Data-driven cyber-physical energy systems
Li, Qi
Li, Qi
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2024
Date Submitted
Keywords
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
Distributed solar energy resources (DSERs) in smart grid systems are rapidly increasing due to the steep decline in solar module prices. Smart cities, utilities, and government agencies are having pressure on managing stochastic power generation with this distributed solar energy penetration, such as predicting and reacting to the variation in the smart grid.
Recently, there is a rising interest in solar energy analytics to improve the efficiency of DSERs management. Unfortunately, DSERs management is still suffering from low efficiency and low fairness. For instance, DSERs performance models are not very accurate and cannot detect damage or degradation conditions accurately. The energy sharing among DSERs is suffering low efficiency and low fairness.
To address these problems, this dissertation presents data-driven cyber-physical energy systems that leverage data analytics techniques to improve the efficiency of DSERs management. My insight is that the smart grid ecosystem produces large volumes of data, which can be analyzed to improve DSERs management and CPS system sustainability. Building on this insight, I make the following contributions:
DSERs Identification Using Big Satellite Imagery Data. I design a new approach that can automatically detect distributed solar arrays in a given geospatial region (such as location and radius, zip code, or county name) without any extra cost. Rather than using Very High Resolution(VHR) aerial images, I leverage regular or low-resolution imagery datasets. I designed a two-step identification method to combine the benefits from both machine learning modelings and deep learning approaches. This dissertation shows that my SolarFinder can detect distributed solar arrays accurately.
However, prior DSER detection approaches, including SolarFinder, are still suffering low accuracy due to insufficient sample and feature learning. To further improve the detection performance, I design a new automatic system that can accurately detect and profile distributed solar photovoltaic arrays in a given region without any extra cost. I leverage multiple data augmentation techniques (e.g., CycleGAN, latent diffusion models, and Generative Adversarial networks) to build a large rooftop satellite imagery dataset (RSID) to learn more samples and features. This dissertation shows that my SolarDetector can accurately and reliably report multi-panel solar deployments.
Peer-to-Peer Solar Energy Sharing and Trading. To mitigate the impact of the intermittent DSERs while also benefiting from distributed generation for more reliable and profitable grid management, I design a new solar energy trading system that can enable unsupervised, distributed, scalable, and long-term fair solar energy trading in residential Virtual Power Plants(VPPs). I leverage a new multi-agent reinforcement learning approach that enables peer-to-peer solar energy trading among different DSERs to ensure that both the DSER users and the VPPs maximize benefit. The dissertation shows that SolarTrader is capable of enabling its users to trade solar energy with their neighbors more efficiently to achieve optimal monetary benefits while empowering the VPP with reliable, online, distributed DSERs management.
DSERs Profiling and Evaluation. Homeowners may have to spend up to $375 to diagnose their damaged rooftop solar PV system. To help inspect potential damage on solar PV arrays automatically and passively, I design a new system that automatically detects and profiles damages on rooftop solar PV arrays using their rooftop images at a lower cost. The dissertation shows that SolarDiagnostics can detect damaged solar PV arrays accurately.
For DSERs identification, I plan to expand to solar farms and smart and connected communities. For peer-to-peer energy trading, I plan to deploy a prototype of SolarTrader in two community-shared solar PV systems. I also plan to design new policies and expand the current DSER agent definition to abstract and integrate more DSERs into SolarTrader, such as EVs, utility battery banks, and wind-generated energy.
Associated Publications
Rights
Copyright of the original work is retained by the author.