Automated Estimation of Fault Locations in Bipolar HVDC Systems Employing Artificial Neural Networks

With a continuous escalation in demand of power, the Indian Electrical system is in constant demand for long transmission lines to fulfil its requirement due to extremely distributed demand and generation location. Advanced HVDC system is one such possibility that finds its utility, especially during long-distance transmission. Such electrical transmission systems are prone to short circuit faults, which subsequently leads to a large current, which will eventually harm or damage the system’s equipment. Thus, the system requires a quick restoration in order to re-establish power transmission and assure system safety. Hence, this paper presents a model, which can precisely assess the location of the fault in a Bipolar HVDC system. The model provides accurate results which is also collectively optimal. A Bi-polar transmission line which is 814 km long and operates at 700 kV, 1500 MW is developed on PSCAD/EMTDC software based on CIGRÉ benchmark guidelines. The designed model is further simulated for short circuit fault with fault ON resistance of 0.01 Ω and fault OFF resistance of 1.0 x 10
⁶
Ω with varying fault location along transmission line at an interval of 1 km. The acquired data is collected and processed for feature extraction. Data from both the ends of the transmission line is used for training and testing of deep neural network models. The evaluation of the proposed system has been done based on the mean squared error and accuracy of fault estimation. It is shown that the proposed system outperforms contemporary baseline approaches.