Forecasting Anonymized Electricity Load Profiles

TL;DR

This paper investigates GDPR-driven privacy concerns for granular electric load profiles and evaluates microaggregation, focusing on MDAV, as a non-perturbative anonymization method. Using the Low Carbon London dataset and a broad set of forecasting models (baseline, statistical, ML, and neural networks), it shows that forecasting accuracy at an aggregated level is largely preserved after anonymization, while information loss and volatility decline with larger anonymization groups. The study identifies an optimal privacy-utility region (roughly mid-range $k$ values) where privacy benefits stabilize without sacrificing utility, and it discusses the practical impact for privacy-preserving data sharing in the energy sector. These findings support integrating privacy-preserving analytics into smart-meter data workflows under GDPR, with caveats about geographic/time limitations and the need for broader anonymization techniques and household-level evaluation.

Abstract

In the evolving landscape of data privacy, the anonymization of electric load profiles has become a critical issue, especially with the enforcement of the General Data Protection Regulation (GDPR) in Europe. These electric load profiles, which are essential datasets in the energy industry, are classified as personal behavioral data, necessitating stringent protective measures. This article explores the implications of this classification, the importance of data anonymization, and the potential of forecasting using microaggregated data. The findings underscore that effective anonymization techniques, such as microaggregation, do not compromise the performance of forecasting models under certain conditions (i.e., forecasting aggregated). In such an aggregated level, microaggregated data maintains high levels of utility, with minimal impact on forecasting accuracy. The implications for the energy sector are profound, suggesting that privacy-preserving data practices can be integrated into smart metering technology applications without hindering their effectiveness.

Figures (3)

• Figure 1: Visual representation of the different datasets under different anonymization levels.
• Figure 2: Overall performance results by $k$ anonymization level in terms of mean ans standard deviation.
• Figure 3: Volatility and information loss for each of the $k$ levels in logarithmic scale.