Semi AI-based protection element for MMC-MTDC using local-measurements

Abstract

The multi-terminal HVDC system based on the modular multilevel converter (MMC-MTDC) is a promising technique for flexible power transmissions to multiple regions. As such a system is quite sensitive to DC faults, there is an acute need to propose a protection element that can trip the local DC circuit breaker (CB) within several milliseconds once there is an internal DC line fault. However, the existing main protection scheme faces a dilemma balancing selectivity and sensitivity. To solve this problem, a novel semi artificial-intelligence (AI) based protection element is proposed, including a start-up criterion and a fault-identification criterion. The start-up criterion is based on the propagation characteristics of the initial fault-induced surge. To enhance the real-time performance of the protection element, it will not trip the fault-identification process unless the fault is identified as a forward one. The fault-identification criterion is based on artificial intelligence (AI), and further determines whether the forward fault is internal, which only works if the start-up criterion trips. Simulation results indicate that the proposed protection element has satisfactory speed, sensitivity, and selectivity against internal DC faults and is quite secure under external fault conditions. The impact of disturbances, such as the white noise, abnormal samplings, etc., on the security of the proposed protection element is also discussed.