Main bearing degradation prediction in wind turbines via quantum support vector machine regression

To mitigate energy production losses in wind energy sector, maintenance activities should be scheduled during periods of low wind. An alternative to help achieve this goal is by predicting the remaining useful life (RUL) of components, which indicates the estimated time a component can function before experiencing critical failure. In this work, a framework is applied to predict the main bearing degradation using real operational data, collect by the SCADA system, of three wind turbines. Quantum support vector machine (QSVM), a field that merges principles from quantum mechanics with the classical SVM model, is explored with six distinct feature maps. The methodology employs a strategy to overcome data scarcity, by partitioning data into training, validation, and test subsets. The results demonstrated that the angular QSVM overcame the classical SVM model, with R2 score of 0.8532 and RMSE of 0.0796, which highlights the potential of quantum computing to achieve accurate predictions.