The System Identification Of High-power LED Based On The Optimization Of BP Network

With the progress of the times and the development of science and technology,people’s need for energy is also increasing, together with the economic crisis, theworldwide energy shortage and environmental pollution problems. Develop rationally,reduce the waste, improve energy’s utilization ratio and find new energy savingtechnology has becomes the mainstream ideology of countries.Light emitting diode gained wide attention as the newest generation green lightinglight source because of the high light effect and long life, small volume, light weight, lowradiation, low power consumption and green energy-saving advantages. LED is asolid-state cold light source based on light-emitting principle. It has become the most idealgreen light source because of photoelectric conversion rate is ten times higher thantraditional lighting sources. Because of the restriction of semiconductor optoelectronictechnology makes the luminous efficiency of LED lamps is low. A large amount ofelectrical energy is converted to heat and can not be dispersed in time. The control of nodetemperature is an effective way to promote LED luminous efficiency.Based on the research on the LED lighting technology’s status quo and development,this paper introduces the principle of neural network identification and the advantages ofthe LED light source. In order to provide theoretical support for the subject of furtherresearch, it mainly states the LED’s thermal properties, the node temperature and itsmeasuring method starting from the analysis of the LED light electric characteristic.High-power LED lighting physical model is established to draw system block diagram ofthe model which need identified. Testing platform of120W high-power LED lightingsystem is built to measure input power and junction temperature, Analysis the relationshipbetween input power and the junction temperature, providing data support for systemidentification.Established a input-output BP neural network model of the high-power LED lightingsystem, input and output variables are identified by BP network identification algorithm,and obtained the junction temperature and input power network model, then testing theperformance of the network; Optimize BP network algorithm by using Particle Swarm which is improved based on Genetic Operators,The experiment data is used to simulationand system identification test, the input-output network model after optimizationalgorithm is obtained and then testing the performance of the network, to verify theaccuracy and reliability of the model, the two simulation results of the network models arecompared to show the optimized network model has the higher accuracy. The resultprovides the theory basis to establish feedback control system of the LED lightingluminaire and it have practical significance.