Study On Harmonic Detection Method Based On Improved Wavelet Neural Network For Power Systems

The rapid development of China's living standard and production pace makes its demand for electricity show an increasing trend year by year,while the stability and safety requirements of the power grid are also increasing day by day.The use of a large number of non-linear equipment in a power system makes the power system polluted by harmonics.Harmonics will cause additional energy losses in the power grid,make the lines in the equipment overheat,accelerate the aging of equipment and lines,and even cause fire.It is also possible to cause resonance in the power grid,and there is a risk of burning capacitors and inductances in the power grid.It may also cause the misoperation of relay protection and some automatic devices,and affect the stability of power.Therefore,it is important to study on the detection methods of harmonic signals for power systems.The results of literature research show that relevant research has provided a rich means for the detection of harmonics in the power grid,but there are still unresolved problems,mainly including:(1)The convergence speed of the wavelet neural network is low,and the accuracy is greatly affected by its parameters.(2)The existing harmonic detection methods have low accuracy and are sensitive to noise,especially when the environment is harsh or there is strong noise.As a result,the performance of harmonic detection methods is not satisfied for power systems.(3)The fluctuation of fundamental wave in power system affects the accuracy of harmonic detection.In order to solve above problems,the wavelet neural network is intorudced in the research to improve the performance of the harmonic detection for power system,and main contents and results are summarized as follows: of(1)The wavelet neural network is studied,including its learning algorithm,structure setting and optimization of initial parameters.The simulation results demonstrate that the improved wavelet neural network has higher detection accuracy and better convergence performance,which provides an available method for harmonic detection in power systems.(2)In view of the problem of pulse noise in power system,the medium filter is used for pre-processing.For the problem of fundamental wave fluctuation in power system,the GA-BP(Genetic-Algorithm-Back-Propagation)neural network method is used to detect the fundamental wave in power system,which is necessary inputs for the harmonic detection.These results provide clear signals and important information on the fundamental wave for the following harmonic detection.(3)In order to improve the performance and noise resistance of harmonic detection,the improved wavelet neural network method is applied to detect the harmonics and compared with other commonly used harmonic detection methods.The simulation results indicate that the improved wavelet neural network has better detection performance and computational efficiency in the harmonic detection of power system.Through the research work in this thesis,the proposed methods are available for the harmonic detection in power systems.Even though the environment is harsh and there is noise in the signal,the proposed method still has high detection accuracy and may be further extened and applied in the field of power systems.