Detection And Identification Research Of Power Quality Disturbances In Power System

With the rapid development of the large-scale AC/DC hybird in smart grid,the grid-connected operation of new energy power generation,and the operations of nonlinear unbalanced loads,the high-quality of power quality has been destroyed,and the problem of power quality disturbance in power system has become more and more serious.However,grave consequences such as abnormal operation of power equipment,equipment overheating and misoperation of relay protection device,will be caused by the power quality disturbance with its non-stationary,mutability and short duration.Moreover,it reduces the service life of compensation capacitor and increase metering error.Thus,it is necessary to detect and identify power quality disturbance to evaluate and improve power quality,which has great significance to ensure the safe supply of electric energy and the optimal allocation of power resource.The major works of the thesis are as follows:(1)Research challenge and key technology of power quality disturbance.The cause and harm of power quality disturbance in power system were analyzed and summarized.The mathematical model,classification and characteristic of power quality disturbance were described in detail according to the relevant standards of power quality at home and abroad.In the first chapter of this thesis,the methods of detecting and identifying power quality disturbance and its applications in the field of power quality analysis were discussed and compared.(2)Power quality disturbance detection and identification based on modulus maximum and energy difference.Power quality disturbance in power system was analyzed by using wavelet transform.Modulus maxima of wavelet transfor,which was used to localize mutation peak of power quality disturbance and achieve the detection of its start-stop time,was extracted through the multi-resolution decomposition.Then,wavelet energy distribution in different decomposition scales,which was used to achieve the identification of power quality disturbance,was constructed according to the differences between high and low frequency wavelet coefficients.The results of simulation and case study show that the proposed method can accurately detect start-stop time of power quality disturbance and identify its type.However,there are some disadvantages such as large relative error in its detection and large amount of calculation in identification.(3)Power quality disturbance start-stop time detection based on lifting Db4 complex wavelet transform.The lifting complex wavelet scheme was designed by using Euclidean decomposition algorithm.The adaptive lifting factor of Db4 complex wavelet was obtained to construct the decomposition and reconstruction model.Start-stop time of power quality disturbance was detected through the difference between the lifting phase of the disturbance signal and the standard phase of the normal voltage signal.Compared with the wavelet and complex wavelet,the simulation results show that the proposed method can better detect the disturbances' start-stop time with relatively high speed and accuracy.The efficiency of the algorithm is improved by 59.52% and 34.55%,and the relative error is reduced by6.67% and 3.14% respectively.The promising prospect in engineering applications has been verified through case study.(4)Power quality disturbance type identification based on improved BP neural network.Traditional BP algorithm was improved by combining the increasing momentum method with the self-adaption learning rate method.The network learning rate and the connection weights of each layer are adjusted automatically.Improved BP neural network can effectively overcome the shortcomings of traditional BP neural network which is easy to fall into local minimum and has slow convergence speed.Compared with the wavelet energy classification method,the simulation results show that the proposed method can identify power quality disturbance in power system with high discrimination ratio and speed.The average recognition rate is up to 97.8% and the convergence rate of BP neural network is increased 3.1times.Some practical reference value has been verified through case study.