Identification of unknown modes for air-conditioning based on hybrid clustering algorithm

Fault detection of Air Conditioning (AC) system is one of the most important problems in building energy-saving and safety. Unfortunately, the identification of unknown modes is still a difficult issue in fault detection. Expectation-maximization (EM) algorithm has been applied widely in modes identification of AC system. However, the traditional EM algorithm is much more complicated in iteration procedure and is more sensitive to the initial parameters, more importantly, the efficiency and the accuracy decreases with the rising of the data volume. A hybrid cluster algorithm is proposed in this manuscript with the intention of overcoming the drawbacks of traditional EM algorithm above. The simulation and the experiment show that the method can solve the identification problem of unknown multi-modes for air-conditioning system, besides the result turns out to be more efficient and more accurate than the traditional EM algorithm.