A Novel Transient Disturbance Waveform Compression Method in Distribution Networks Based on Dictionary Learning

The distribution network with a large-scale integration of distributed renewable energy sources exhibits more complex dynamic behavior. The widespread deployment of high-resolution waveform monitoring in distribution networks is hindered by the massive data transmission and storage requirements. To mitigate this communication bottleneck, this paper proposes a novel transient disturbance waveform compression method in distribution networks based on dictionary learning. A new framework for disturbance waveform compression on terminal side and reconstruction on master station side has been proposed to outline the offline learning and online decomposition steps. Then a determination method of the start and end times of transient disturbance waveforms based on the relative energy acceleration operator (REAO) has been proposed for disturbance waveform segmentation, which clarifies the disturbance processes that need to be monitored. To improve the decomposition and reconstruction efficiency, an overcomplete dictionary design method that considers feature redundancy within the dictionary based on clustering pruning optimized K SVD (CPK-SVD) has been proposed. The redundant atoms in the dictionary are pruned based on Gaussian mixture model (GMM) clustering method after feature extraction. Simulation and field recorded waveform data tests show that the proposed method can achieve a compression ratio of over 10 while ensuring measurement accuracy.