EMT-Based Machine Learning Model for Fault Ride-Through Assessment in Type IV Offshore Wind Turbine Generators

The increasing penetration of converter-based generation in the power system is driving the need for more reliable operation in the power system. In recent blackouts and other large power systems events, the proper use of models in asserting grid compliance and stability played a key role. During the design of new products, Electromagnetic Transient (EMT) models are developed and validated against full-scale wind turbine prototypes for different fault ride- though scenarios. Initially, the paper briefly describes the EMT model used, developed in a commercial EMT software tool, that incorporates the actual converter control source code as a .dll as well as the fault ride-through (FRT) logic. The validation of EMT models for type IV Wind Turbine Generators (WTGs) during both the design and operation phases are also shown. Differently than in design, additional challenges in grid compliance verification during operation are often present, especially the short duration of real power system events and the challenges of performing model validation in a large fleet of WTGs and WPPs, including plant model unavailability, changes in software parameters and versions, etc. To address these challenges and speed up assessment during operation, the paper proposes a novel approach involving the feature extraction of relevant information during FRTs and an EMT-based Machine Learning (EMT-ML) model for playback of FRT characteristics, enabling grid compliance assessment. The EMT-ML model consists of a hybrid approach using long-short term memory (LSTM) and recurrent neural network (RNN) and was trained against 6000 simulated scenarios. The results are compared to real power system faults in two offshore WTGs in a WPP, validating the EMT-based ML model. The study concludes that the proposed EMT-ML method can be used in preliminary assessments of a large fleet of WTGs and potentially enhance WPP reliability in terms of grid compliance and stability.