AUTHOR=Liu Xiaofeng , Yan Zhiyong , Leng Fangling , Bao Yubin , Huang Yijie TITLE=Machine learning predictive model for electronic slurries for smart grids JOURNAL=Frontiers in Energy Research VOLUME=10 YEAR=2023 URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.1031118 DOI=10.3389/fenrg.2022.1031118 ISSN=2296-598X ABSTRACT=

Electricity is a fundamental energy that is essential to the growth of industrialization and human livelihood. Electric power resources can be used to meet living and production needs more steadily, effectively, and intelligently with the help of an intelligent power grid. The accuracy and stability of component requirements have increased due to the rapid growth of intelligent power networks. One of the fundamental components for component production is electronic slurry, so optimizing electronic paste’s properties is crucial for smart grids. In the field of materials science, the process of discovering new materials is drawn out and chance-based. The traditional computation process takes a very long time. Scientists have recently applied machine learning techniques to anticipate material properties and hasten the creation of novel materials. These techniques have proven to offer amazing benefits in a variety of fields. Machine learning techniques, such as the cross-validated nuclear ridge regression algorithm to predict double perovskite structure materials and the machine learning algorithm to predict the band gap value of chalcopyrite structure materials, have demonstrated excellent performance in predicting the band gap value of some specific material structures. The performance value of other structural materials cannot be predicted directly by this targeted prediction model; it can only forecast the band gap value of a single structural material. This study presents two model techniques for dividing data sets into element kinds using regression models and dividing data sets into clusters using regression models, both of which are based on the fundamental theory of physical properties, band gap theory. This plan is more efficient than the classification-regression model. The MAE dropped by 0.0455, the MSE dropped by 0.0425, and the R2 rose by 0.022. The effectiveness of machine learning in forecasting the material band gap value has increased, and the model trained by this design strategy to predict the material band gap value is more reliable than previously.