Parameter identification of DC–DC converter based on dendrite net under fluctuating input voltages

Abstract

Abstract The traditional identification methods of direct current‐to‐direct current (DC–DC) converters usually consider the common situation where the training and testing sets have the same input voltage. In real‐world applications, the actual input voltage may fluctuate beyond its original value, thus negatively affecting the parameter identification accuracy. Therefore, this paper presents a parameter identification method under fluctuating input voltages. Firstly, to increase the identification accuracy, the paper proposes the coefficient of variation method to extract key features from all time‐ and frequency‐domain features as training data. Then, for fluctuating input voltages, the paper adds the input voltages into the training data to build the parameter identification network based on Dendrite Net. When using the established network to identify component parameters in practical applications, the actual input voltage may change to an unknown value because of irregular fluctuation. Hence, the actual input voltage needs to be identified before identifying the component parameters of a DC–DC converter. Finally, the paper builds the input voltage identification network. By combining the identification results of the actual input voltage, the parameter identification under fluctuating input voltages can be completed. The simulation and hardware experiments results validate the practicability and effectiveness of the proposed method.

Keywords

• DC–DC power convertors
• fault diagnosis
• identification
• power conversion