Vibration Signal-Based Diagnosis of Wind Turbine Blade Conditions for Improving Energy Extraction Using Machine Learning Approach

Wind power capacity is rapidly expanding across the world. In many nations, however, wind energy profit margins are being reduced. As a result, many wind farm operators are looking for ways to save costs and reduce maintenance issues. This research provides a condition monitoring and predictive maintenance framework for wind turbines based on artificial intelligence. This paper aims to create a model that categorizes various blade defects using statistical attributes with acquired vibration signals. The fault classification uses machine learning approaches, including attribute extraction, selection, and classification. First, statistical characteristics or attributes are extracted from wind turbine quaver or vibration signals utilizing a data acquisition system, then feature selection is performed using a decision tree algorithm to choose the best attributes. Next, feature classification is performed with 15-fold cross-validations using different models of tree classifiers. Then, based on their accuracy percentage, the results of machine learning classifiers are compared to provide a good model of the turbine blade for the real-time monitoring system. The objective of this learning is to design a prototype that will work best for the fault classification of turbine blades with less computational time. The logistic model tree shows the best classification accuracy of 91.57 %, with 1.72 seconds of computation time.