A Continual Learning-based Framework for Developing A Single Wind Turbine Cybertwin Adaptively Serving Multiple Modeling Tasks

This paper proposes a generalized deep continual learning-based cybertwin (GDC) modeling framework for developing one wind turbine (WT) cybertwin serving multiple modeling tasks in the wind farm operations and maintenance (O&M). A generalized WT cybertwin modeling problem is first formulated. Fully connected deep neural networks (DNNs) are adopted as the backbone for developing the GDC. The online elastic weight consolidation (OEWC) method is incorporated to mitigate the catastrophic forgetting phenomenon among different modeling tasks. Computational experiments are conducted to validate the effectiveness of the proposed GDC framework based on the supervisory control and data acquisition (SCADA) data. Three classical tasks, the WT gearbox failure detection, WT blade breakage detection, and wind power prediction, studied in the wind farm O&M are considered in the experiment. Compared with considered benchmarking models, the proposed GDC can achieve high accuracies on new tasks while maintain high accuracies on previous tasks.