State Recognition of Wind Turbines Based on K-means and BPNN

In order to achieve the goal of “double carbon”, the development of wind power generation technology is essential. At the same time, with the increasing complexity of power grid, the real-time detection and accurate evaluation of the state of wind turbines and other power equipment are becoming increasingly important. In recent years, the development of big data technology and the improvement of power equipment data monitoring technology makes possible the application of big data technology in power equipment state recognition. Compared with the conventional methods, the above-mentioned methods are independent of accurate empirical thresholds or quantitative models, and have better adaptability to the rapid increase and variability of data. Thus, this paper applies the unsupervised (K-means) and supervised (BPNN) machine learning methods to state recognition of wind turbines, while exploring the variation of accuracy and computational efficiency after the application of dimensionality reduction methods. The results show that both machine learning methods are effective in state recognition of wind turbines, while the dimensionality reduction method can effectively improve the computational efficiency with limited accuracy loss.