Intelligent protective relaying for the series compensated line with high penetration of wind energy sources

• Protection issues of TCSC compensated line connected with large scale DFIG based wind farm. • A novel protection algorithm based on fast discrete S-transform assisted Back propagation neural network is proposed which uses three phase current signal for fault detection and classification. • This proposed method sensitivity is also tested in a New England IEEE 39-bus bigger system connected with large scale wind farm. • Use of the back propagation neural network based technique for detection and classification of fault in complex TCSC compensated transmission line connected to wind farm is presented here for the first time. • This back propagation neural network based algorithm is very sensitive for major fault cases like close in fault, CT saturation, and high resistance far end fault attached with wind farm. The bulk amount of power generated from the present-day large-scale DFIG-installed wind farms are preferably transmitted to utility grid through series compensated transmission lines. Currently, TCSC compensation is more attractive compared to fixed series compensation due to its numerous technical advantages. However, the nonlinear relationship of the output power verses wind speed and the different operating modes of the DFIG and TCSC cause rapid variation in the line impedance during both normal as well as fault conditions. Consequently, the widely used fixed impedance-based distance relays when used for protection of such lines find limitation. In this paper, a fast discrete S-transform feature-assisted back propagation neural network technique is proposed using the relay end current measurements for effective detection and classification of faults in such crucial transmission lines. The efficacy of the scheme is evaluated on numerous fault and non-fault cases simulated through MATLAB/Simulink on different standard test systems under varying system operating conditions. The results clearly show the superiority of the proposed method in comparision to the existing approaches in terms of its low computational burden, fast fault detection time (< 10 ms) and accuracies (= 100 %) and fast fault classification time (< 10 ms) and accuracies (99.99 %).