Power signature analysis

Is disaggregation the holy grail of energy efficiency? The case of electricity

“…Multiple studies attempt to extend the types of identifiable loads to lower-power, more complex devices found in modern homes and offices (e.g., computer, copier) or industrial buildings (e.g., variable speed fans) (Laughman et al, 2003;Lee, 2003;Shaw et al, 1998;Lee et al, 2005). In Shaw et al (1998), current waveforms were collected at a sampling rate of several kHz and used to compute in-phase and quadrature spectral envelopes (time-dependent, averaged Fourier coefficients) that characterize harmonic behavior.…”

“…Transient signal characteristics are highly dependent on the operational principle of the device and could provide more reliable, near real-time identification capability given that data is available that is sampled at high enough frequency (Laughman et al, 2003). Variations in the spectral envelopes (timevarying average Fourier coefficients) of current (Leeb et al, 1995;Shaw et al, 1998) or voltage (Laughman et al, 2003) waveforms are used to define signatures that are matched against prototypes ("exemplars") obtained for each appliance in a previous calibration phase. The match is typically performed by finding appropriate shifting and scaling parameters that minimize a least-square error criterion between events and exemplars.…”

Compressive Sampling for Non-Intrusive Appliance Load Monitoring (NALM) using Current Waveforms

“…The most well developed NALM technique utilizes the signature given by the real and reactive power of the appliance during its steadystate operation (i.e., excluding a short duration of the offto-on transient state) [1]. To resolve the ambiguity when two appliances have the same power signatures, more advanced techniques are proposed which exploit other signatures, including the steady-state current waveform [5], the transient-state current waveform [3], etc.…”

“…The current signature based approaches rely on the high frequency components to make the appliance more unique/distinguishable, which typically require a higher sampling rate (e.g., 40 kHz in [3]) than the power signature based approaches (e.g., 2-10 Hz in [1]). Compressive sampling (CS) is a method of acquiring and reconstructing sparse signals [2].…”

Power Signal Recognition with High Order Moment Features for Non-Intrusive Load Monitoring

“…[3] . 구체적으로, 유효 전력과 무효 전력을 쓰는 방법 [4] , 고조파를 (harmonics) 쓰는 방법 [6] , 푸리에 (Fourier) 계수와 전력의 변화량을 쓰 는 방법 [7] , 그리고 스위치 전환 때 생기는 광대역 잡음 을 쓰는 방법 [8] 처럼 나타낼 수 있다. 여기서,…”