The intelligent fault diagnosis of main circulation pumps is crucial for ensuring their safe and stable operation. However, limited research has been conducted on this topic, and applying existing fault diagnosis methods designed for other equipment may not yield optimal results when directly used for main circulation pump fault diagnosis. To address this issue, we propose a novel ensemble fault diagnosis model for the main circulation pumps of converter valves in voltage source converter-based high voltage direct current transmission (VSG-HVDC) systems. The proposed model employs a set of base learners already able to achieve satisfying fault diagnosis performance and a weighting model based on deep reinforcement learning that synthesizes the outputs of these base learners and assigns different weights to obtain the final fault diagnosis results. The experimental results demonstrate that the proposed model outperforms alternative approaches, achieving an accuracy of 95.00% and an F1 score of 90.48%. Compared to the widely used long and short-term memory artificial neural network (LSTM), the proposed model exhibits improvements of 4.06% in accuracy and 7.85% in F1 score. Furthermore, it surpasses the latest existing ensemble model based on the improved sparrow algorithm, with enhancements of 1.56% in accuracy and 2.91% in F1 score. This work presents a data-driven tool with high accuracy for the fault diagnosis of main circulation pumps, which plays a critical role in maintaining the operational stability of VSG-HVDC systems and satisfying the unmanned requirements of offshore flexible platform cooling systems.
Keywords: VSG-HVDC; converter station; converter valve; ensemble learning; fault diagnosis; machine learning; main circulation bump.