Data-Driven Decision-Making Method of Intelligent Supervision and Command Platform in Offshore Wind Power Operation and Maintenance

INTRODUCTION: Innovations in offshore wind farm operation and maintenance require intelligent monitoring platforms that can leverage high-resolution SCADA data to enhance predictive precision and operational effectiveness. With the use of deep learning algorithms, specifically LSTM, this work achieves improved forecasting and anomaly detection accuracy on a real wind turbine dataset recorded in Turkey in 2018. OBJECTIVES: The proposed approach involves extensive data preprocessing, including cleaning, synchronization, and normalization, followed by advanced feature extraction using signal processing transforms such as the Fast Fourier Transform and wavelet transforms. METHODS: Different predictive models, including Linear Regression, Random Forest Regression, Support Vector Regression, Gradient Boosting Machines, and LSTM, were trained and tested within a Python setting. The LSTM model achieved a remarkable improvement, with a Mean Absolute Error of 78.6 kW, compared to traditional machine learning methods such as RF Regression, SV Regression, and Gradient Boosting Machines. The enhanced accuracy results from the LSTM's ability to derive intricate temporal patterns and nonlinear relationships inherent in sequential turbine operational data. RESULTS: The results affirm the potential of deep learning approaches in reshaping offshore wind turbine management and highlight the importance of tailored temporal modeling for resolving the specific challenges of renewable energy systems. CONCLUSION: The system not only accurately predicts power production but also performs anomaly detection and optimises maintenance scheduling, resulting in enhanced reliability and energy production for offshore wind farms. By integrating these data-oriented approaches with an intelligent command and supervision system, the strategy facilitates proactive decision-making and real-time operation control.