Design Of A TEC Temperature Control System Without Feedback Based On Neural Network Prediction

In the field of laser industry or scientific research,many high power semiconductor devices such as lasers and photoelectric detectors usually run at a constant temperature for a long period of operation.Therefore,control of temperature within a given range must be guaranteed to avoid overheat and damage.With the rapid development of semiconductor technology,a thermoelectric semiconductor cooling device has been invented.The device is called thermoelectric cooler(TEC).The thermoelectric cooler has the features of high efficiency of temperature control,low thermal inertia easy to control,no noise during operation,and no need of freezing medium.Whether the cooling or heating function is determined by the direction of its input current.The properties of TEC make it suitable to control a device that needs to operate within a proper temperature range.This thesis carried out the analysis of related thermoelectric driving devices on the market,and extended and upgraded the existing circuit system and control algorithm,making it more flexible and reliable.The hardware uses the C2000 series digital signal controller to generate multiple high precision PWM signals,control a two phase interleaved parallel synchronous buck converter,and meet the cooling or heating need of the thermoelectric cooler.In the software design,the anti-integration saturation and integral separation PID algorithm is used in the control strategy,so there will be no overshoot and oscillation when the constant temperature control of the target device is operating.In applications,a temperature sensor is required to be tightly coupled with the controlled device for the driver to acquire the temperature of target device.However,there is a risk that the temperature sensor will be broken due to the problem of mechanical assembly or the quality of the sensor component.If that happens to be an open-loop,the temperature of target device will be out of control.To deal with this problem,this thesis presents a non-contact high precision measurement method,and the method can replace the main sensor when it is open-looped.The accuracy of infrared sensor is low,and it needs to be calibrated.The temperature data and distance data in the temperature control process will be acquired by the upper computer software,then the temperature and distance data will be fitted by RBF neural network.When the sensor is open-looped,the temperature data given by the neural network will be used to make the control process close-looped.If the infrared sensor is broken too,the history data from the DSC will be uploaded to the upper computer software,will be trained by the neural network to predict the trend of temperature data variation and the data will be used for the temperature control.The experiment results have shown that the developed system has the advantages of reliable operation,fast response,high accuracy,and high power density for the temperature control,and it can meet needs for most of the temperature control applications.Even in the case of open loop of temperature sensor,the system can also meet the accuracy requirement of temperature control.