A super-junction retroviral core battery multi-physics field simulation data prediction system

Abstract

This invention discloses a multiphysics simulation data prediction system for superjunction radiation-voltage nuclear batteries, belonging to the field of data prediction technology. It includes: a data preprocessing module for collecting physical field simulation data, performance indicators, and related parameters of the nuclear battery, and outputting a preprocessed dataset; a feature extraction and decomposition module for calculating the mutual information entropy between physical fields, decomposing the data into high-frequency, mid-frequency, and low-frequency modal components based on variational mode decomposition, and constructing a multi-dimensional, multi-scale coupled feature set; a multi-model prediction and fusion module for adapting to long short-term memory networks, random forests, and gradient boosting models, predicting short-term, medium-term, and long-term results respectively, and fusing the prediction results through a weighted voting method; and a model optimization and iteration module for hierarchical analysis of prediction errors, collaborative optimization of variational mode decomposition parameters and model parameters, outputting the optimal parameter configuration and prediction results, significantly improving the accuracy and robustness of nuclear battery performance prediction.

Images