Time aggregation for privacy-protecting EMU based on model-distribution predictive control

Smart meter adoption in electricity networks introduces privacy risks for consumers due to the increased measurement frequency and granularity. In particular, consumer behaviour and lifestyle choices may be inferred from their metering data using various Non-Intrusive Load Monitoring techniques. To protect consumer privacy, energy storage controlled by a Model-Distribution Predictive Control scheme can be utilised to mask the actual energy consumption by measuring privacy loss through the mutual information between actual and grid-visible energy consumption. However, the poor scalability of this approach limits the size of the prediction horizon, which is important for both energy cost and privacy loss reduction. In this paper, we propose using time aggregation to increase the reach of the controller's prediction horizon, and describe how to correctly model the statistics in this setting. Results show that with the proposed time aggregation, information leakage and energy costs can be further reduced without increasing the controller's computational requirements.