Bearing Fault Detection Method in Gravity Energy Storage System Based on Improved VMD Fusion-Optimized CNN

Against the backdrop of increasing global energy demand, efficient energy storage technologies are of significant importance for advancing a low-carbon economy. Gravity energy storage systems, as an advanced energy storage method, rely on the performance of key components such as bearings, which directly influence the system's reliability and efficiency. Therefore, it is necessary to monitor the bearing's condition to ensure the stable operation of the system. However, during the actual detection of bearing vibration signals, a considerable amount of noise may be present. This paper utilizes the Variational Mode Decomposition (VMD) for signal denoising and applies the Sparrow Search Algorithm (SSA) to optimize the decomposition parameters of VMD. By using these optimized decomposition parameters, the original vibration signals of the bearings are effectively separated. Finally, we employ K-Fold to optimize the hyperparameters of the Convolutional Neural Network (CNN) and utilize the optimized CNN for detecting bearing faults. This approach has shown certain improvements in enhancing the efficiency and accuracy of bearing fault detection.