Mean estimation of electricity consumption data based on PID and local differential privacy

In the collection and analysis of electricity consumption data, malicious attacks from untrusted third party are prone to incomplete protection of user privacy. However, the periodic and correlation of the electricity consumption data will be ignored with direct random perturbation under local differential privacy, causing the potential for leakage of user life details. Therefore, this paper proposed a mean estimation method for electricity consumption data based on proportion-integral-derivative (PID) and local differential privacy. Firstly, the data was divided into a smooth period and a peak period; Secondly, the classified data was sampled with PID updating the sampling period. Finally, the sampled data was randomly perturbed by the Piecewise Mechanism (PM) and aggregated for mean estimation. The theoretical analysis shows that the method satisfies the local differential privacy. The experimental results demonstrate that the mean estimation error can be controlled within 0.008. As the privacy budget increases, the error becomes smaller and can reach 0.0001 when the privacy budget is maximum with better data utility.