Testing Open-Source Libraries for Private Counts and Averages on Energy Metering Time Series

Open markets for renewable energy sources usually need demand forecasting and dynamic pricing. This requires from metering aggregators analytics capabilities on detailed consumption data. Smart grid's operators can access metering information without accidentally leaking someone's consumption patterns, if good implementations of differential privacy are adopted when publishing statistics. However, this emergent technology is complex and hard to implement correctly. This paper analyzes three differential privacy libraries used for computing differentially private averages and counts on synthetic energy metering data. Libraries are compared by their ability to preserve privacy and utility on time series for energy consumption. Results indicate a delicate trade-off between privacy and utility in energy metering time series, where increased utility leads to decreased privacy, eventually revealing visually distinguishable patterns.