Nima Mousavi, Baris Aksanli, A. S. Akyurek, T. Simunic
{"title":"物联网中边缘设备的精度-资源权衡","authors":"Nima Mousavi, Baris Aksanli, A. S. Akyurek, T. Simunic","doi":"10.1109/PERCOMW.2017.7917627","DOIUrl":null,"url":null,"abstract":"Modern power grid has evolved from a passive network into an application of Internet of Things with numerous interconnected elements and users. In this environment, household users greatly benefit from a prediction algorithm that estimates their future power demand to help them control off-grid generation, battery storage, and power consumption. In particular, household power consumption prediction plays a pivotal role in optimal utilization of batteries used alongside photovoltaic generation, creating saving opportunities for users. Since edge devices in Internet of Things offer limited capabilities, the computational complexity and memory and energy consumption of the prediction algorithms are capped. In this paper we forecast 24-hour demand from power consumption, weather, and time data, using Support Vector Regression models, and compare it to state-of-the-art prediction methods such as Linear Regression and persistence. We use power consumption traces from real datasets and a Raspberry Pi 3 embedded computer as testbed to evaluate the resource-accuracy trade-off. Our study reveals that Support Vector Regression is able to achieve 21% less prediction error on average compared to Linear Regression, which translates into 16% more cost savings for users when using residential batteries with photovoltaic generation.","PeriodicalId":319638,"journal":{"name":"2017 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2017-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Accuracy-resource tradeoff for edge devices in Internet of Things\",\"authors\":\"Nima Mousavi, Baris Aksanli, A. S. Akyurek, T. Simunic\",\"doi\":\"10.1109/PERCOMW.2017.7917627\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Modern power grid has evolved from a passive network into an application of Internet of Things with numerous interconnected elements and users. In this environment, household users greatly benefit from a prediction algorithm that estimates their future power demand to help them control off-grid generation, battery storage, and power consumption. In particular, household power consumption prediction plays a pivotal role in optimal utilization of batteries used alongside photovoltaic generation, creating saving opportunities for users. Since edge devices in Internet of Things offer limited capabilities, the computational complexity and memory and energy consumption of the prediction algorithms are capped. In this paper we forecast 24-hour demand from power consumption, weather, and time data, using Support Vector Regression models, and compare it to state-of-the-art prediction methods such as Linear Regression and persistence. We use power consumption traces from real datasets and a Raspberry Pi 3 embedded computer as testbed to evaluate the resource-accuracy trade-off. Our study reveals that Support Vector Regression is able to achieve 21% less prediction error on average compared to Linear Regression, which translates into 16% more cost savings for users when using residential batteries with photovoltaic generation.\",\"PeriodicalId\":319638,\"journal\":{\"name\":\"2017 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)\",\"volume\":\"10 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2017-03-13\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2017 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/PERCOMW.2017.7917627\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2017 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/PERCOMW.2017.7917627","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Accuracy-resource tradeoff for edge devices in Internet of Things
Modern power grid has evolved from a passive network into an application of Internet of Things with numerous interconnected elements and users. In this environment, household users greatly benefit from a prediction algorithm that estimates their future power demand to help them control off-grid generation, battery storage, and power consumption. In particular, household power consumption prediction plays a pivotal role in optimal utilization of batteries used alongside photovoltaic generation, creating saving opportunities for users. Since edge devices in Internet of Things offer limited capabilities, the computational complexity and memory and energy consumption of the prediction algorithms are capped. In this paper we forecast 24-hour demand from power consumption, weather, and time data, using Support Vector Regression models, and compare it to state-of-the-art prediction methods such as Linear Regression and persistence. We use power consumption traces from real datasets and a Raspberry Pi 3 embedded computer as testbed to evaluate the resource-accuracy trade-off. Our study reveals that Support Vector Regression is able to achieve 21% less prediction error on average compared to Linear Regression, which translates into 16% more cost savings for users when using residential batteries with photovoltaic generation.