Lithium battery parameter estimation and SOC
19 views (last 30 days)
I have been working on Battery Modeling and I checked several webinars on how to perform parameter Estimation of Battery to fill up the lookup tables. But I am a bit confused.
- How do you judge the minimum and maximum range of R1, tau1, R2, tau2, etc? I have checked the white paper and webinars; and they have not explained how they set-up the minimum and maximum ranges. The minimum and maximum range for Em is pretty straightforward since you can use the battery datasheet for this but I have no idea for the other ECM components.
- From the resources, it has always been mentioned that having 3 Time Constants are better than 1 Time Constant. But when I run the Battery Modeling project, the cost of the "1 Time Constant option" is a lot smaller than the "3 Time Constants" option. Shouldn't the latter be smaller than the former because more time constants should be able to simulate the experimental data?
- Some webinars and the white paper has explained how the SOC breakpoints for the lookup tables should be distributed. Could someone provide a simple example on how to distribute them? I am also confused because the SOC breakpoints in the plots of the webinars and whitepaper are in increments of 20% even though the explanations of how to distribute them implied that they shouldn't be 20%. I apologize I forgot which webinar explained this.
One out of many webinars has mentioned that the methods in Battery Modeling project is not ideal when it is implemented on lithium-polymer batteries. The explanation was not in depth as to why it was not ideal. Could someone provide a white paper or a detailed reference elaborating why implementing this on the lithium-polymer batteries is not good? I am interested in doing the Battery Modeling project on lithium-polymer batteries.
I'd really appreciate it if one of the white paper authors would be able to enlighten me on this issue. Thank you!
Juan Sagarduy on 22 Feb 2021
I’ll give a shot to the questions you have.
1. When this type of work is done with Simulink Design optimization, we leave these values to within 0 and Inf. So strictly positive. We trust that the values will converge in the process.
2. The decision on the number of constants is a difficult one. You want as few RC pairs as possible, but sometimes an ambitious fit with real data requires more than one.
3. Here the choice of 5%, 10% or 20% is a trade-off between pragmatism and fidelity. If you use 20% SOC segments, this means that the LUTs will be sufficient with only 6 elements in the SOC dimension. This would make the estimation effective. In practice, a segmentation in 10% is more common, thus making the process more demanding.
All in all, i recommend starting with a small segment of SOC and maybe two Time constants. If u find a good match, continue to a new SOC interval and so on. If not, you need another time constant and try out.
Basically, sort out the time constant decision before attempting to estimate for the whole SOC range.
Good luck Juan