Design and Analysis of Gearboxes for Wind Power Systems

Abstract

The effective and dependable functioning of wind turbines depends on the construction and assessment of gearboxes in wind power installations. In transferring rotational energy through the wind turbine rotors to the electric power source, gearboxes are essential, and their performance has a direct bearing on the total efficiency of energy conversion. Yet, gearbox malfunctions can result in a lot of lost productivity and expensive repairs. To achieve the best overall efficiency and dependability of wind power networks, early identification and prediction of gearbox defects is essential. In order to address this problem, we introduce in this paper a new enhanced harmony search optimization-based feed-forward neural network (EHSO-FNN) technique. First, 20800 cases total, with 2600 examples for each of the 8 health categories. These instances included typical and unusual fault circumstances with variable speeds and workloads. In this investigation, 2000 records from each sample were provided, recording important operational factors, including temperature, motion, and oil quality. By using min-max normalization to record the basic gearbox health details, this data is cleaned up and turned into useful features. By using MFCC to analyze the motion and Acoustic information collected by wind turbines, we are able to identify a group of specific characteristics that are highly effective in describing the state of the system. The most insightful and pertinent features from the retrieved MFCC feature set are then chosen using EHSO. At last, a FNN model based on the selected elements is created to carry out the fault prediction. The suggested method's performance is assessed using the metrics of accuracy (98.98%), precision (98.92%), recall (99%), f1-score (98.96%), RMSE (0.021), MAE (0.028), and MAPE (0.032). The experimental findings show that, when compared to other methods(1DCNN-PSO-SVM, LSTM,TSVR, WF-MMD-JDA,SVM, and SCADA-DBN), the suggested method obtains the best prediction performance.Early fault detection is made possible by the recommended way, which also enables preventive repairs and reduces downtime for wind turbine installations.