Predicting Output Power of Hybrid Solar-Wind Energy system using ML based Neural Network Regression Model

Abstract: In order to maintain the power load, demand the hybrid solar-wind power systems are recently beingstudied. The proposed methodology establishes a machine learning (ML) based predictive framework toaccurately estimate the output power of a hybrid solar–wind energy system (HSWES) by integrating torqueparameters from the wind subsystem and solar load characteristics from the PV subsystem. Performance istested open Koggle Indian Solar-Wind data as well on synthetically generated wind system torque and solar PVsystem load datasets, The data for various features like wind torque, wind speed, solar irradiance, solar load,and ambient temperature are generated. It suggests using linear interpolation to fill missing values and apply z-score standardization to normalize data. The proposed approach trained and test the Neural network (NN) based regression model for achieving the high degree of production.The performance is investigated by trainingof multiple regression models such as linear regression, support vector regression (SVR), ensemble learning, andneural networks (NN). Comparative performance evaluation is carried out using metrics like RMSE and R2 it isfound thatNN model works best with validated R2value ranging from0.94358 to 0.95241 for low to high loadconditions. The Bayesian optimization is applied with SVR and ensemble learning methods, while for improvingthe NN prediction performance the Grid search optimization with increased epochs is used and the performanceof R2 is improved to 0.95696. The model can be tested in Realtime scenarios with actual true data.