A federated learning framework for cyber-resilient defense in EV-integrated microgrids

Continuous deployment of moving target defense in modern power grids is often limited to intermittent use in practice because deliberately perturbing the network to expose stealthy cyber attacks pushes the dispatch away from its economic optimum and inflicts an operating-cost penalty that is paid with certainty in every hour of operation, whereas the attacks it prevents are sporadic. In the cluster studied here this penalty amounts to about 298 dollars per hour, roughly 1.4 percent of the base operating cost and 6.2 percent in the most congested microgrid, or about 2.6 million dollars per year. This paper proposes a unified cyber-physical architecture that absorbs most of this penalty while enforcing a minimum subspace-separation margin for attack exposure and preserving the data sovereignty of every participant. The defense layer selects each reactance reconfiguration through a decision-based multi-objective optimizer that simultaneously maximizes the geometric separation needed to expose hidden attacks, minimizes the induced operating cost, and minimizes the physical actuation effort of the flexible-transmission devices. The cost penalty created by the reconfiguration is then absorbed by mobilizing the bidirectional flexibility of electric vehicle fleets, which are dispatched precisely at the buses where the defense creates artificial congestion. Detection of the exposed attacks is delegated to a federated detector that is trained collaboratively across the cluster by exchanging only model parameters, so that raw measurements never leave a local microgrid. The framework is evaluated on a cluster of five interconnected IEEE 30-bus microgrids, each hosting five charging stations with one hundred electric vehicles. The results show that full participation of the vehicle fleet reduces the cluster defense penalty from about 298 to roughly 16 dollars per hour, a reduction close to 95 percent that leaves a small residual cost, while the federated detector reaches a detection area under the curve of about 0.92 and outperforms an isolated, data-scarce local detector. The exposed attacks raise the detection statistic far above its alarm threshold, supporting reliable and economically sustainable protection.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-16
DOI
https://doi.org/10.1038/s41598-026-78565-4
Primary Topic
Smart Grid Security and Resilience
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A federated learning framework for cyber-resilient defense in EV-integrated microgrids

Borzou Yousefi, Javad Gholami, Arash Yazdani, abdoreza noori shirazi
4 links
Scientific Reports
Smart Grid Security and Resilience
article

A federated learning framework for cyber-resilient defense in EV-integrated microgrids

Borzou Yousefi, Javad Gholami, Arash Yazdani, abdoreza noori shirazi
article en

Abstract

Continuous deployment of moving target defense in modern power grids is often limited to intermittent use in practice because deliberately perturbing the network to expose stealthy cyber attacks pushes the dispatch away from its economic optimum and inflicts an operating-cost penalty that is paid with certainty in every hour of operation, whereas the attacks it prevents are sporadic. In the cluster studied here this penalty amounts to about 298 dollars per hour, roughly 1.4 percent of the base operating cost and 6.2 percent in the most congested microgrid, or about 2.6 million dollars per year. This paper proposes a unified cyber-physical architecture that absorbs most of this penalty while enforcing a minimum subspace-separation margin for attack exposure and preserving the data sovereignty of every participant. The defense layer selects each reactance reconfiguration through a decision-based multi-objective optimizer that simultaneously maximizes the geometric separation needed to expose hidden attacks, minimizes the induced operating cost, and minimizes the physical actuation effort of the flexible-transmission devices. The cost penalty created by the reconfiguration is then absorbed by mobilizing the bidirectional flexibility of electric vehicle fleets, which are dispatched precisely at the buses where the defense creates artificial congestion. Detection of the exposed attacks is delegated to a federated detector that is trained collaboratively across the cluster by exchanging only model parameters, so that raw measurements never leave a local microgrid. The framework is evaluated on a cluster of five interconnected IEEE 30-bus microgrids, each hosting five charging stations with one hundred electric vehicles. The results show that full participation of the vehicle fleet reduces the cluster defense penalty from about 298 to roughly 16 dollars per hour, a reduction close to 95 percent that leaves a small residual cost, while the federated detector reaches a detection area under the curve of about 0.92 and outperforms an isolated, data-scarce local detector. The exposed attacks raise the detection statistic far above its alarm threshold, supporting reliable and economically sustainable protection.

Scientific Reports
Islamic Azad University, Tehran (IR)
Industry, innovation and infrastructure, Affordable and clean energy
Smart Grid Security and Resilience
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon