The wind energy industry is growing rapidly as a response to the global call to combat climate change by rapidly decarbonising the energy sector. However, with the increasing deployment of wind turbines in offshore locations, they are exposed to harsh environmental conditions which can lead to leading edge erosion on the blades. This erosion can result in up to 25% loss of Annual Energy Production (AEP) and cost operators up to €130 million annually.
To address this issue, researchers have developed a tool using Machine Learning technology to predict wind turbine blade leading edge erosion. The study involved reviewing existing numerical and analytical models on leading edge erosion to find one capable of producing the necessary erosion data to train an ML model. An analytical surface fatigue model was used to create an erosion dataset.
The dataset was then used to train, test, and validate three ML algorithms: a Support Vector Machine, an Artificial Neural Network, and a Random Forest algorithm. While the SVM produced an inferior result, the ANN and RF algorithms showed promise with optimal accuracies of 80.41% and 98.87%. Based on these results, a final ML model was developed using the RF algorithm and used to predict the erosion damage rate for an unfamiliar set of input data. A final optimal accuracy of 99.42% was recorded, showing that the ML model could successfully predict the erosion damage rate of a wind turbine blade leading edge.
This study highlights the potential for Machine Learning technology to address a significant challenge faced by the wind energy industry, leading edge erosion on wind turbine blades. With an optimal accuracy of 99.42%, this ML model can be used by operators to predict erosion damage rate and take proactive measures to minimize AEP loss and costs. The use of machine learning technology in wind turbine blade erosion prediction holds promise for the wind energy industry.
