An Approach To Perform Comprehensive Wind Turbine Performance Analytics By Means Of Machine Learning (Sci-Kit Learn)

Abstract

With the present day’s emphasis on sustainable and secure energy supply, wind power sector is growing rapidly all over the world. Along with the rapid expansion of the wind turbine sector, the wind turbine industry is also growing. Understanding the power response of these systems to the variations in wind velocity is essential for the optimal selection and efficient management of these turbines. This is defined by the power analysis and prediction of wind turbines output. Previous approaches such as mechanical aspects of a turbine are used to work with wind turbine power output. However, these techniques are not capable of analyzing big dataset with hefty number of parameters. In this regard, Machine Learning is a handy tool to analyze a dataset and design a model that can perform prediction related to turbines parameters. Python is one of the Open Source convenient tools to implement ML tasks. In this thesis work, we proposed Scikit Learn based Machine Learning models, which is based on Python for the power analysis. Three different machine learning methods such as XGBoost, LightGBM, Catboost were used for the modeling. The comprehensive dataset based on Western Wind Firms are being used to design the analysis model. The accuracies of these models are validated by estimating the error between the model output and the field observations from these turbines from the comprehensive dataset.