Divisional fault diagnosis of large-scale power systems based on radial basis function neural network and fuzzy integral

• A method for divisional fault diagnosis of large-scale power systems is proposed.
• An overlapping network division method is proposed to divide a given power system.
• The FRA can faster construct better-performing RBF neural networks.
• Fuzzy integral can integrate historical experience and current state information.
• The proposed method has strong fault tolerance and high diagnostic accuracy.

This paper proposes an effective method for fault diagnosis of large-scale power systems based on radial basis function (RBF) neural network (NN) and fuzzy integral. It aims at effectively diagnosing the tie lines which connect different adjacent sub-networks in the context of divisional fault diagnosis. First, an overlapping network division method is proposed to divide a large-scale power system into a desired number of eligible sub-networks. Then, for each sub-network, a local RBF NN diagnostic module which is constructed by an exhaustive search-assisted forward recursive algorithm is allocated. Finally, a Choquet fuzzy integral fusion module is constructed for any pair of connected sub-networks. When a fault occurs, local RBF NN diagnostic modules will be selectively triggered according to local alarm information. If it involves a tie line, the corresponding Choquet fuzzy integral fusion module will be triggered to fuse the diagnostic outputs derived from the adjacent sub-networks which are connected by the tie line. Case studies with a 14-bus power system are presented to evaluate the feasibility and efficiency of the proposed method under various complex fault scenarios. The diagnostic results demonstrate that this proposed method is efficient in identifying faults within local sub-networks as well as those on the tie lines with strong fault tolerance and high diagnostic accuracy.