A CNN-LSTM-Attention model for risk prediction of offshore wind turbines

Abstract

Offshore wind turbines are exposed to complex environmental loads such as wind, waves and ocean currents, which can lead to progressive structural degradation and increased failure risk over time. To achieve accurate risk prediction and intelligent structural health management, this paper proposes a hybrid deep learning model—CNN-LSTM-Attention—that integrates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and an Attention mechanism. The model leverages CNN to extract local spatial features, LSTM to capture temporal dynamics, and Attention to enhance the learning of critical patterns, thereby improving overall prediction accuracy. The proposed model is trained and validated on datasets representing four critical limit states of offshore wind turbine structures. Comparative experiments with traditional LSTM and CNN-LSTM models demonstrate that the CNN-LSTM-Attention model achieves superior performance across multiple evaluation metrics, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). Furthermore, the predicted probability distributions of structural failure reveal clear risk trends under different limit conditions, offering meaningful insights for risk-based maintenance planning and operational decision-making.