hull, and probability-driven data-correlated formulations. These sets provide adjustable conservativeness, tighter feasible regions, and improved out-of-sample performance compared with traditional approaches. | en_US |
| dcterms.abstract | Finally, tailored solution algorithms are developed for large-scale robust optimization involving renewable uncertainty, hydrogen storage, and EVs. Innovations include semi-convex relaxations, enhanced column-and-constraint generation, and quad-level decomposition with efficient uncertainty exploration. | en_US |
| dcterms.abstract | By bridging appliance-level data analytics and system-level decision-making, this thesis provides both theoretical foundations and practical tools for enabling flexible, intelligent, and robust operation of renewable-penetrated hybrid microgrids. | en_US |
| dcterms.extent | xvi, 262 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2026 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
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:
https://theses.lib.polyu.edu.hk/handle/200/14386