Predicting the Temperature of the Electrolytic Capacitor Based on Neural Network Algorithm

Air-cooled power can rapidly dissipate heat, eliminate more heat, and prevent equipment from overheating. It is widely applied in the field of engineering. Capacitors have the highest failure rate of the overall power supply. This article presents a noninvasive method for estimating the temperature of electrolytic capacitors in air-cooled power supplies. The technology is suitable for power sources that cannot be turned off for intrusive detection of the temperature of electrolytic capacitors, such as nuclear power supplies. The proposed method makes it possible to measure the ambient temperature of electrolytic capacitors in nuclear power supplies. Therefore, the safety performance of nuclear power is greatly improved. It is sufficient to measure the output voltage, output current, air inlet temperature, air outlet temperature, and fan speed. This method only needs to build a model based on historical data or data from the same series of power supplies and then input these parameters into the trained neural network model. The ambient temperature of the capacitor can be estimated by using this model. Finally, the prediction results of four neural networks are compared using a 3-kW air-cooled power supply as an example, and the applicability of the neural network technique is demonstrated.