Predictive Model of Fiber Composite Overhead Ground Wire Galloping Based on Improved KVSD and CS-BP Method

The optical fiber composite overhead ground wire (OPGW) holds significant importance in electric power transmission systems. However, external factors like weather can lead to abnormal waving of the OPGW, characterized by high amplitude and low frequency. This waving can result in tripping, damage, or even cable fracture. Hence, timely identification of OPGW waving signals and assessing the associated security risks is crucial to ensure uninterrupted power transmission. Considering that waving monitoring data often contains random noise from the external environment, this study proposes an OPGW waving signal denoising method based on an enhanced KSVD algorithm fusion. To address the limitations of the BP neural network model, the Cuckoo Search Algorithm is employed to optimize the initial weights and thresholds of neurons, establishing an optimized BP neural network model. This optimization improves efficiency and helps overcome the problem of local optimal solutions. Experimental evaluation using a measured sample model confirms that the proposed model exhibits significantly smaller errors compared to traditional BP neural networks and generalized regression neural networks.