Advanced devices and algorithms for energy disaggregation in buildings

能量分解（也称为非侵入式负载监测）是信号处理和模式识别技术的组合，用于从总能耗信号中估计各个电器的能耗。它是通过识别聚合信号中的区别特征并将其分解为组成部分来实现的。将能源消耗数据分解到电器的层面，是通过提高国民意识、利用电力需求预测工具、智能自动控制电器等来提高能源效率的契机。本研究课题旨在推进能源分解领域的发展，以便在实际生活中有效地确定各个电力负载的电力消耗。这将通过建立在申请人的团队在霍尔效应电流传感器设备和弱监督机器学习算法的最新发展来实现。具体来说，我们将1）使用来自低精度电流传感器的弱标记数据设计用于住宅应用的新的能量解聚算法，2）分析解聚算法对传感器信息的准确性和数量的敏感性，3）设计下一代超低功耗霍尔效应电流传感器; 4）为商业和工业应用设计具有最佳数量的低精度传感器的新能量分解算法。这些进步将提高现有技术的准确性，可扩展性和适应性。该研究计划将在工业和学术界需求很高的领域培养四名HQP，获得信号处理，机器学习和微电子学方面的重要技能。研究结果也可以成为新合作的起点，因为能源分解正在引起工业和公用事业的兴趣，最近在这一领域的重要投资证实了这一点。加拿大自然资源部确定，“加拿大建筑部门有责任负责任地使用我们的能源资源，并采取行动，作为一种机制，将加强和丰富我们的经济，为子孙后代。“这项研究计划是朝着这个方向迈出的重要一步，通过开发新技术来更有效地监测能源消耗，并最终为智能建筑提供领先的能源管理技术。

Energy disaggregation (also referred as nonintrusive load monitoring) is a combination of signal processing and pattern recognition techniques to estimate the energy consumption of individual appliances from the total energy consumption signal. It is achieved by identifying discriminating features in the aggregated signal and decomposing it into its constituent parts. Disaggregation of the energy consumption data down to the level of appliances has been largely identified as an opportunity for enhanced energy efficiency through citizen awareness, energy demand prediction tools for utilities and smart automatic control of appliances, just to name a few.This research program seeks to advance the field of energy disaggregation to efficiently determine the power consumption of individual electrical loads in real-life scenarios. This will be achieved by building on the applicant's team latest developments in Hall-effect current sensor devices and weakly-supervised machine learning algorithms. Specifically, we will 1) design new energy disaggregation algorithms for residential applications using weakly labeled data from low-precision current sensors, 2) analyze the sensitivity of the disaggregation algorithms to the accuracy and quantity of sensor information, 3) design the next generation of ultra-low-power Hall-effect current sensors and 4) design new energy disaggregation algorithms for commercial and industrial applications with an optimal number of low-precision sensors. These advances will increase the accuracy, the scalability and the adaptability of existing techniques.This research program will train four HQP in domains which are in high demand in industry and academia, gaining important skills in signal processing, machine learning and microelectronics. The research results could also be the starting point of new collaborations, as energy disaggregation is gaining interest from industry and utilities, as confirmed by recent important investments in this field.Natural Resources Canada established that "the Canadian buildings sector has a duty to use our energy resources responsibly and take up the call to action as a mechanism that will strengthen and enrich our economy for future generations." This research program is an important step in that direction, through the development of novel technologies to monitor energy consumption more efficiently, and eventually, enabling leading-edge energy management technologies for smart buildings.