Abstract:
This work proposes the application of artificial neural networks (ANNs) as a working principle of a transmission line protection relay. The proposed relay has the functions of fault classification and location, distance protection and neutral overcurrent. Each application is performed by an ANN multilayer perceptron (MLP). The methodology used in the training of protection functions allows the performance of well-defined characteristics such as quadrilateral and mho. The architecture of the networks was defined aiming at optimization and performance speed, making use of the smallest possible number of neurons. The training process of the ANNs was carried out through the Solver tool, from Microsoft Excel, with the help of several applications developed in the Visual Basic language. One of the programs developed in VBA is used in the evaluation of neural networks in the face of real disturbances of the electrical system using oscillography files. The five designed neural networks were implemented in hardware through an ESP32 microcontroller. All program code steps are detailed. The final part of the work analyzes the results of the prototype relay tests carried out in a real-time simulator RTDS. The practical tests performed prove that the proposed relay is fast and safe, confirming the feasibility of the proposal for the protection of transmission lines.