Design And Optimization Of The Temperature Drift Compensation Of Switching Power Supply Based On Least Squares Support Vector Machine

With the continuous economic development, switching power supply has been widely used in the fields of power industry, medical treatment, communications and others. Studies have shown that the reliability of the components of power electronics would decrease by 10% when its temperature rises in every 2 degrees. In order to meet the requirements of working in a wide temperature range and reduce the impact of temperature drift of the switching power supply, this dissertation presents a new method to compensate the temperature drift based on Least Squares Support Vector Machine(LSSVM). It mainly focuses on the design and optimization of the temperature drift compensation of the switching power supply.Firstly, on the basis of referring to a great quantity of relevant literature, research of state of art and trend as well as main problems of the switching power supply in China and abroad is carried out. Then the significance and the main work of this dissertation is introduced briefly.Secondly, design of the digital control loop is implemented by combining the hardware platform with the control loop method, including the main program of the system, the A/D interrupt service routine, the fuzzy PID and the compensation program of the temperature drift.Thirdly, output voltage is taken as the research subject of the study, then the difference between fuzzy control and classic PID control strategy is briefly introduced.Meanwhile, this dissertation also covers the implementation of the fuzzy PID in the control system.Finally, aiming at reducing the influence of temperature drift on the loop control in wide temperature environment, the dissertation presents a method of temperature compensation of the switching power supply based on LSSVM. A suitable LSSVM prediction machine is established on the basis of the law of temperature drift in a wide temperature environment. In addition, the particle swarm optimization(PSO)algorithm is introduced to improve the performance of the LSSVM predictor, and its effectiveness is verified by simulation. For the sake of speeding up the LSSVM, the grey relational analysis(GRA) method is employed to preprocess the data of the temperature drift. Then the temperature compensation table is produced according to the prediction value of the LSSVM and the expectation value of the output voltage.The method is implemented in a DSP controller, simulation and experiment results show that the presented method can improve the fluctuation range of the output voltage under different temperature.