Lightweight Non-Intrusive Load Monitoring Employing Pruned Sequence-to-Point Learning

Abstract

I]bj A]qadq+ Gdqhadqsn Bt]x©gthsk Rbgnnk ne Bnlotsdq Rbhdmbd+ Tmhudqrhsx ne Khmbnkm+ TJ 06522842?rstcdmsr-khmbnkm-Wb-tj gbtWwWgthsk?khmbnkm-Wb-tj Lhmfitm Ygnmf Cdo]qsldms ne Bnlotshmf Rbhdmbd+ Tmhudqrhsx ne @adqcddm @adqcddm+ Tmhsdc Jhmfcnl lhmfitm-ygnmf?Wacm-Wb-tj Vdmodmf Kt]m Bnkkdfd ne Dkdbsqhb]k ]mc Hmenql]shnm Dmfhmddqhmf+ Sh]mihm Tmhudqrhsx Sh]mihm+ Bghm] vdmodmf-ktWm?sit-dct-bm :arsqWbs Mnm,hmsqtrhud kn]c lnmhsnqhmf &MHKL( hr sgd oqnbdrr hm vghbg ] gntrdgnkc-r sns]k onvdq bnmrtloshnm hr trdc sn cdsdqlhmd sgd onvdq bnmrtloshnm ne gntrdgnkc ]ookh]mbdr-Oqduhntr vnqj g]r rgnvm sg]s rdptdmbd,sn,onhms &rdp1onhms( kd]qmhmf hr nmd ne sgd lnrs oqnlhrhmf ldsgncr enq s]bjkhmf MHKL-Sghr oqnbdrr trdr ] rdptdmbd ne ]ffqdf]sd onvdq c]s] sn l]o ] s]qfds ]ookh]mbd-r onvdq bnmrtloshnm ]s sgd lhconhms ne sg]s vhmcnv ne onvdq c]s]-Gnvdudq+ lncdkr oqnctbdc trhmf sghr ldsgnc bnms]hm tov]qcr ne sghqsx lhk, khnm vdhfgsr+ ld]mhmf sg]s sgd lncdkr qdpthqd k]qfd unktldr ne qdrntqbdr sn odqenql chr]ffqdf]shnm-Sghr o]odq ]ccqdrrdr sghr oqnakdl ax oqtmhmf sgd vdhfgsr kd]qmdc ax rtbg ] lncdk+ vghbg qdrtksr hm ] khfgsvdhfgs MHKL ]kfnqhsgl enq sgd otq, onrd ne adhmf cdoknxdc nm lnahkd cduhbdr rtbg ]r rl]qs ld, sdqr-Sgd oqtmdc rdp1onhms kd]qmhmf ]kfnqhsgl v]r ]ookhdc sn sgd QDEHS c]s]+ dwodqhldms]kkx rgnvhmf sg]s sgd odqenq, l]mbd v]r qds]hmdc bnlo]qhmf sn sgd nqhfhm]k rdp1onhms kd]qmhmf vghkrs sgd mtladq ne vdhfgsr v]r qdctbdc ax 76$-Bncd9gssor9..fhsgta-bnl.I]bjA]qadq87.oqtmdc,mhkl ::K :neSdosr ' Bnlotshmf ldsgncnknfhdr • Rtodp, uhrdc kdXpmhmf ax pdfpdrrhnm: MdtqWk mdsvnqir-Cdwtnpcr Dmdqfx chr]ffqdf]shnm+ cddo kd]qmhmf :BL Pdedqdmbd EnqlWs9 I]bj A]qadq+ Gdqhadqsn Bt]x©gthsk+ Lhmfitm Ygnmf+ ]mc Vdmodmf Kt]m-

Review on Deep Neural Networks Applied to Low-Frequency NILM

“…That means that the DNN-NILM inference has to work on an embedded system, even though that can be quite challenging in terms of computational, storage, and energy resources. This direction has been investigated by [47][48][49]65,106]: Ref. [106] is to our knowledge the first to publish the implementation of DNN-NILM inference on an embedded device.…”

Review on Deep Neural Networks Applied to Low-Frequency NILM

“…That means that the DNN-NILM inference has to work on an embedded system even though that can be quite challenging in terms of computational, storage, and energy resources. This direction has been investigated by [132,[137][138][139]145]: [132] is to our knowledge the first to publish the implementation of DNN-NILM inference on an embedded device. Both [132] and later [145] used for that purpose a Raspberry Pi computer.…”

“…The authors report that "disaggregation accuracy deviates up to ≈9.4 % from original disaggregation model, but, on average, remains satisfactory". Both [138] and [139] investigate different pruning methods based on the network from [80]. Pruning methods aim at reducing neurons in the network that contribute little to the final output.…”

“…The final goal is to result with sparse networks that have lower storage and computational requirements but similar performance compared to the original networks. Both publications find that networks can be heavily pruned with only slight decrease in performance: [138] reports a reduction of the number of network weights by 87 % and [139] reports a 100-fold reduction in model size and a 25-fold reduction in inference times. [139] additionally investigates multi-task learning and vector decomposition as further paths towards efficient computations in embedded systems.…”

“…There remain therefore a multitude of research questions in this direction. From our perspective, an interesting question would be to see how best performing approaches (see section 4.1) could be adapted to embedded devices because the architectures of these approaches are more elaborate than the ones used by [132,[137][138][139].…”