Research On Power Quality Disturbance Recognition Method Based On High Time-frequency Generalized S-Transform And LGWO-SVM

With the rapid development of smart grid and the extensive use of various impact,nonlinear loads and power electronic devices,power quality problems are becoming more and more serious.Accurate extraction and clear and effective characteristics of power quality disturbance signals and completion of disturbance signal identification are the basis and premise of solving power quality problems.It is of great significance to improve the quality of electricity consumption and maintain the safe and stable operation of power system.At present,time-frequency analysis methods such as wavelet transform and S-transform are mainly used to obtain characteristic parameters at home and abroad,and pattern recognition methods such as neural network are adopted.Because the time-frequency resolution of the standard S-transform cannot be optimized at the same time,the precision of the time-frequency characteristic of the disturbance signal is affected.Focusing on the feature parameter extraction of S-transform in the process of power quality detection,this paper studies the power quality disturbance detection and classification method of high time-frequency generalized S-transform and LGWO-SVM.Firstly,to solve the problem that standard S-transform is used to detect the non-optimal time-frequency characteristics of power quality disturbance signals,a high time-frequency generalized S-transform detection method is proposed.First of all,two groups of Gaussian window function regulating factorsλ₁（29）1、p₁（29）1 andλ₂（27）1、p₂（27）1 are introduced.By adjusting the parameters,the mode-time-frequency matrix with high time and frequency resolution is obtained respectively.Then,the maximum,minimum,mean and standard deviation of the time amplitude curve of the former fundamental frequency are extracted.And the three peak values of the maximum frequency amplitude envelope curve of the latter and their corresponding frequency values.The time-domain and frequency-domain features are combined to form the optimal combination features,which serve as the input of the subsequent classifier.Secondly,in order to solve the problem of selecting penalty factor C and kernel function parameter?of support vector machine（SVM）,the improved gray Wolf optimization algorithm（LGWO）based on Levy flight was used to optimize the parameters of support vector machine,and the LGWO-SVM classifier was constructed.With its strong global search capability,Levy flight makes up for the defect of GWO algorithm,which is easy to fall into local optimum.The LGWO algorithm formed by the combination of the two algorithms has a high level of solving accuracy and convergence speed.Through the optimization of LGWO algorithm,the optimal combination parameters and classification model of SVM are obtained to further improve the recognition performance of power quality disturbance.Finally,the combined feature samples extracted after the high time-frequency generalized S-transform are input into the LGWO-SVM classifier for training and testing,and the disturbance signal recognition is completed.The experimental results show that the proposed method can give consideration to both time and frequency domain resolutions,and has high recognition accuracy and strong anti-interference ability.Compared with the traditional S-transform time-frequency analysis method,the average recognition rate is 99.66%in the case of no noise,about 5%higher than S-transform,and the average recognition rate is 98.75%in the 20d B high-noise environment.