Georgia Institute of Technology
A Decentralized Privacy-Preserving Data-Driven Methodology for Energy Trading in the Smart Grid
Abstract
dc:description.abstractTo mitigate the climate change crisis, the Clean Power Plan (CPP) policy, the 2030 electricity generation decarbonization scenario, and the 2050 net-zero economy projection instruct and predict greenhouse gas emissions reduction and the need for grid reliability support. It has been suggested by the North American Reliability Corporation (NERC) that distributed load, i.e. commercial and residential buildings that are responsible for up to 32% of the total carbon dioxide emissions in the United States, are ideal candidates not only for energy savings but also for grid reliability improvement. Recent electricity market regulatory changes have allowed for the distributed load to participate in the demand-response market via an energy broker, the aggregator, and facilitate the aforementioned issues. This dissertation proposes a two-horizon, decentralized, and privacy-concerned methodology for extracting energy savings from the distributed load (prosumer) without the use of personally identifiable data. In the day-ahead horizon, a Stackelberg game-theoretical mechanism with bidirectional transaction ability is designed for the prosumer-aggregator game and an online and scalable with the number of prosumers algorithm is proposed for the extraction of approximate equilibria for all market players, while the cost of decentralization and privatization is theoretically estimated. In the real-time horizon, a demand correction mechanism is proposed for the improvement of demand deviations from the day-ahead bid to prevent market-imposed fines, based on a Physics-Informed Pattern Recognition Machine (PI-PRM) that automates the detection and demand correction suggestion processes with the use of non-personally identifiable prosumer measurements. Simulations with real-world data from the CAISO market and local prosumer data demonstrate the increased flexibility, energy savings, monetary rewards, and grid-side net demand ramping improvement achieved with the proposed methodology. The scalability and privacy concerns of the overall methodology make it not only applicable on a large scale but also promising for the incorporation of the market bidding process and the prosumer's local energy management system in the same customer end product.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Electrical and Computer Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kampezidou, Styliani Ioanna
- Advisors dc:contributor.advisor
-
- Mavris, Dimitri N.
- Romberg, Justin
- Committee members dc:contributor.committeemember
-
- Vachtsevanos, George
- Vamvoudakis, Kyriakos
- Patino-Echeverri, Dalia
- Puranik, Tejas
Subjects
dc:subject × 21- Smart Grid
- Distributed Load
- Prosumer
- Aggregator
- Emissions Reduction
- Energy Savings
- Ramping Support
- Grid Reliability
- Demand Response
- Sensor-based Demand Management
- Bidirectional Energy Trading
- Supervised Machine Learning
- Machine Learning Pipelines
- Pattern Recognition Machines
- Scalable Algorithms
- Decentralized Algorithms
- Bilevel Optimization
- Game Theory
- Approximate Stackelberg Equilibria
- Privacy Preservation
- Non-Personally Identifiable Data
Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/76734