An unsupervised learning algorithm for the classification of the protection device in the fault diagnosis system

Power protection devices achieve a rapid removal of the grid accident, but the numerous applications of the devices had brought data disaster for the fault diagnosis information system. It costs a great deal of efforts to ensure that the information uploaded by the protection devices correspond to the function of the devices. This pa per presents an unsupervised learning algorithm for the classification of the protection device to facilitate the fault diagnosis information system to locate accurately the event reports of every protection device. The algorithm classifies the protection devices without samples or with small number of samples according to the automation relaying settings. The classification of the protection devices is not just separating the devices according to the types of different companies, but also differentiates these devices with the same type from the same company but different functions. There are two innovations in the proposed unsupervised learning algorithm for the classification of the protection device. Firstly, the automation relaying settings can solve the classification of the protection device in essence and eliminate mistakes from the source. The mistakes are usually caused by the classification of the device´s name in manual mode. Secondly, addressing to the characteristics that the relaying settings of the protection devices have uncertain entries under different functions, the algorithm realizes the classification from massive devices.