Adaptive load signature coding for electrical appliance monitoring over low-bandwidth communication channels

Abstract

Collecting and analyzing power consumption data from electrical appliances is a key enabling element for grid-related services, e.g., load forecasting or anomaly detection. Device-level sensors (smart plugs) have found widespread use to collect such data. However, they prevalently report an electrical appliance's power consumption at a rate of one reading per second in order to limit the resultant communication traffic. With mains voltage frequencies of 50/60 Hz, undersampling and the consequent loss of spectral information result from the use of such reporting rates. Moreover, as most smart plugs only report real power consumption values, important supplementary features (e.g., the phase shift between voltage and current or the magnitude of reactive power) are not available when using such devices. In this work we present a data processing system design that exploits the recurring nature of electrical current waveforms in order to facilitate the provision of data at a high resolution whilst keeping the corresponding data rate requirements low. Our design, called ALSCEAM, is applicable to voltage and current waveforms collected at high sampling rates, thus spectral components are implicitly included in collected traces. Instead of transferring raw readings to external processing services, however, local data processing routines are being employed to detect and eliminate redundancies. Thus, a high data fidelity is maintained while network traffic is reduced by more than 95% in many cases. All functionalities are implemented in a proof-of-concept system design and evaluated in practice.