Scheduling the energy consumption of residential loads, the impact of missing data

We consider the problem of dispatching domestic power loads in presence of renewable energy sources. Renewables are largely and quickly fluctuating according to weather conditions, thus the imbalance between production and consumption may cause overvoltage, and overload of grid components. One of the technical measures for controlling these unbalances is the smart management of available flexibility in the grid. In order to let the distribution system operators face this challenge, we developed, in the context of the Optiflex project, a solution aimed to actively control domestic heat pumps and electric heaters. The main challenge of the project is that we want to use the already installed Smart Metering base as measurement, communication, and switching infrastructure, without the need for any additional hardware, containing investments for Smart Grids deployments. The commercial smart metering infrastructure we have available in the project is capable of collecting samples at up to a data point per minute. The approach consists of four main steps: power metering and non-intrusive load monitoring, photovoltaic power prediction, power demand estimation of the controllable loads, and load actuation scheduling. In this contribution, we discuss the practical implications of such a low sampling rate for the purposes of the project, we will showcase some examples from real-world pilot programs where we implemented the solution and we will discuss the robustness of adopted solutions in case of data-holes in collected data. We observed that missing data from the monitoring infrastructure can have a significant impact on non-intrusive load monitoring tasks, anyway the overall effect on the system performance is mitigated by the inherent robustness of the demand estimation component