Calibration of parameters and predictive control strategy of a wind turbine for improvement of energy harvesting

Applied mathematical modeling of the various energy resources has contributed widely in their design and operation improvements. Among renewable energy sources, wind energy systems have gained a significant attention for their rapid and continuous growth, which resulted from their cost effectiveness and clean power generation. However, research effort on the industry and university levels is still needed to enhance the characteristics of existing wind turbines and allow larger amounts of their share. This thesis presents an accurate model of a wind turbine (TWT-1.65) system for time-based dynamic simulations. The parameters of the power coefficient have been identified, calibrated and verified using different meta-heuristic optimization techniques, which are the Wild Horse Optimizer (WHO), Whale Optimizer (WO), and Genetic Algorithm (GA) to enhance the model accuracy over the previously related studies. The new version of the parameters has resulted in a higher accuracy while being practically meaningful, with the best achieved MSE of 0.006. The system model has been then integrated with two predictive controllers which are the linear Model Predictive Control (MPC) and Neural Network Predictive Control (NNPC) to regulate the pitch angle trends for the target of maximum energy harvesting. With proper selection of the controllers’ parameters, off-line simulation studies have shown improved production trends of the power output with constrained and safe trajectory of the pitch angle, which can be translated to be an improve in the total average harvested power of 34% and 44.5% over the complete time window of 5760 min using NNPC and Linear MPC respectively.