This work proposes Inertia-Informed Weighted FedAvg (IIWFedAvg), a physics-informed aggregation strategy that embeds generator inertia directly into global model fusion for transient stability control in transmission networks.
Abstract
Resilient-by-design smart grid control demands frameworks capable of maintaining stability under physical disturbances and communication failures, without reliance on centralized coordination. While Centralized Training Decentralized Execution (CTDE) enables a learning-based control paradigm at the grid edge, individually trained models fail to generalize across unseen fault contingencies and fall short of fully decentralized deployment. Federated learning (FL) restores generalization through collaborative training; however, standard aggregation strategies remain agnostic to the physical heterogeneity of synchronous generators. This work proposes Inertia-Informed Weighted FedAvg (IIWFedAvg), a physics-informed aggregation strategy that embeds generator inertia directly into global model fusion for transient stability control in transmission networks. The proposed framework further integrates interpretable Chebyshev Kolmogorov-Arnold Network (ChebyKAN)-based controllers, augmented with Rate-of-Change-of-Frequency (RoCoF) features to enhance dynamic response awareness. Evaluated on the IEEE 39-bus benchmark under full decentralized deployment, IIWFedAvg achieves a 75% generalization success rate across unseen fault contingencies. It also surpasses the centralized baseline in two out of three stabilized faults, while delivering a 3x improvement in stabilization speed at zero centralized coordination overhead.
Modern power systems, with the large-scale integration of renewable energy (RE) and distributed energy resources (DERs), have evolved into networked microgrid systems (NMGSs). While this transition aligns with sustainable development goals, it also introduces significant reliability challenges, including frequency instability caused by low system inertia, slow stochastic variations in load demand, communication time delays, and cyber–physical disturbances. Such scenarios make traditional centralized control strategies ambiguous, as system performance can drop significantly even in a simple case of communication disruption. To overcome these issues, a new distributed control framework for NMGS is developed based on federated learning fuzzy neural network (NN) optimized fractional-order PID (FLFNN FOPID). It is a federated architecture in which agents exchange only parameter updates, not raw operational data. This strategy preserves data confidentiality, reduces communication overhead, and enables control of synchronization frequency among DER. The proposed controller is then rigorously validated across normal load conditions, random disturbances, and phasor measurement unit (PMU) failure cases, while its practical implementation feasibility is additionally ascertained via hardware-in-the-loop (HIL) experiments on an OPAL-RT real-time simulator. Compared to the NN FOPID baseline, the proposed FLFNN FOPID reduces integral absolute error (IAE) by more than 69% for <inline-formula> <tex-math notation="LaTeX">$\Delta $ </tex-math></inline-formula>f1, with its value (<inline-formula> <tex-math notation="LaTeX">$2.4235\times 10^{-3}$ </tex-math></inline-formula>) compared with that of NN FOPID’s (<inline-formula> <tex-math notation="LaTeX">$0.7313\times 10^{-3}$ </tex-math></inline-formula>), while keeping track of improvement in almost similar ratios for each <inline-formula> <tex-math notation="LaTeX">$\Delta $ </tex-math></inline-formula>f2 and <inline-formula> <tex-math notation="LaTeX">$\Delta $ </tex-math></inline-formula>f3 independently too. It is important to note that the integral time absolute error (ITAE) in <inline-formula> <tex-math notation="LaTeX">$\Delta $ </tex-math></inline-formula>f2 decreased by almost 93% from <inline-formula> <tex-math notation="LaTeX">$0.1513\times 10^{-5}$ </tex-math></inline-formula> to <inline-formula> <tex-math notation="LaTeX">$0.0099\times 10^{-5}$ </tex-math></inline-formula>, which demonstrates a much-improved transient behavior. In all three MGs, integral squared error (ISE) and integral time-weighted squared error (ITSE) are also minimized, indicating reduced oscillatory behavior and improved system stability. Settling time reduced from 6.2 to 6.6 s under NN FOPID to 3.5–3.9 s under the proposed controller, corresponding to faster stabilization of about ~44%. Additionally, reductions in peak magnitude, rise time, peak time, and absolute error collectively indicate improved steady-state accuracy. These results have made the proposed FLFNN FOPID framework a powerful, privacy-preserving, and communication-friendly solution for frequency regulation of next-generation NMGSs under various circumstances.
Jeevitha Kandasamy, Sheila Mahapatra, Fahima Hajjej· IEEE Canadian Journal of Ele...· 0 citations
This paper formulate networked grid operation as a constrained decentralized partially observable Markov decision process and proposes a safe multi-agent collaborative learning framework that aims to reduce operating cost, load shedding, renewable curtailment, and carbon-relevant corrective burden.
Jia-Yi Zhang, Bing Fang, Huan-Xiu Xiao et al.· International journal of pat...· 0 citations
The rapid growth of large-scale interconnected systems, such as smart cities, industrial automation, and environmental monitoring, demands intelligent decision-making frameworks that are resilient, scalable, and resource-efficient. Traditional centralized intelligence approaches suffer from communication bottlenecks, high energy demands, and vulnerability to single points of failure, making them unsuitable for realworld deployment. This work introduces an event-triggered decentralized intelligence framework with energy-aware federated learning designed to address these challenges. In the proposed system, distributed nodes collaborate by exchanging model updates only when significant events or anomalies occur, rather than relying on continuous communication. This event-driven strategy substantially reduces bandwidth consumption while enabling timely adaptation to dynamic environments. To further enhance sustainability, the framework integrates energy-aware scheduling, allowing devices with limited power resources to contribute adaptively based on their energy profiles. A multilayer coordination mechanism ensures local autonomy and global consensus without centralized control. Experimental evaluations on representative real-world datasets demonstrate that the proposed method achieves competitive accuracy compared to conventional federated learning while reducing communication overhead by more than 40% and extending device lifetime in energy-constrained settings. Additionally, the framework incorporates Byzantine-resilient aggregation and is analyzed under communication latency and varying network topology conditions.
M. Kishore, N. Velmurugan· 2026 7th International Confe...· 0 citations
The rapid growth of renewable energy integration has increased the complexity of smart grid operation due to the intermittent nature of distributed energy resources and continuously varying load demand. Existing approaches often rely on centralized control or combine only selected intelligent technologies, limiting scalability, data privacy, and autonomous decision-making. This paper proposes a digital twin-driven federated multi-agent intelligence (DT-FMAI) framework that integrates virtual system synchronization, privacy-preserving distributed learning, and cooperative multi-agent control within a unified architecture. Digital twins continuously mirror physical grid assets, federated learning enables collaborative forecasting without sharing raw data, and intelligent agents coordinate energy management in real time. The framework was implemented in MATLAB/Simulink with TensorFlow Federated and evaluated using renewable generation, weather, battery, and load datasets. Results demonstrate a 15-25% reduction in forecasting error, 10-18% improvement in voltage regulation, 92-96% load-matching efficiency, and 12-20% higher energy efficiency. These outcomes demonstrate the potential of the proposed framework for scalable, secure, and intelligent renewable-integrated smart grid operation.
Pushpa Sreenivasan, K. Gattaiah, N. Hemalatha et al.· International Journal of Pow...· 0 citations
The results show that hierarchical feedback, bounded unit influence, stochastic aggregation, and compressed message passing can support stable privacy-preserving distributed coordination in distributed agent networks.
Atef Gharbi, Ahmad Alshammari, Nasser S. Albalawi et al.· Frontiers Comput. Neurosci.· 0 citations
In AC microgrids interconnected through a flexible DC system, false data injection attacks (FDIAs) targeting the control process of AC/DC converters may cause state deviations and even loss of synchronization among microgrids. To address these issues, this article proposes a distributed adaptive resilient restoration strategy. First, the FDIA mechanism model is established to characterize the effects of random and unbounded attack signals on distributed control. Second, a distributed secondary voltage and frequency control framework is developed based on consensus theory. The local adaptive compensation mechanism is further incorporated to enable coordinated correction and restoration of the voltage magnitude and frequency of AC microgrids. Finally, case studies are conducted on a multi-terminal AC system interconnected through a flexible DC network. Simulation results demonstrate that the proposed strategy effectively suppresses FDIA-induced state deviations and limits cross-regional disturbance propagation. It also restores voltage and frequency to stable operating conditions in real time, thereby enhancing coordinated stability and cyber resilience under various FDIA scenarios.
Yan-Song Zhao, Qian Xiao, Wen-Biao Lu et al.· Frontiers in Energy Research· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
