<p>This paper presents an analytical power forecasting model for pitch-controlled wind turbines using real-time wind-farm monitoring data. This study highlights the significance of employing real-time monitoring data and advanced analytical techniques for short-term wind turbine power curve forecasting to optimize wind turbine performance. The model incorporates meteorological features (wind speed, wind direction, and temperature) and operational factors (blade pitch and yaw error) to assess performance efficiency and predict wind power output. A hybrid method utilizing Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and a locally weighted regression technique called Locally Estimated Scatterplot Smoothing (LOESS) were employed to address anomalies and improve the data quality. The robustness of the models was assessed using statistical performance checks, including root mean square error (RMSE), mean relative standard error (RSE), absolute error (MAE), and correlation coefficient (R). The variable importance index (Ii) was calculated to determine the most influential input features for wind-power prediction. The results indicated that wind speed and blade pitch angle were the most influential parameters, followed by yaw errors, wind direction, and temperature. The new analytical model, developed from sub-expression trees based on optimized mutations, demonstrates an outstanding correlation between the predicted and measured wind power with minimal errors.</p>