Load Disaggregation Through Particle Filtering of Harmonic Features

The need of electrical energy saving policies is becoming increasingly important nowadays, specially after the latest electricity market tensions in Europe. The energy consumption impacts directly in our electricity bills, thus optimization strategies are required to decrease its negative consequences on the economy and encouraging the consumption of renewable energies and reducing the carbon footprint. For this reason, non-intrusive load monitoring is gaining more attention as a modern strategy to assist in controlling electricity consumption. Several techniques have been proposed using basic information from current or voltage waveforms. However, the use of harmonic features is not widespread due to the need for a device capable of measuring power signals at a high sampling rate and for the high complexity of the disaggregation problem. The use of new high-end IoT devices like openZmeter, can help on this task. In this paper, several sets of harmonic features (current, active power, and custom combination) used along with a particle filter algorithm are compared. The results show that increasing the number of odd harmonics increases the accuracy of the result provided by the algorithm.