An Improved Metaheuristic Method-Based Neural Network for Predicting Wind Turbine Power

Abstract

Wind power production has advanced rapidly in recent years as a supreme renewable energy source that is safe, reliable, pollution-free, and simple to integrate into the power grid. Furthermore, employing data acquisition and supervisory control to predict wind turbine power may not result in the best governing approach, as non-calibrated data may be generated due to sensor degradation. Hence, to solve the adverse impact, it’s planned to incorporate the wind turbine Supervisory Control and Data Acquisition (SCADA) data into the improved optimization based NN, resulting in the development of a highly accurate prediction model. Thus, SCADA data of the wind turbine such as active power, bearing shaft, gear box shearing, gear box oil temperature, generator rpm, generator windings 1 and 2, hub temperature, reactive power, rotor rpm and wind speed are fed as input features to the predictive model. In order to achieve better prediction, the NN will be trained by using a new Coefficient Factor Updated Coyote Optimization Algorithm (CFU-COA), which is the conceptual advancement of traditional Coyote Optimization Algorithm (COA). Finally, the supremacy of the presented approach is proved with respect to varied error measures.