Sensor-driven analysis of retrofit options for low energy buildings**

低能耗建筑改造方案的传感器驱动分析**

• 批准号: 536485-2018
• 负责人: Evins, Ralph
• 金额: $ 1.82万
• 依托单位: University of Victoria
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=650358
• 关键词: Sensor; driven; analysis; retrofit; options

Energy assessment and providing retrofit options of existing buildings is currently costly and time consuming. This project aims to use the internet of things and big data to provide a cheaper way to find optimal retrofit options for residential buildings. A web of cheap sensors based on internet of things technology will be tested by monitoring environmental conditions in a 1870s era residential building. The data obtained from the sensors is stored in a database that will be analysed by machine learning algorithms trained using the available sensor data. The goal is that, when sufficient data is obtained, predictions can be made about optimal retrofit options for the monitored building. To validate this newly developed pathway traditional physics-based models will also be used. These provide a well-understood analysis method for existing buildings, but take significant time and expertise to develop. If these models are not properly calibrated, the error in energy consumption predictions may be in the order of 100%. The machine learning methods proposed require large amounts of sensor data but less time and expertise to develop. Furthermore, they provide accurate predictions of building performance based on available historical data. When these methods are fully developed, simple sensor data from residential buildings can be used to provide tailored retrofit options as a fast and inexpensive alternative to conventional physics-based modelling methods. This will provide quick retrofit solutions for the existing building stock. Improving building efficiency will help Canada meet its emissions goals and prevent further catastrophic climate change impacts.

能源评估和提供现有建筑物的改造方案目前是昂贵和耗时的。该项目旨在利用物联网和大数据提供一种更便宜的方式来寻找住宅建筑的最佳改造方案。一个基于物联网技术的廉价传感器网络将通过监测19世纪70年代住宅楼的环境条件进行测试。从传感器获得的数据存储在数据库中，该数据库将通过使用可用传感器数据训练的机器学习算法进行分析。目标是，当获得足够的数据时，可以预测被监测建筑物的最佳改造方案。为了验证这一新开发的途径，还将使用传统的基于物理学的模型。这些为现有建筑物提供了一种易于理解的分析方法，但需要大量的时间和专业知识来开发。如果这些模型没有被适当地校准，则能量消耗预测的误差可能在100%的量级。提出的机器学习方法需要大量的传感器数据，但开发时间和专业知识较少。此外，它们还可以根据现有的历史数据对建筑性能进行准确预测。当这些方法得到充分发展时，来自住宅建筑的简单传感器数据可用于提供量身定制的改造方案，作为传统基于物理的建模方法的快速和廉价的替代方案。这将为现有建筑提供快速改造解决方案。提高建筑效率将有助于加拿大实现其排放目标，并防止进一步的灾难性气候变化影响。