Energy monitoring and smart grid applications have rapidly developed into a multi-billion dollar market. The continued growth and utility of monitoring technologies is predicated upon the ability to economically extract actionable information from acquired data streams. One of the largest roadblocks to effective analytics arises from the disparities of scale inherent in all aspects of data collection and processing. Managing these multifaceted dynamic range issues is crucial to the success of load monitoring and smart grid technology. This thesis presents NilmDB, a comprehensive framework for energy monitoring applications. The NilmDB management system is a network-enabled database that supports efficient storage, retrieval, and processing of vast, timestamped data sets. It allows a flexible and powerful separation between on-site, high-bandwidth processing operations and off-site, low-bandwidth control and visualization. Specific analysis can be performed as data is acquired, or retroactively as needed, using short filter scripts written in Python and transferred to the monitor. The NilmDB framework is used to implement a spectral envelope preprocessor, an integral part of many non-intrusive load monitoring workflows that extracts relevant harmonic information and provides significant data reduction. A robust approach to spectral envelope calculation is presented using a 4-parameter sinusoid fit. A new physically-windowed sensor architecture for improving the dynamic range of non-intrusive data acquisition is also presented and demonstrated. The hardware architecture utilizes digital techniques and physical cancellation to track a large-scale main signal while maintaining the ability to capture small-scale variations.
展开▼
