A Self-Improved Optimizer-Based CNN for Wind Turbine Fault Detection

Abstract

In comparison to other alternative energy sources, wind power is more affordable and environmentally friendly, making it one of the most significant energy sources in the world. It is vital to monitor the condition of each wind turbine in the farm and recognize the various states of alert since difficulties with the operation as well as maintenance of wind farms considerably contribute to the rise in their overall expenses. The Supervisory Control and Data Acquisition (SCADA) data-based continuous observation of wind turbine conditions is the most widely used existing strategy to detect the fault early by preventing the wind turbine from reaching a shutdown stage. Several parameters irrelevant to the faults are saved in the SCADA system while the wind turbine is operating. To increase the efficacy of wind turbine fault diagnostics, optimally selected SCADA data parameters are required for fault prediction. Hence, this paper introduces an optimized Convolutional Neural Network (CNN)-based wind turbine fault identification method. For more precise detection, a Self-Improved Slime Mould Algorithm (SI-SMA) is used for the optimal selection of SCADA parameters as well as weight optimization of CNN. The proposed SI-SMA method is an enhanced form of the standard Slime Mould Algorithm (SMA). Eventually, an error analysis and a stability analysis are carried out to check the overall effectiveness of the suggested approach. In particular, the root mean square error (RMSE) of the implemented algorithm is lower, and it is 0.69%, 1.58%, 0.81% and 1.71% better than the existing FF, GWO, WOA and SMA models.