Research On Harmonic And Interharmonic Detection Method With Neural Network

With the development of modern industry,rectifiers, frequency converters devices, electric furnace and a variety of power electronic devices influx into grid. These loads'features, such as nonlinear, unbalance and impacting, generate lots of harmonics and interharmonics in power. Also cause great harms on the power system and power users. In the meantime, power users demand high power quality increasingly. These show that it is significant to detect harmonic and interharmonics accuratly.This paper introduces the generation, harm and depression of harmonic and interharmonics firstly. Then analysis and compares these harmonic and interharmonics detection methods used at home and abroad commonly. Compare these methods synthetically from their calculation, accuracy, reliability, implementation, adaptive capacity, range and several other aspects.Based on those, this paper puts forward the neural network detection method of harmonics and interharmonics. Described specifically as follows:â‘ Using the BP neural network to analysis the sampled signals, obtain the high-precision of fundamental frequency.â‘¡Using the principles and methods of analysis fundamental frequency with BP neural network to detect fundamental frequency in integer harmonics high accuracy, then achieve the integer harmonics detection with the traditional linear neural network. In order to more close to actual electrical signals, joining the even number harmonics and random noise also. Compare the FFT algorithms with rectangular window and hanning window.â‘¢Using the FFT algorithm with hanning window to obtain the number of interharmonics and frequency with low precision.Set the pretreatment number of interharmonics as the number of neurons in the linear neural network with adjustable parameters activation function.To the pretreatment frequency as the basis to set the initial value of the iterative interharmonic number in the linear neural network with adjustable parameters activation function .â‘£Use the linear neural network with adjustable parameters activation function, that is the frequency, amplitude and phase of interharmonics as weights participating in learning and adjusting. Achieve high-precision detection of interharmonics. And analyze the detection in random noise environment .Compare the FFT algorithms with rectangular window and hanning window.In order to validate the proposed method to detect harmonics and interharmonics using neural network, program in the Matlab software environment and simulate. These simulation results show that the proposed methods which analysis harmonics and interharmonics in power system have higher detection accuracy and adaptive ability. Theoretical analysis and simulation result show the proposed methods'feasibility and effectiveness. When detect integer harmonics, the advantages of even harmonics detection are more obvious, especially the phase detection of even harmonics. When detect interharmonics, the advantages of interharmonic which are closed even harmonics detection are more obvious, especially the frequency detection of interharmonics which are closed even harmonics.Although only train and test 1 to 11 harmonics and interharmonics in grid signals. But these principles and methods are also fit for the grid signal including higher harmonics and interharmonics.