The Development Of High Reliability Switching Power Supply Based On Neural Network Temperature Drift Compensation

Traditional experiment shows that the electronic components may have a 10% drop in reliability when they have a two degree rise in temperature. For the high frequency switch power supply with high-power heating components, the treatment design of temperature drift is of great importance. Traditionally, it may be considered to reduce the temperature rise by natural cooling and fan cooling. However, natural cooling needs larger mental radiator to conduct heat directly; fan cooling is demanding for mean time between failures, also it is difficult to lower the temperature when the surrounding temperature is high. In view of the limited scope of use for traditional power supply and disability in low temperature environment, this article proposes a research for the reliability of switching power supply based on neural network temperature drift compensation, expects to get more stable performance in practical application of high frequency switch power supply. In order to improve the reliability of switch power supply, this article mainly involves the following aspects:(1) The heat sources on the PCB is designed reasonably by consult a large number of power thermal design documents, we simplify and optimize the circuit design to ensure that electronic components under extreme conditions and the thermal stress caused by temperature change does not exceed the allowed range. We also select the components with long-life and durability to improve redundancy of circuit design and components work safety to ensure the whole system work safely and stably in a long term.(2) In view of the inherent defect that analog control circuit structure is complex and unchangeable once it is designed, we design a digital power supply loop control which is based on DSP chip F28027. We can program the power control strategy through chip F28027 to compensate the temperature drift, and also to reduce the impact on the output voltage of the power supply.(3) In order to solve the low power factor of traditional power supply and big harmonic current, this paper design a power factor correction circuit based on chip L6562, which makes the power supply power factor increase to 0.97, and also improve the utilization efficiency of the electric energy.(4) In view of the worse loop stability of power control strategy in the condition of wide temperature range, a temperature drift compensation method based on neural network is presented. According to the law of the temperature drift in the different load, a parameter model is built which is based on BP neural network, we can get the temperature drift compensation data query table after training, and the temperature drift can be compensated more quickly through the query table.The experiment is done in the wide temperature range(- 30 ℃ to 60 ℃), compare the data before and after adding the temperature drift compensation in different load. For the AC-24 VDC switching power supply(80W, power factor 0.93), the mean square error of output voltage is improved.