IoT intelligent agent based cloud management system by integrating machine learning algorithm for HVAC systems.

• A novel IoT intelligent agent based cloud management system was proposed to improve energy efficiency and safety in building energy systems. • The residuals of actual condition and predicted fault-free condition were calculated as input of FDD model to improve generalization performance. • A k-fold cross validation method was presented to especially testify the generalization capacity in different operation conditions. • The diagnostic accuracy was over 99% and 71% in source conditions and cross conditions respectively. • The proposed smart cloud management framework can be applied in general energy systems. Collection of extensive sensor data from HVAC chillers has facilitated the development of smart cloud management systems to increase the energy efficiency of buildings. However, limited by computing resource and algorithm performance, smart energy management technique and its fault diagnosis method in the literature suffer from low response speed and generalization capacity. To this end, a novel IoT intelligent agent based cloud management system for HVAC systems to improve operational efficiency and safety. The smart cloud management system integrates fundamental framework with machine learning algorithm for fault detection and diagnosis. After preprocessing the data collected by IoT agents, an algorithm is constructed to predict virtual sensor values based on fault-free conditions. The calculated residuals of the actual values and virtual values on both normal and faulty conditions are used as inputs to an extreme gradient boosting algorithm that predicts the fault level. The diagnosis results are compared with other methods such as support vector machine, multi-layer perceptron and random forest. The k-fold cross validation indicated that the proposed methodology can achieve superior overall generalization performance with 67.8%, 70.5% and 71.6% while that of the conventional method were 59.4%, 63.9% and 68.3%. This study will contribute to the practical applications of smart cloud management system in building energy systems. [ABSTRACT FROM AUTHOR]