Non Intrusive Load Monitoring for Industrial Chiller Plant System - A Long Short Term Memory Approach

With the development of better computational powered metering infrastructure, there is an increasing interest in using machine learning methods for non-intrusive load monitoring. In this paper, variations of input-output data model and long-short term memory, a deep learning layer, are implemented and discussed for energy disaggregation of a chiller plant system of a commercial building for energy monitoring purposes. Then, error analysis is carried out for the multi-mode, multi-state cooling tower operation. The opportunities and challenges of towards energy monitoring are then briefly discussed. Unlike NILM residential buildings applications for on-off equipment status determination, NILM for industrial system focuses on estimating the power level of the disaggregated components due to high cost of sub-metering.