Skip to the main content

Original scientific paper

https://doi.org/10.20532/cit.2026.1006581

Local Differentially Private and Robust Edge–Fog Aggregation for Smart Grid Data

Song-Yao Feng ; Information Center of Guangxi Power Grid Co., Ltd, Nanning, China
Zheng-Yan Huang *
Ai Zhou ; Information Center of Guangxi Power Grid Co., Ltd, Nanning, China
Jun-Hao Song ; Information Center of Guangxi Power Grid Co., Ltd, Nanning, China

* Corresponding author.


Full text: english pdf 1.223 Kb

page 175-202

downloads: 0

cite


Abstract

Smart grid high-frequency electricity consumption data, while supporting load forecasting, regional status monitoring, and anomaly detection, may reveal users' behavioral patterns and power consumption characteristics. Therefore, a balance must be struck between privacy protection, communication efficiency, and robust aggregation performance. Addressing limitations of existing privacy-aggregation methods — including high communication overhead, unclear privacy guarantees, and interference from anomaly reports — we propose a local differential privacy-robust aggregation method for distributed smart grids. This approach applies trimming, random projection, and Laplacian noise to smart meter data at the user or edge side, ensuring the uploaded representations satisfy ε-local differential privacy under single-release conditions. Subsequently, robust fusion is performed at the fog node using coordinate median values and anomaly-aware weights to mitigate the impact of tampered summaries on regional aggregation results. Experiments evaluating 1-hour regional load forecasting and active power estimation demonstrated the method's effectiveness: MAPE scores of 4.89% and 5.02% across both datasets; cumulative communication overhead of 13.4×10⁵ bits in global experiments — lower than FL-PPDA, HE-PDA, and DP-EDA; and an aggregate error increase of 8.92% under data tampering attacks, below DP-EDA's 12.46%. The comprehensive results demonstrate that this method reduces communication load while maintaining acceptable aggregation utility, enhances aggregation stability under limited tampering conditions, and provides a reproducible, lightweight implementation approach for privacy-sensitive smart grid data aggregation.

Keywords

smart grid; edge computing; privacy protection; data aggregation; differential privacy

Hrčak ID:

353528

URI

https://hrcak.srce.hr/353528

Publication date:

17.11.2026.

Visits: 0 *