Portal de Periódicos da CAPES

Artificial intelligence-based fault detection and diagnosis methods for building energy systems: Advantages, challenges and the future

2019; Elsevier BV; Volume: 109; Linguagem: Inglês

10.1016/j.rser.2019.04.021

1879-0690

Yang Zhao, Tingting Li, Xuejun Zhang, Chaobo Zhang,

Building Energy and Comfort Optimization

Artificial intelligence has showed powerful capacity in detecting and diagnosing faults of building energy systems. This paper aims at making a comprehensive literature review of artificial intelligence-based fault detection and diagnosis (FDD) methods for building energy systems in the past twenty years from 1998 to 2018, summarizing the strengths and shortcomings of the existing artificial intelligence-based methods, and revealing the most important research tasks in the future. Challenges in developing FDD methods for building energy systems are discussed firstly. Then, a comprehensive literature review is made. All methods are classified into two categories, i.e. data driven-based and knowledge driven-based. The data driven-based methods are abundant, including the classification-based, unsupervised learning-based and regression-based. They showed powerful capacity in learning patterns from training data. But, they need a large amount of training data, and have problems in reliability and robustness. The knowledge driven-based methods show powerful capacity in simulating the diagnostic thinking of experts. But, they rely on expert knowledge heavily. It is concluded that new artificial intelligence-based methodologies are needed to be able to combine the advantages of both kinds of methods in the future.