The setting values of thresholds for fault feature parameters are critical in all kinds of protection schemes. When the detected feature parameter value exceeds the setting value

the protection will trip. However

the setting value based conventional protection schemes sometimes cannot satisfy the protection requirements of neutral ineffectively earthed power systems (NIEPS) due to wide variations in operating conditions and the complexities of fault cases. In this paper

a novel single phase grounding fault protection scheme without threshold setting is proposed. The fault detection is achieved based on operating states rather than setting values. A fuzzy c-means algorithm is used to divide the operating state of the protected feeder into non-fault states and fault states. The cluster center of each state is then obtained by classifying the historical feature samples of the protected feeder extracted under various operating conditions into their corresponding states in a constructed multi-dimensional fault feature space. The distances between the detected feature samples and the cluster centers of the non-fault and the fault states are calculated. If the distance to the fault state is shorter than that to the non-fault state

a fault is detected. Otherwise

the feeder is considered normal. A PSCAD/EMTDC simulator is used to simulate a 35 kV NIEPS under various operating conditions

non-linear loads

and complex fault cases. Results show that the proposed single phase grounding fault protection scheme without threshold setting can protect the system correctly under all kinds of faults.