Non-Intrusive Signature Extraction for Major Residential Loads

Abstract

This paper presents a technique to extract load signatures non-intrusively by using the smart meter data. Load signature extraction is different from load activity identification. It is a new and important problem to solve for the applications of non-intrusive load monitoring (NILM). For a target appliance whose signatures are to be extracted, the proposed technique first selects the candidate events that are likely to be associated with the appliance by using generic signatures and an event filtration step. It then applies a clustering algorithm to identify the authentic events of this appliance. In the third step, the operation cycles of appliances are estimated using an association algorithm. Finally, the electric signatures are extracted from these operation cycles. The results can have various applications. One is to create signature databases for the NILM applications. Another is for load condition monitoring. Validation results based on the data collected from three actual houses and a laboratory experiment have shown that the proposed method is a promising solution to the problem of load signature collection.

A framework for classification of non-linear loads in smart grids using Artificial Neural Networks and Multi-Agent Systems

“…Among various techniques the ones using the classical approach such as least squares calculation [7] and those based on intelligent systems [8] are found in the literature. The idea of using less-invasive methods for tracking load behaviours [9] is of special interest because it avoids the inconvenient installation of indoor sensors. Each home appliance has a unique harmonic current signature and the better understanding of the characteristics can be provided by the evolving smart meters [10].…”

“…The load classification is performed from the data collected from smart meters installed outdoors; [25] Fridge, furnace, microwave, stove, oven, kettle, cloth dryer and washer Liang et al [26] LED, LCD and plasma TV, LCD monitor, set-top box, heater, portable fan, microwave oven, desktop and laptop computer, DVD player and cellphone He et al [27] Electric heat, furnace, heat pump, lighting, TV, monitor, projector, fan, desktop computer and printer Bouhouras et al [28] Air conditioner, coffee machine, hair dryer, heater, home theatre, electric iron, laptop, refrigerator, washing machine, halogen lights and led lights Wang and Zheng et al [29] Washing machine, fan, mixer, personal computer, TV, stereo, air conditioner, heater, refrigerator, cooker, microwave ovens and hair dryer Lin et al [30] Vacuum cleaner, electric boiler, microwave oven and hair dryer…”

Development of distinct load signatures for higher efficiency of NILM algorithms

“…instantaneous current) is used for the development of the load signatures and this is expected to make the proposed algorithm more simple and time efficient for large scale applications. In [28], the authors also utilize three features of each appliance, active and reactive power and total harmonic distortion (THD), but for the computation of the THD they utilize the harmonic orders up to the ninth which means that for some appliances there could be a lack of information regarding their harmonic content and in turn their signatures would not be as distinct as they could be. Finally in [29] a different approach for a NILM approach with EMF sensors is presented.…”

Comprehensive feature selection for appliance classification in NILM

“…Dong et al [21] use total harmonic distortion (THD) of current waveforms along with P and Q for load discrimination.…”