Data-driven battery state of health diagnostics and prognostics

Unless referenced, this is paraphrased from my research proposal or taken from the original project summary put forward for the application by Professor David Howey:

Battery degradation has a large impact on the performance of the energy storage device. As renewable energies, electric vehicles (Oxford has just pledged an all-electric zone by 2020 which will cover the whole city by 2035) and viable second life applications become more and more prevalent and available, the need to maximise the capabilities of our energy storage becomes a larger and larger problem. There has been research into the area to keep up with the increasing demand, but a validated algorithm for general use in the real world modelling battery degradation has yet to be created. Degradation is non-uniform, path-dependant and multi-faceted and involved chemical, electrical and mechanical factors. An understanding of physics, chemistry, electronics and statistics needs to be brought together in order to appreciate this problem. The issue is intensified at pack level where detail of the internal configurations may not be fully known.

My research aims to combine the large quantities of real world data, compiled by the Department, with the increasingly available new techniques for producing a diagnosis during battery degradation in order to produce a data focused technique to analyse a given battery. The purpose of this technique would be to perform both prediction of future performance (eg battery capacity, faults, etc...) and diagnosis after testing a given battery. Consequently mapping the difference between laboratory behaviour to real world examples will form a very large part of this research. As stated in my proposal, a positive result would be the creation of software capable to using the data produced in future to predict the state of health of a battery in a reliable fashion. This software, and indeed much of my research, will be using existing models, or indeed a slightly altered form of current models.

A more detailed and chronological group of aims follows. Following a meeting with Siemens to decide the specific targets of this research project, I will use the existing literature to form an idea of where in the market the largest gaps are to be found. I will need to ensure that the data requirements from Siemens, the previous data and the physical models are all established and well understood. Then research into various methods of life prediction can be used as small trials of the overall problem. Hopefully, an algorithm will be produced at this point and this can be run iteratively where required. After large improvements, I will aim to run practical experiments to verify the algorithm on real-world systems.

This project falls within the EPSRC engineering research area. Specifically, this is within the Energy and Power group in the Engineering Department.

This project is in collaboration with Siemens.