Power Quality Disturbance Recognition Based On Feature Combination And Optimized Support Vector Machine

With the rapid development of modern science and technology,industrial intelligence and power grid interconnection technology,the demand for power quality is more stringent.At the same time,the power quality problems are increasingly prominent in some fields such as:aerospace,automobile,electronic device manufacturing industry,etc.Therefore,the power quality problem has become the focus of scientific research and industrial field.To improve and solve the problem of power quality,it is necessary to take reasonable and targeted methods to guarantee the excellent power quality supplied by the system,and the key basis of its realization lies in in-depth study of the influence factors of each power quality disturbance,accurate extraction of disturbance information,and finally accurate identification of power quality disturbance types.This paper analyzes and summarizes the causes and harms of various power quality problems.Combined with the research status and standards of power quality,the identification of power quality disturbances is mainly focused on the feature extraction and classification identification:For the power quality disturbance types discussed in this paper,the time-frequency detection methods,namely wavelet and S-transform are used to analyze the disturbance signals.Firstly,the characteristics of wavelet multi-resolution are used to analyze each disturbance signal.By constructing the relative wavelet energy of each decomposition layer,the distribution of the relative wavelet energy of each decomposition layer is obtained,and the relative wavelet energy which differ greatly from each disturbance signal on the decomposition layer are extracted as the feature quantities.Then,combined with the characteristics of each disturbance signal,S-transform can be used to analyze the feature information of the specific frequency and specific moment of the signal,extract the amplitude and frequency information related to each disturbance signal with great difference,and combine them with the feature quantity extracted by wavelet transform to form the combined features.The support vector machine(SVM)is used as the classifier for identifying power quality disturbances,the feature quantities extracted by wavelet and S-transform are input to SVM to identify the types of power quality disturbances.Aiming at the problem of parameter selection of the classifier,an improved particle swarm optimization(IPSO)algorithm is proposed to dynamically adjust the inertia weight and learning factor,So that the particles have reasonable motion inertia and strong learning ability in the search space,so as to accurately obtain the optimal parameter combination and improve the accuracy of disturbances identification.The accuracy of power quality disturbance recognition based on feature combination and optimization SVM proposed in this paper is verified by simulation.The results show that the recognition accuracy of features extracted by wavelet and S-transform is 2.4167%higher than that of features extracted by S-transform.By comparing the classification results of SVM,PSO optimized SVM and IPSO optimized SVM in power quality disturbances,IPSO optimized SVM can availably enhance the accuracy of disturbance recognition and has certain noise resistance.