Privacy-preserving load forecasting in smart grids using federated learning: a comparative analysis of aggregation strategies

Entropy-Regularized Weighted Federated Meta-Learning Approach for Power Load Forecasting in Non-IID Scenarios

Federated learning-based privacy-preserving multisource data fusion for smart grids

This study addresses the challenges of privacy leakage and data silos in multi-source heterogeneous data interaction within smart grids by designing a federated multi-source data fusion architecture that combines adaptive local differential privacy with feature space alignment. This architecture utilizes Hessian matrix trace perception to adaptively adjust the local noise budget and introduces a dynamic aggregation selection mechanism based on the maximum mean difference, reconciling the conflict between differential privacy perturbations and feature manifold losses. Experimental results show that, while ensuring strict differential privacy boundaries, the system improves test accuracy by 7.45%, achieves a model inference speed of 45 FPS, and reduces communication resource overhead by 36.5%. Even under extreme conditions such as nonindependent identically distributed skew and 15% Byzantine poisoning attacks, it maintains a 98.40% attack interception rate and robust generalization fusion performance, providing a feasible system solution for building a highly reliable and resilient situational awareness and control foundation for the distribution IoT.

Initialization Is Critical: Advancing Federated Short-Term Load Forecasting under Load Heterogeneity via Model Initialization

Short-term load forecasting (STLF) provides essential information for numerous applications in modern power systems. However, accurate STLF often relies on fine-grained smart-meter data from distributed users, raising increasing concerns about data privacy. Federated learning (FL) has therefore emerged as a promising privacy-preserving paradigm for STLF. Nevertheless, this paper reveals structured heterogeneity in clients'load data. Specifically, clients exhibit different responses to exogenous factors and distinct temporal load profiles, which can degrade forecasting performance in FL. To mitigate these issues, this paper studies the role of model initialization in federated STLF, and proposes two initialization strategies from global and local perspectives. For global model initialization, when auxiliary public load data are available, a pretrained initialization strategy is developed to initialize the global model before federated training, thereby reducing client drift during the training process. For local model initialization, we propose SLIAvg, a sequential local initialization strategy that promotes a more consistent training process by allowing participating clients to start from progressively adapted models within each communication round. Since the proposed strategies only modify the initialization process, they are compatible with most existing FL frameworks and privacy-enhancing techniques. Experiments on real smart-meter data with two representative forecasting architectures demonstrate that the proposed strategies effectively improve forecasting performance, as evidenced by reduced client drift, improved convergence behavior, and lower forecasting errors.

A Causal-Driven Hierarchical Decentralised Federated Learning Framework for Resilient Load Forecasting in Distributed Microgrids

Modern microgrids require distributed intelligence and edge computing to handle variable demand and renewable generation, but heterogeneity, communication limits, and privacy hinder centralised forecasting. This paper proposes a causally guided hierarchical decentralised federated learning (H-DFL) framework for resilient short-term load forecasting, integrating a hybrid TCN–BiLSTM with MCMC-based probabilistic causal feature selection. A three-tier architecture enables local training and hierarchical aggregation without raw data sharing, improving scalability and communication efficiency through sparse, interpretable feature selection driven by key factors such as solar and weather dynamics. Experiments on the Ausgrid dataset show improved stability and efficiency over Granger Causality (GC), Dynamic Causal Modeling (DCM), and Markov chain Monte Carlo (MCMC) baselines, with intervention tests confirming robustness under solar, demand, and outage disturbances. Overall, the results demonstrate that combining probabilistic feature sparsity with hierarchical decentralised federated learning to enable scalable, privacy-preserving, and resilient load forecasting for future microgrid systems.

Confidence-Aware Federated Learning for Smart Load Characterization in Cyber-Physical DER Systems

The integration of smart meters into residential environments has allowed smooth collection of electricity consumption data, which is critical for demand response and coordinated operation in cyber-physical distributed energy resource systems. However, existing centralized methods of consumer characteristic identification pose significant risks to data privacy and confidentiality. Furthermore, the inherent limitations imposed due to noisy data cause a steep degradation in the accuracy and system-level decision-making. Addressing the issues of accuracy and data confidentiality, this paper proposes a confidence-aware federated learning framework for privacy-preserving inference of electricity consumer characteristics from raw smart meter data. The proposed method employs decentralized retention of smart meter data, using federated learning to refine network performance while ensuring that raw data remain localized at the client level. By evaluating a combined distribution of noisy and clean labels, erroneous data points are identified and excluded, thereby enhancing the model's efficacy and robustness. The effectiveness of the approach is validated using the Irish Commission for Energy Regulation dataset. The proposed framework demonstrates notable improvements, achieving an average gain of 3.97% in accuracy and 3.68% in MCC score compared to state-of-the-art methods, while maintaining strong data privacy guarantees.

Formal Logic Inference Guided Uncertainty Quantification for Personalized Federated Learning

Federated Learning (FL) enables privacy-preserving model training across heterogeneous distributed systems, such as smartgrid forecasting or traffic-flow prediction from geographically dispersed sensors and devices. A key challenge in such settings is capturing client-specific patterns while addressing data heterogeneity and uncertainty at scale. Existing approaches, including Bayesian Neural Networks (BNNs) and clustering-based methods, struggle with scalability and consistent personalization. We propose LogiCP, a novel FL framework that integrates formal logic reasoning with uncertainty quantification (UQ) to support scalable and personalized learning with theoretical guarantees. LogiCP uses Signal Temporal Logic (STL) to extract temporal patterns and form semantically coherent client clusters, controlling intra-cluster heterogeneity. Within each cluster, LogiCP applies decentralized Conformal Prediction (CP) to produce distribution-free prediction intervals with mathematical guarantees that encompass the real value. LogiCP dynamically assigns clients to clusters at runtime without retraining, improving practicality. Evaluations on three real-world datasets—traffic, temperature, and electricity—show that LogiCP consistently outperforms BNN-, clustering-, and CP-based baselines, achieving up to a 95% improvement in client-level MSE while maintaining strong scalability.