HVDC technology is increasingly important for long distance bulk power transmission, but existing protection relaying techniques for such a system are subject to limitations. This paper presents a novel Artificial Neural Network (ANN) based on an algorithm for fault detection, location and classification in VSC-HVDC systems. Taking advantage of the ability of ANNs to identify and classify patterns, the proposed algorithm is able to detect and correctly classify a fault occurring at either the rectifier substation on the DC line or at the inverter substation. Therefore, such a scheme can be used as a decision support tool or as a backup protection. Only local signals are used at the rectifier substation and no communication link is necessary, thus improving the system's protection reliability and reducing the overall cost of the hardware implementation. A detailed VSC-HVDC system is described and used to simulate a number of fault scenarios in the system. Using the resulting fault waveforms, a comprehensive decision support scheme is developed and described, paying particular attention to the signal processing chain and design of the specific ANNs for each relaying task. Finally, a detailed analysis of the influence of key fault parameters on the limits of the algorithm's performance is carried out.