An AI-based fault detection and classification method for hybrid parallel HVAC/HVDC overhead transmission lines.

• Presenting an artificial intelligence-based protection method for hybrid parallel HVAC/HVDC overhead transmission lines (HPOTLs) using only the sending-end data (not needing a telecommunication channel) • Using a Decision Tree (DT) algorithm and a fault energy index to select the best features for identification of faulty transmission line (HVAC/HVDC line), type of fault (AC/DC/intersystem fault), and faulty phase(s)/pole(s) • Accurate modeling of HPOTL considering electromagnetic coupling between the HVAC and HVDC lines • Low sampling frequency, high speed and high accuracy in fault detection and classification • Robustness of the proposed method to fault resistance, fault angle, transmission line length and noise In this paper, a novel method is presented to detect and classify the AC, DC, and AC/DC intersystem faults in hybrid parallel HVAC/HVDC overhead transmission lines (HPOTLs) on the same tower. In HPOTLs lines the presence of electromagnetic coupling between the AC and DC conductors induces components in the AC and DC lines which can affect the performance of their protection systems. To tackle these challenges, a protection method based on the decision tree (DT) algorithm and a fault energy index (FEI) is proposed for HPOTLs. The proposed method uses the sending-end voltages and currents of the AC and DC lines and some transformations to calculate the FEIs. The fault detection and classification algorithms and their criteria are derived by the DT algorithm. PSCAD and MATLAB software are used for HPOTL modeling, and voltage and current signals processing, respectively. Simulation results show that the proposed protection method can detect and classify different faults (AC and DC line faults and AC/DC intersystem faults) accurately, and is able to discriminate between transmission line internal and external faults. Simplicity, high speed and accuracy, robustness to the fault resistance and noise, and not needing a telecommunication channel are important features of the proposed method. [ABSTRACT FROM AUTHOR]