Classifying Sensor Data Using Recurrent Neural Networks
(2020)Department of Automatic Control
- Abstract
- Nearly 40 percent of overall energy usage in the European Union is used by buildings and 85 percent of that is for heating and cooling them[4]. This massive amount of energy is thought to be able to be lowered by introducing smarter control of the building systems. Technology to do this is continuously
being developed and improved. The newest analysis system developed by Schneider Electric uses cloud technology to apply a centrally developed algorithm to buildings around the world. It is fed information from the physical buildings through a Building Management System, BMS. As there are no universal naming schemes for building sensors, the process of connecting an existing building to the service is done manually. With the massive amount... (More) - Nearly 40 percent of overall energy usage in the European Union is used by buildings and 85 percent of that is for heating and cooling them[4]. This massive amount of energy is thought to be able to be lowered by introducing smarter control of the building systems. Technology to do this is continuously
being developed and improved. The newest analysis system developed by Schneider Electric uses cloud technology to apply a centrally developed algorithm to buildings around the world. It is fed information from the physical buildings through a Building Management System, BMS. As there are no universal naming schemes for building sensors, the process of connecting an existing building to the service is done manually. With the massive amount of sensors in a single building, this process is tedious, time consuming and error prone. This thesis is a continuation of previous work [1] on the same topic. It aims to find a way to fully- or at least semi-automate the connecting process using Recurrent Neural Networks, RNN, to analyze time series data and sensor names.
The goal for this thesis was reaching over 90% accuracy on a simple data set and having a top 3 accuracy good enough to simplify the connection problem significantly on a more complex data set. This goal was achieved, although some concerns about the data set not accurately portraying the real world scenario remain. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9024718
- author
- Niles, Oscar
- supervisor
- organization
- year
- 2020
- type
- H3 - Professional qualifications (4 Years - )
- subject
- report number
- TFRT-6096
- other publication id
- 0280-5316
- language
- English
- id
- 9024718
- date added to LUP
- 2020-07-16 09:03:05
- date last changed
- 2020-07-16 09:03:05
@misc{9024718,
abstract = {{Nearly 40 percent of overall energy usage in the European Union is used by buildings and 85 percent of that is for heating and cooling them[4]. This massive amount of energy is thought to be able to be lowered by introducing smarter control of the building systems. Technology to do this is continuously
being developed and improved. The newest analysis system developed by Schneider Electric uses cloud technology to apply a centrally developed algorithm to buildings around the world. It is fed information from the physical buildings through a Building Management System, BMS. As there are no universal naming schemes for building sensors, the process of connecting an existing building to the service is done manually. With the massive amount of sensors in a single building, this process is tedious, time consuming and error prone. This thesis is a continuation of previous work [1] on the same topic. It aims to find a way to fully- or at least semi-automate the connecting process using Recurrent Neural Networks, RNN, to analyze time series data and sensor names.
The goal for this thesis was reaching over 90% accuracy on a simple data set and having a top 3 accuracy good enough to simplify the connection problem significantly on a more complex data set. This goal was achieved, although some concerns about the data set not accurately portraying the real world scenario remain.}},
author = {{Niles, Oscar}},
language = {{eng}},
note = {{Student Paper}},
title = {{Classifying Sensor Data Using Recurrent Neural Networks}},
year = {{2020}},
}