The Prediction On Electric Breakdown Strength And Energy Storage Performance Of Polymer Composite With Data Analysis Method

As the most important component of clean energy structure,electric energy plays a vital role in promoting social development and scientific progress.The demand for energy storage devices with high energy density is continually increasing in electric grid system,high-power electronic equipment,new energy vehicles and other fields.Dielectric capacitors have higher energy storage density and larger charge-discharge power than traditional electrostatic capacitors.The polarization and insulation properties of the composite can be improved by adding functional nanofillers into the polymer matrix.Although a large number of researchers are deeply engaged in this field,there are so many factors affecting the energy storage performance of composite media such as structure design and filler selection,which requires a lot of time,energy and resources.Relying on the advantages of cross-disciplines in big data processing,it is effective that by analyzing a large number of experimental data to predict specific properties of related materials.In this thesis,the data including PVDF composite media structure and energy storage information were collected firstly.The data set including packing shape,packing content,molecular descriptor,molecular fingerprint and actual physical properties,and the prediction model was established by combining with machine learning algorithm.Through experimental calculation,the accuracy of test set prediction reaches 77 %.Then we compared the performance data of composite dielectric films prepared in the laboratory with the predicted results of the prediction model to verify the reliability of the prediction model.In order to expanding the database to study the energy storage characteristics of different substrates and fillers.The data set was constructed by combining 1254 groups of composite energy storage density information and 869 groups of composite visual information,which including more than 20 specific attributes descriptors from three perspectives of matrix,filler,composite structure and interface design.The machine learning algorithm was used to establish the prediction model,and the highest test set prediction accuracy reached91.9 %.The composite dielectric film prepared by the laboratory was used for comparison test,it is found that the prediction model can correctly predict the range of energy storage density of composite dielectric.Then analyzing the weight proportion of each descriptor in the prediction model,from high to low,the weight of matrix performance,filler content,filler coating structure,filler properties,filler shape,multiple filler,multilayered structure were 29.3 %,24.1 %,14 %,13 %,12.2 %,5.6 %,1.8 % respectively.Finally,the prediction model was used to explore the potential effective space of structure of composite,and the prediction results were statistically analyzed to give some suggestions on the shape and coating structure of nanofillers.Then,in order to increase the types of data collection methods and expand the scale of the database,combining with the phase field theory and related energy equation,a breakdown simulation model of composite was written in MATLAB software.By adjusting the values of various parameters in simulation model with measured data,the breakdown process of one-dimensional and two-dimensional fillers in matrix was simulated at different temperatures.