서브메뉴
검색
Differentiable Electricity Models for Power Grid Sensitivity Analysis and Planning
Differentiable Electricity Models for Power Grid Sensitivity Analysis and Planning
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202104741
- ISBN
- 9798290649580
- DDC
- 300
- 서명/저자
- Differentiable Electricity Models for Power Grid Sensitivity Analysis and Planning
- 발행사항
- [Sl] : Stanford University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 132 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-01, Section: B.
- 주기사항
- Advisor: El Gamal, Abbas.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2025.
- 초록/해제
- 요약Understanding, upgrading, and efficiently managing the electricity grid is essential to developing a clean, reliable, and affordable electricity system. Modern electricity systems are managed through a complex series of optimization problems that maximize overall welfare subject to reliability and physical constraints. The complexity of these systems makes analyzing the sensitivity of various outcomes, such as greenhouse gas emissions or nodal prices, mathematically and computationally challenging. Likewise, it is equally difficult to understand how future grid investments, e.g., new power generators or transmission lines, affect these outcomes.In this thesis, we show how techniques proven successful in machine learning can be applied to problems in electrical power systems. In particular, we apply implicit differentiation to develop differentiable models of electricity system operations. This enables us to effortlessly derive sensitivities such as marginal emissions rates or market power measures. Then we demonstrate how implicit differentiation can also be used to derive a gradient algorithm for long-term grid capacity expansion planning problems. This algorithm, which can be interpreted as an incremental process for upgrading system capacities, enables flexible and scalable modeling of bilevel, Stackelberg-like expansion planning problems. Furthermore, this method allows for warm starts, making it suitable for interactive grid planning. Finally, we show how to accelerate this framework for a broad class of electricity dispatch models using a differentiable, GPU-compatible implementation of the alternating direction method of multipliers (ADMM). Our implementation is linear-system-free, written in pure PyTorch, and end-to-end automatic differentiation compatible, allowing us to solve expansion planning problems with hundreds of millions of variables in a matter of minutes.We develop an open-source software package implementing these methods in a unified and modular framework. Our software is flexible and supports a wide array of system constraints used in practice. Since the algorithms we develop also scale well to large problems, the methods and software developed in this thesis are well-equipped to solve real-world power grid operations, sensitivity analysis, and planning problems.
- 일반주제명
- Load
- 일반주제명
- Expansion
- 일반주제명
- Sensitivity analysis
- 일반주제명
- Electricity
- 일반주제명
- Emissions
- 일반주제명
- Electricity distribution
- 일반주제명
- Carbon
- 일반주제명
- Prices
- 일반주제명
- Energy
- 기타저자
- Stanford University.
- 기본자료저록
- Dissertations Abstracts International. 87-01B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017358715
■00520260202104741
■006m o d
■007cr#unu||||||||
■020 ▼a9798290649580
■035 ▼a(MiAaPQ)AAI32149706
■035 ▼a(MiAaPQ)Stanfordpz322rh1205
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a300
■1001 ▼aDegleris, Anthony.
■24510▼aDifferentiable Electricity Models for Power Grid Sensitivity Analysis and Planning
■260 ▼a[Sl]▼bStanford University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a132 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-01, Section: B.
■500 ▼aAdvisor: El Gamal, Abbas.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2025.
■520 ▼aUnderstanding, upgrading, and efficiently managing the electricity grid is essential to developing a clean, reliable, and affordable electricity system. Modern electricity systems are managed through a complex series of optimization problems that maximize overall welfare subject to reliability and physical constraints. The complexity of these systems makes analyzing the sensitivity of various outcomes, such as greenhouse gas emissions or nodal prices, mathematically and computationally challenging. Likewise, it is equally difficult to understand how future grid investments, e.g., new power generators or transmission lines, affect these outcomes.In this thesis, we show how techniques proven successful in machine learning can be applied to problems in electrical power systems. In particular, we apply implicit differentiation to develop differentiable models of electricity system operations. This enables us to effortlessly derive sensitivities such as marginal emissions rates or market power measures. Then we demonstrate how implicit differentiation can also be used to derive a gradient algorithm for long-term grid capacity expansion planning problems. This algorithm, which can be interpreted as an incremental process for upgrading system capacities, enables flexible and scalable modeling of bilevel, Stackelberg-like expansion planning problems. Furthermore, this method allows for warm starts, making it suitable for interactive grid planning. Finally, we show how to accelerate this framework for a broad class of electricity dispatch models using a differentiable, GPU-compatible implementation of the alternating direction method of multipliers (ADMM). Our implementation is linear-system-free, written in pure PyTorch, and end-to-end automatic differentiation compatible, allowing us to solve expansion planning problems with hundreds of millions of variables in a matter of minutes.We develop an open-source software package implementing these methods in a unified and modular framework. Our software is flexible and supports a wide array of system constraints used in practice. Since the algorithms we develop also scale well to large problems, the methods and software developed in this thesis are well-equipped to solve real-world power grid operations, sensitivity analysis, and planning problems.
■590 ▼aSchool code: 0212.
■650 4▼aLoad
■650 4▼aExpansion
■650 4▼aSensitivity analysis
■650 4▼aElectricity
■650 4▼aEmissions
■650 4▼aElectricity distribution
■650 4▼aCarbon
■650 4▼aPrices
■650 4▼aEnergy
■650 4▼aIndustrial plant emissions
■690 ▼a0791
■71020▼aStanford University.
■7730 ▼tDissertations Abstracts International▼g87-01B.
■790 ▼a0212
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358715▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


