AI methods for development and condition monitoring of energy storage devices

This project aims to develop a physics-based AI framework for the characterization, performance evaluation, and lifetime prediction of energy storage devices, including high-voltage power capacitors, batteries, and supercapacitors. By utilizing AI techniques such as Physics-Informed Neural Networks (PINNs), the project will enhance material selection and condition monitoring, ensuring reliable operation under various stress conditions. 

The outcome will contribute to the development of a new generation of energy storage devices that can meet the growing demands of modern power systems, particularly with the integration of renewable energy sources and HVDC technology.

This research project focuses on developing a physics-based AI framework to improve the performance and lifespan of energy storage devices like high-voltage power capacitors, batteries, and supercapacitors. These devices are vital in modern power systems for short- and mid-term energy storage, enabling stable grid operation and power quality control. However, during operation, they are often subjected to overstresses, which reduce their lifespan. We aim to address these challenges by using AI-driven methods for material characterization, performance evaluation, and lifetime prediction. Energy storage devices play different roles in power systems depending on their properties. Batteries, for instance, are efficient in applications that require high energy density but suffer from slow response times and relatively low power density, making them unsuitable for transient high-power demands such as those posed by renewable energy sources. On the other hand, supercapacitors offer a much faster response and higher power density, making them more suitable for applications like grid stabilization, where high amounts of energy need to be released quickly. Meanwhile, metalized film capacitors provide the highest power density and are crucial for high-voltage power grid applications due to their efficiency and reliability.

To address current shortcomings in the design and performance of these devices, the project proposes using a physics-based AI method, specifically Physics-Informed Neural Networks (PINNs). These AI models incorporate the fundamental physics governing the devices, described by partial differential equations (PDEs), into the machine learning framework. The models are expected to improve upon traditional methods, such as finite element modeling (FEM), which struggle with numerical problems in highly nonlinear conditions or when dealing with steep gradients in charge density distribution.

The project will explore the use of PINNs and other AI techniques to simulate the transport of electrical charges, dielectric polarization, and aging mechanisms in energy storage materials. The goal is to develop models that can predict the performance and aging of materials based on their responses to electric field stress and temperature variations. These models will ensure that energy storage devices can operate reliably within specified ranges of electric field strength and temperature, while also predicting their lifetime. The project will benefit from data provided by industrial partners and ongoing research, and its applicability will be tested across various energy storage technologies, including batteries and supercapacitors.

This research is highly relevant to the evolving energy landscape, where the integration of renewable energy sources and the use of High Voltage Direct Current (HVDC) technology demands more reliable, efficient, and durable energy storage devices. By developing AI-based tools for material characterization and condition monitoring, the project will contribute to a more sustainable and resilient energy system.

Involved in the project

Chalmers: PhD student Emir Esenov, Professor Yuriy Serdyuk (project coordinator, examiner), Associate Professor Thomas Hammarström (main supervisor), Assistant Professor Christian Häger (co-supervisor)
Hitachi Energy: Adjunct Professor Olof Hjortstam (co-supervisor)
Karlstad University: Associate Professor Jorge Solis (researcher)
Linköping University: Professor ReverantCrispin (advisor)

Partners

Hitachi Energy, Redox Me, Ligna Energy, Chalmers University of Technology, Karlstad University, Linköping University