Research On Power Quality Detection Algorithm Based On Neural Network

With the continuous improvement of production efficiency and quality of life,the demand for power quality of electrical equipment is becoming higher and higher,and some problems of power quality have also attracted the common attention of a large number of research institutions and scholars.However,accurate detection of various power quality disturbances is an effective measure to improve power quality.Therefore,the detection and analysis of power quality disturbance signal have certain theoretical significance and practical value.BP neural network is a kind of multi-layer feedforward neural network with strong non-mapping ability and self-learning ability.This algorithm is used frequently in the field of harmonic detection and has high precision in harmonic detection.In this dissertation,based on BP neural network,the detection algorithms for harmonic/inter-harmonic and disturbance signals are discussed respectively by improving the problems in it.The dissertation based on the BP neural network,through the improvement of the problem,the detection algorithm researchs for harmonic,inter-harmonic and disturbance signals are discussed respectively.The following three points of research have been carried out and some achievements have been achieved.1.In the harmonic detection process of BP neural network,oscillation is not difficult to occur in the learning process of the network,and the accuracy of harmonic detection was affected by the selection of the initial value of the network as well as the number of intermediate nodes.A detection method conbined double adaptive neural network and with TLS-ESPRIT is proposed.On the basis of the traditional BP neural network,the momentum factor and learning rate are double adaptive,and the double adaptive BP neural network is used to improve the instability of the network and improve the convergence speed of the network.In order to select the initial value and the number of intermediate nodes accurately and improve the detection accuracy of the signal,the TLS-ESPRIT algorithm is introduced to get the initial value of frequency and the number of intermediate nodes accurately and quickly,determine the network structure,and lay the foundation for the precise detection of the signal by BP neural network.Compared with the traditional BP neural network,the improved dual adaptive BP neural network can optimize the harmonic frequency,greatly reduce the computing time,enhance the convergence,and have a certain anti-noise ability.2.When the dual-adaptive neural network processes inter-harmonic signals and complex signals mixed with harmonics and inter-harmonics under noise interference,the network falls into local minimum value,which leads to low precision of detection parameters.An improved harmonic detection algorithm based on double adaptive BP neural network is proposed.Compared with the traditional BP neural network,the adaptive ability of the dual adaptive BP neural network has been improved.However,the momentum factor and learning rate of the dual adaptive BP neural network only change at a limited fixed point without considering the continuous increase or decrease of errors,which leads to the lack of adaptive ability and the decline of precision in the harmonic detection process.So the network on the basis of further specialisation,introducing the ideas of "pointer in C language of BP neural network,to create a network error" pointer ",through the vector pointer "error" and momentum factor further double adaptive adjustment,continuous increase/decrease the error of considering it a moving point,further extend the value of the space,is advantageous to the adaptive adjustment of the network at a deeper level.Interharmonic,complex harmonic and experimental signals were used to simulate the detection method,and the simulation results were compared with those in the unimproved double-adaptive BP neural network.The results show that the improved harmonic detection method has better network adaptability,higher convergence speed and higher detection accuracy.3.To solve the problem that BP neural network could not perform transient signal perturbation location,a perturbation location method based on Hankel matrix singular value decomposition(SVD)was introduced,and a perturbation detection method based on improved dual adaptive neural network and Hankel matrix singular value decomposition was proposed.Firstly,the component signal with the most obvious mutation point was obtained after SVD decomposition,and the mutation information was extracted from the component signal,and the time when the mutation point changed the most obvious was taken as the start and end of the composite disturbance signal.Through the improved double-adaptive BP neural network algorithm,the sampling data before and after the initial point of disturbance were detected and analyzed,and the disturbance parameters were obtained.Finally,the simulation results of single/compound perturbation and perturbation experimental data signal show that the proposed algorithm can effectively locate the perturbation signal,and the detection accuracy of signal parameters is high.