We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Privacy-preserving Spatiotemporal Scenario Generation of Renewable Energies: A Federated Deep Generative Learning Approach

Abstract: Scenario generation is a fundamental and crucial tool for decision-making in power systems with high-penetration renewables. Based on big historical data, a novel federated deep generative learning framework, called Fed-LSGAN, is proposed by integrating federated learning and least square generative adversarial networks (LSGANs) for renewable scenario generation. Specifically, federated learning learns a shared global model in a central server from renewable sites at network edges, which enables the Fed-LSGAN to generate scenarios in a privacy-preserving manner without sacrificing the generation quality by transferring model parameters, rather than all data. Meanwhile, the LSGANs-based deep generative model generates scenarios that conform to the distribution of historical data through fully capturing the spatial-temporal characteristics of renewable powers, which leverages the least squares loss function to improve the training stability and generation quality. The simulation results demonstrate that the proposal manages to generate high-quality renewable scenarios and outperforms the state-of-the-art centralized methods. Besides, an experiment with different federated learning settings is designed and conducted to verify the robustness of our method.
Comments: Accepted by IEEE Transactions on Industrial Informatics
Subjects: Machine Learning (cs.LG); Signal Processing (eess.SP); Systems and Control (eess.SY)
Journal reference: IEEE Transactions on Industrial Informatics 18 (2022) 2310-2320
DOI: 10.1109/TII.2021.3098259
Cite as: arXiv:2107.07738 [cs.LG]
  (or arXiv:2107.07738v1 [cs.LG] for this version)

Submission history

From: Yang Li [view email]
[v1] Fri, 16 Jul 2021 07:18:22 GMT (2353kb,D)

Link back to: arXiv, form interface, contact.