Reference : Federated Learning Meets Contract Theory: Economic-Efficiency Framework for Electric ...
Scientific journals : Article
Engineering, computing & technology : Electrical & electronics engineering
http://hdl.handle.net/10993/45097
Federated Learning Meets Contract Theory: Economic-Efficiency Framework for Electric Vehicle Networks
English
Saputra, Yuris M. [University of Technology Sydney]
Nguyen, Diep N. [University of Technology Sydney]
Dinh, Thai Hoang [University of Technology Sydney]
Vu, Thang Xuan mailto [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SigCom >]
Dutkiewicz, Eryk [University of Technology Sydney]
Chatzinotas, Symeon mailto [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SigCom >]
2021
IEEE Transactions on Mobile Computing
Institute of Electrical and Electronics Engineers
Yes (verified by ORBilu)
International
1536-1233
United States
[en] Federated learning ; privacy ; contract theory ; EV network ; demand prediction
[en] In this paper, we propose a novel energy-efficient framework for an electric vehicle (EV) network using a contract theoretic-based economic model to maximize the profits of charging stations (CSs) and improve the social welfare of the network. Specifically, we first introduce CS-based and CS clustering-based decentralized federated energy learning (DFEL) approaches which enable the CSs to train their own energy transactions locally to predict energy demands. In this way, each CS can exchange its learned model with other CSs to improve prediction accuracy without revealing actual datasets and reduce communication overhead among the CSs. Based on the energy demand prediction, we then design a multi-principal one-agent (MPOA) contract-based method. In particular, we formulate the CSs' utility maximization as a non-collaborative energy contract problem in which each CS maximizes its utility under common constraints from the smart grid provider (SGP) and other CSs' contracts. Then, we prove the existence of an equilibrium contract solution for all the CSs and develop an iterative algorithm at the SGP to find the equilibrium. Through simulation results using the dataset of CSs' transactions in Dundee city, the United Kingdom between 2017 and 2018, we demonstrate that our proposed method can achieve the energy demand prediction accuracy improvement up to 24.63% and lessen communication overhead by 96.3% compared with other machine learning algorithms. Furthermore, our proposed method can outperform non-contract-based economic models by 35% and 36% in terms of the CSs' utilities and social welfare of the network, respectively.
http://hdl.handle.net/10993/45097

File(s) associated to this reference

Fulltext file(s):

FileCommentaryVersionSizeAccess
Open access
TCM20.pdfAuthor preprint2.66 MBView/Open

Bookmark and Share SFX Query

All documents in ORBilu are protected by a user license.