Analyzing feature impacts in wind turbine power output modelling using quantile regression

The wind power industry is experiencing significant growth, establishing itself as a sustainable and environmentally friendly energy resource. A predictive model for wind turbine power output is essential to optimize operational costs. In practical scenarios, the technical specifications of turbines alone are inadequate to estimate power output under varying environmental conditions. This research conducts feature impact analysis to investigate the complex relationships of various meteorological variables. The traditional ordinary least squares regression model that is frequently used for predictive modelling cannot effectively capture the changing effects observed at different levels of the response variable's distribution. This limitation motivates the adoption of quantile regression, which offers the analysis of features' impact on response variables by estimating the model parameters at various quantiles. By conducting feature impact analysis across multiple quantiles, this study reveals various trends at different quantiles, which can optimize model parameters in wind turbine power output modelling.