Level-Oriented Diagnosis for Indoor Climate Installations

Abstract

Indoor climate installations are widely used in modern office buildings. Statistical data show that the energy consumption in buildings has grown quickly in the past decades. A considerable part is due to inefficient and improper operations of installations. Therefore, Fault Detection and Diagnosis (FDD) for indoor climate installations has fallen increasingly into the research scope. A new approach called Level-Oriented Fault Detection and Diagnosis (LOFDD) is presented. It analyses the performance of the installations on different levels. 1. Global level on which the overall performance of energy consumption of the whole building is analysed. Fuzzy Neural Networks modelling has been adopted. By analysing the peak response from the model, the fault will be detected. When a fault is detected on this level, the next level is activated for further diagnosis. 2. Domain level on which the performance of rooms or zones are analysed. Physical model based method is discussed and applied. With the help of the characteristic parameters, the fault may be detected and diagnosed on this level or further diagnosis can be done on the next level. 3. Sub-domain level on which the detailed components are analysed. Several faults are introduced into the real installations. A new Pattern Match Method is presented and it shows good capability for fault detection and diagnosis. On this level the fault will eventually be diagnosed. Experiments are carried out to validate the method in this thesis.