A Highly Efficient Compression Algorithm for Power Quality Disturbance Data Using Sparse Decomposition and Hybrid Entropy Encoding

The power quality disturbance (PQD) data are useful information for the power grid. Generally, different PQD waveforms correspond to different faults in the power grid, so the data can be used to distinguish the types of PQD signals. For example, the transient oscillation signal is caused by heavy load switching, the voltage notch originates from phase commutation of power electronic devices, voltage sag or interruption occurs because of short circuit faults, and the impulse signal is due to the thunder striking the lines, etc. In addition, the PQD waveforms can be used to locate the disturbance of the distribution network. However, with the scale of the power network increasing, PQD events are inevitably becoming more frequent than ever before. If there are many power grid monitoring devices in a certain distribution grid network, the bandwidth of Ethernet will be tremendous to transmit and receive the uncompressed PQD data, and the disk space for storage will also be huge, which leads to high expense on Ethernet and disk. Therefore, a highly efficient compression algorithm for PQD data calls for much attention and has become more urgent than ever before. The principle of the algorithm can be described as follow. First, this paper uses joint dictionary based sparse decomposition algorithm to separate the transient and steady-state components in PQD, secondly, the transient components are compressed by wavelet analysis, Huffman coding and run-length coding algorithms, for the steady-state component, i.e. the fundamental and harmonic components, the values that are greater than the threshold is reserved, and the compression is completed. To evaluate the proposed method, two indicators, compression ratio (CR) and percentage of root-mean-square difference (PRD) are utilized. Three kind of PQD signals, including impulse signal, sag with transient oscillation, and harmonic with decaying amplitude are generated in Matlab and compressed. The CR and PRD of impulse signal of the proposed algorithm is 27.64 and 1.19%, which are better than that of the competing methods. Similarly, the CR and PRD of sag with transient oscillation of the proposed algorithm is 31.12 and 1.88%, and the CR and PRD of harmonic with decaying amplitude of the proposed algorithm is 7.98 and 2.02%, showing better compression results. What’s more, a real-life sag signal measured from Power System Dynamic Simulation Laboratory, which is affiliated with Huazhong University of Science and Technology, are compressed and recovered. The CR are as high as 50, while the PRD is less than 2%, which are better than the results of sparse decomposition. The following conclusions can be drawn from the simulation and experiment results: (1) Compared with the competing methods such as “fast Fourier transform+wavelet transform”, “sparse decomposition”, and one-class support vector machine, the CR in the proposed method is much higher than that of the compared methods without PRD declining significantly. Therefore, it is appropriate to apply the proposed method to the real-life power quality signals. (2) the experiment performed in the Dynamic power system simulation laboratory results show that the CR is higher than the competing methods, which showed better application prospect. © 2023 Chinese Machine Press. All rights reserved.