Compression Reconstruction And Disturbance Identification Of Power Quality Data Based On Dictionary Learning

With the construction of the world-class power grid project started, the rapid development of electric power system has brought a variety of power quality problems, which has aroused wide concern of researchers. As an important part of power quality problems, the compression reconstruction and disturbance identification of power quality data are the most significant research methods. At present, the research methods are generally limited by the Shannon sampling theorem. Compressed sensing theory, as a "big idea" in signal processing community, will bring new inspiration to the study of data signal processing. With the deepening of the research, the theory of compressed sensing is becoming increasingly superiority in all aspects.Compressed sensing theory mainly consists of three parts: sparse representation of signal,design of measurement matrix and selection of reconstruction algorithm. Among them, the sparse representation of the signal is the premise of theory and has a great relationship with the reconstruction accuracy of the data. Due to the particularity of the power quality data, the orthogonal basis of the common function is used to express the sparse representation and the optimal sparse representation can not be self-adaption. So this paper has carried on the deep research to this problem, which was combining compression perception with dictionary learning. Dictionary learning is used to obtain the dictionary of the power quality data, using dictionary to sparse representation of power quality data and obtaining the sparse representation of the best power quality data, in order to improve the accuracy of reconstruction precision and disturbance identification of power quality data.The main research contents of this paper are as follows:1. Based on the compressed sensing theory study, compressed sensing theory in signal sparse representation were in-depth study, which combined the characteristics of power quality data with the dictionary learning method. The method of dictionary learning is applied to the sparse representation of power quality data, which lays a solid theoretical foundation for further research.2. Through the study of sparse representation of the power quality by compressed sensing,power quality data compression and reconstruction method based on adaptive dictionary learning is proposed in this paper. This method breaks through the limitation of the traditional data compression, which can quickly and easily recover the original power quality data in the case of a small amount of samples.3. In this paper, a new method of power quality data disturbance recognition based on discriminative dictionary learning sparse representation is proposed, which can identify the sparse representation of various power quality disturbances. Compared with other identification methods, this method does not require feature extraction and recognition classifier, and discriminative dictionary learning is established to identified different types of power quality disturbance data.4. Inspired by the discriminative dictionary, the study further extends the discriminative dictionary, and proposes an algorithm for classifying the power quality disturbance classification based on the sub dictionary, and the non coherence constraint of sub dictionary was added to the cost function of discriminative dictionary learning, which can effectively reduce the coherence between sub dictionaries to obtain better classification results.