A method of fault detection in thermal storage tank and results of applying to data which are generated by simulation and are obtained from an actual thermal storage system are presented. In the fault detection, a physical model is used for a water thermal storage tank of multi-connected complete mixing type, and unknown parameters of the physical model are identified from measured data. Faults can be detected by comparing value of identified parameters with value of the identified parameters under normal condition. In this report, simulation study was conducted in which several faults in storage tank, such as insulation damage of tank wall or abnormal water level in tank, were tried to detect. Furthermore, an analysis of identified parameters of the model is presented for measured data in an actual system.