{"title":"Non-intrusive Load Monitoring in MVDC Shipboard Power Systems using Wavelet-Convolutional Neural Networks","authors":"Soroush Senemmar, J. Zhang","doi":"10.1109/TPEC54980.2022.9750745","DOIUrl":null,"url":null,"abstract":"This paper develops a non-intrusive load monitoring (NILM) method in future shipboard power systems (SPS) using discrete wavelet transform-based convolutional neural networks (CNN). We have applied the proposed NILM method to a two-zone medium voltage direct current (MVDC) SPS, with multiple appliances in each zone such as pulsed load, radar load, motor load, and hotel load. The input to the proposed NILM model only includes the current signal of generators, which will be first processed by a discrete wavelet transform, to form a coefficient matrix that represents the status of all the appliances in each zone. Then, a CNN model is adopted to monitor the load in real time by solving a multi-class classification problem. Results show that the proposed wavelet-based CNN model for NILM could: (i) determine the status of all appliances with an overall accuracy of more than 97%, and (ii) monitor specific pulsed loads with an accuracy of more than 98%.","PeriodicalId":185211,"journal":{"name":"2022 IEEE Texas Power and Energy Conference (TPEC)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE Texas Power and Energy Conference (TPEC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/TPEC54980.2022.9750745","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
This paper develops a non-intrusive load monitoring (NILM) method in future shipboard power systems (SPS) using discrete wavelet transform-based convolutional neural networks (CNN). We have applied the proposed NILM method to a two-zone medium voltage direct current (MVDC) SPS, with multiple appliances in each zone such as pulsed load, radar load, motor load, and hotel load. The input to the proposed NILM model only includes the current signal of generators, which will be first processed by a discrete wavelet transform, to form a coefficient matrix that represents the status of all the appliances in each zone. Then, a CNN model is adopted to monitor the load in real time by solving a multi-class classification problem. Results show that the proposed wavelet-based CNN model for NILM could: (i) determine the status of all appliances with an overall accuracy of more than 97%, and (ii) monitor specific pulsed loads with an accuracy of more than 98%.