# Prediction of Energy Storage Performance in Polymer Composites Using High‐Throughput Stochastic Breakdown Simulation and Machine Learning

## TL;DR

This paper uses simulations and machine learning to predict the energy storage performance of polymer composites for capacitors.

## Contribution

A novel method combining high-throughput simulation and machine learning to predict breakdown strength and energy storage density in polymer composites.

## Key findings

- High-throughput stochastic simulations were performed on 504 data groups to predict breakdown strength.
- Machine learning models accurately predicted energy storage density when combined with classical dielectric formulas.
- Predictions were validated through directional experiments on dielectric constant and breakdown strength.

## Abstract

Polymer dielectric capacitors are widely utilized in pulse power devices owing to their high power density. Because of the low dielectric constants of pure polymers, inorganic fillers are needed to improve their properties. The size and dielectric properties of fillers will affect the dielectric breakdown of polymer‐based composites. However, the effect of fillers on breakdown strength cannot be completely obtained through experiments alone. In this paper, three of the most important variables affecting the breakdown strength of polymer‐based composites are considered: the filler dielectric constants, filler sizes, and filler contents. High‐throughput stochastic breakdown simulation is performed on 504 groups of data, and the simulation results are used as the machine learning database to obtain the breakdown strength prediction of polymer‐based composites. Combined with the classical dielectric prediction formula, the energy storage density prediction of polymer‐based composites is obtained. The accuracy of the prediction is verified by the directional experiments, including dielectric constant and breakdown strength. This work provides insight into the design and fabrication of polymer‐based composites with high energy density for capacitive energy storage applications.

The effects of single variable and multivariable coupling on the breakdown strength of polymer‐based composites are studied by high‐throughput stochastic breakdown simulation, and a machine learning database is established to obtain breakdown prediction. A universal energy storage density prediction is obtained by combining the breakdown prediction with the classical dielectric formula, and the predictions' accuracy is verified.