Research On LED Harmonic Analysis And Control Scheme Based On Neural Network

With the widespread use of new power electronics in the distribution network,the power quality problems caused by the harmonics injected into the network are becoming increasingly prominent.However,the current research on harmonic management is mainly focused on new energy sources and large industrial equipment,while the power quality problems due to lighting equipment have attracted little attention.Although the capacity of a single lighting device is small,the lighting power consumption accounts for approximately20%of global total power consumption,and because LED(Lighting Emitting Diode)lamps are energy efficient and green,many governments have introduced directives to replace energy inefficient incandescent lamps with LEDs or CFLs(Compact Fluorescent Lamps).Therefore,the power quality problems caused by the large use of LEDs and other new lighting equipment cannot be ignored.At present,the research on harmonic analysis and management of LEDs is not yet complete:more in-depth detection and analysis of LED harmonic emission characteristics are needed,a more accurate LED total harmonic distortion(THD)prediction model needs to be established,harmonic detection algorithms with better dynamic response speed and steady-state accuracy and active power filters(APF)with reasonable harmonic compensation are needed.This article mainly takes LED lights as the research object by designing experiments and combined with artificial intelligence algorithms and simulation models to conduct a more in-depth study for analysis and governance plan of LED harmonics.1.Testing of LED lamps with different driver circuits by building a voltage regulation circuit.The Fluke 435 power quality analyser and illuminance meter were used to collect data on the harmonics and illuminance values of LED lamps during voltage changes.The experimental data shows that LEDs generate large harmonic currents and there is a strong correlation with the type of drive circuit.Through the study of the third harmonic current phase,it is found that combining different drive types of LEDs can reduce harmonic distortion.2.Use the data collected in the experiment to train the THD prediction model of LED lamps and based on neural network.A THD prediction method based on an improved Ada Boost algorithm is proposed.In this method,a Generalized Regression Neural Network(GRNN)model is established,and its parameters are optimized by Mind Evolution Algorithm(MEA)to improve the search ability of GRNN.On this basis,the Ada Boost algorithm is utilized to integrate multiple MEA-GRNN individuals to form a strong predictor,which can improve the model’s generalization ability of the neural network.To avoid the integration failure caused by improper selection of threshold value,a sigmoid adaptive factor is added to improve the accuracy of Ada Boost algorithm.Finally,the Ada-MEA-GRNN model is trained and simulated with the LED harmonic data collected by the experiment.From the simulation results it can be found that the accuracy of the proposed method is better than BP and GRNN,which can reach 95.48%.Meanwhile,even if the input dimension is reduced,the error is still small.3.Aiming at the problems of traditional detection algorithms that are prone to spectrum leakage and poor real-time performance,a harmonic detection method based on clustering and fourier basis PSO-Elman is proposed.By resetting the excitation function,data are integrated by clustering algorithm and particle swarm optimization algorithm(PSO)is combined with the Elman neural network which has good dynamic performance.Firstly,the harmonics are sampled and clustered,the data of the same type after clustering is used as the training data of the network,and then,utilizing the global optimization ability of particle swarm optimization(PSO),the optimal weights and thresholds are given to neural networks.Finally,the trained model is applied to predict the harmonic parameters.Simulation results show that,the fourier basis PSO-Elman network trained with 80,000 sets of k-means clustering integrated data has higher identification accuracy than the Elman neural network alone,harmonic current amplitude and phase estimation are accurate,which verifies the reliability of the detection method.4.In terms of active filter design,the APF topology suitable for different scenarios is classified and sorted out,and the key technologies of APF are explained.Starting from the topology,APF parameter design,harmonic detection method and current tracking control method,analyze and give the current problems and solutions.Finally,a simulation model of a parallel three-phase two-level LCL active power filter based on the i_p-i_q method is built,which better realizes harmonic compensation.In summary,this paper investigates the key technologies of LED harmonic analysis and treatment scheme,such as harmonic characteristic analysis,THD prediction method,harmonic detection method and active power filter design.It can provide reference for LED harmonic suppression,lamp selection,improvement of power quality evaluation standard system,and research of harmonic compensation equipment.