Proxy-Calibrated Obfuscation for Power Systems - Obfuscating Capacity Calculation for Swedish DSOs

In Sweden, numerous small, local Distribution System Operators (DSOs) with limited computational resources struggle to independently meet EU data privacy and transparency requirements. For instance, they must publish capacity maps showing available capacity at each network node, typically relying on third-party cloud services outside Sweden for these calculations. However, network data is transmitted unprotected, exposing it to theft that could enable attacks on Swedish infrastructure. This thesis therefore investigates network and electrical data obfuscation techniques for low-voltage distribution grids. Data obfuscation deliberately obscures sensitive information to protect it (and the underlying network) from malicious attacks while preserving utility, i.e., the ability to perform computations on obfuscated data yielding results close to those from the original network. This project particularly emphasizes Proxy-Calibrated Obfuscation (PCO), where utility is approximated using proxy metrics (such as Graph Centrality Measures (GCMs)) derived directly from network structure and topology, without requiring ground- truth capacity computations for validation. First, this work demonstrates the feasibility of obfuscation when the true capacity results are known (Capacity-Calibrated Obfuscation (CCO)), establishing a foundation for introducing proxy-based approximations in the subsequent PCO phase. Although the results do not yet fully validate PCO effectiveness, they reveal a promising approach for lightweight obfuscation, enabling Swedish DSOs to safely publish accurate capacity maps of their networks.