A Recognition Method of Excitation Waveforms for Core Loss Analysis of Renewable Generation Converter Based on Magnetic Flux Density Peak Location and Feature-Optimized Random Forest

The study of core loss characteristics is essential for improving the performance of renewable generation converters, and the classification and identification of excitation waveforms is fundamental to analyzing core loss characteristics. The excitation waveform is primarily characterized by the temporal distribution pattern of magnetic flux density. To accurately identify different excitation waveforms, it is necessary to extract feature variables that reflect the temporal distribution of the magnetic flux density and select an appropriate artificial intelligence algorithm to establish a classification and recognition model for excitation waveforms. However, current feature extraction methods have difficulty capturing deep features beyond traditional characteristics, which results in inaccurate classification outcomes. In response, a classification and recognition method for excitation waveforms based on magnetic flux density peak location and feature-optimized random forest is proposed. Firstly, the magnetic flux density peak location after being normalized along the time axis is extracted as a deep feature. Secondly, an excitation waveform classification model is established using artificial intelligence methods such as feature-optimized random forest. Finally, the effectiveness of the proposed method is verified using magnetic flux density sampling data from four types of core materials as examples.