Power carbon emission factor analysis algorithm based on lightweight differential privacy and association preserving

The ongoing informatization of power systems has heightened the significance of both carbon emission data sharing and privacy protection. With the advancement of smart grid technologies, the collection and analysis of critical data have become central technical priorities for achieving the "dual carbon" goals. Traditional centralized data processing methods are increasingly inadequate to meet operational demands and pose substantial risks to user privacy and business confidentiality. To address these challenges, this study develops a lightweight differential privacy (DP) framework integrated with a data compression algorithm based on sparse Fourier transform (SFT), which significantly reduces computational complexity and improves processing efficiency. By incorporating a dynamic privacy budget allocation (DPBA) mechanism, the framework introduces sensitivity-aware adaptive privacy protection levels. Furthermore, the DP-based correlation correction algorithm (DP-CCA) is innnovatively proposed, which effectively preserves the statistical correlations of carbon emission factors after noise injection. The proposed method is evaluated through simulation experiments. Compared with traditional DP methods, the proposed approach improves data processing speed by approximately 87.7% (equivalent to 12.3% of the time required by the conventional method). The accuracy, measured by association similarity, of carbon emission factor correlation analysis is improved by 42.2%, while memory usage is reduced by 72%. This research not only provides an optimized data sharing solution for the power industry but also, owing to its lightweight characteristics, demonstrates enhanced suitability for deployment in edge computing environments. Thus, it offers a crucial technical foundation for establishing privacy-preserving distributed carbon emission monitoring networks.