Reliability estimation algorithm of power grid data based on blockchain and federated learning
The rapid development of the energy Internet has made power grid data increasingly critical in energy management, supply–demand balancing, and fault prediction, making accurate reliability estimation essential for secure and stable power system operation. However, the complex topological structure and multi-level organizational relationships of power grids render them vulnerable to external interference, often resulting in data island phenomena and delays in data transmission and integration, which collectively undermine the accuracy of reliability estimation. To address these challenges, this paper proposes a power grid data reliability estimation algorithm that integrates blockchain technology with federated learning: blockchain encrypts grid data information by determining transaction fragment counts based on blockchain nodes, calculating input and output values of common parameters, constructing a judgment function for traceability certification, and computing tampering success rates to ensure data integrity and security, while federated learning trains the encrypted data across distributed nodes without exposing raw data, enabling the construction of a comprehensive reliability estimation model. Experimental results demonstrate that the proposed algorithm achieves an accuracy of 98.28 %, a recall of 99.61 %, and an F1-score of 98 %, and its estimated results align closely with actual outcomes after training on only 100 samples, indicating excellent convergence speed and strong reliability.
Authors
- Ning Pang (ORCID: https://orcid.org/0009-0006-9650-3033)
- Baifeng Ning (ORCID: https://orcid.org/0000-0002-2911-7857)
- Bin Chen
- Shuqing Yang
- Shaodong Zhao
Institutions
- Shenzhen Stock Exchange (CN)
- China Southern Power Grid (China) (CN)
Publication Details
- Journal
- Egyptian Informatics Journal
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1016/j.eij.2026.101086
- Primary Topic
- Blockchain Technology Applications and Security
- Type
- article
- Field-Weighted Citation Impact
- 0.00