Battery Cycle Counting

Hi

I am running simulations of PV+Battery and we are very interested in the battery degradation over the lifespan. We have recently found some inconsistencies in how SAM counts the number of cycles.
How does SAM count the discharges below the set depth of discharge (I have set it to be 80% DoD). Ie when DoD is 40%? (As per my attached image).
In the attached image, there are some inconsistencies as to how SAM has calculated the number of cycles. This has raised the question regarding the validity of the battery degradation.

So how does SAM count the number of cycles? Is this value accurate when determining the lifespan of utility scale BESS?

I am more than willing to share my SAM file if you want.

Hi Michael,

Please do send a .sam file to help me understand your question. I'm not sure what you mean by "I have set it to be 80% DoD." Do you mean that you set the minimum state of charge on the Battery Dispatch page to 20%?

Best regards,
Paul.

Hi Paul

I set the "Maximum state of charge" to 90% and "minimum state of charge" to 10%... So effective DoD is 80%. 
I have attached the SAM file. I hope that helps.

Hi Michael,

Thank you for sharing the .sam file.

I'm seeing different results than you originally showed. The cycle count seems to follow the state of charge reasonably well:

 

I'm using a different weather file, which might affect how the battery is charging during the day, but I think it's more likely that the manual dispatch in your original file is different from this one. Could you either share your weather file or a file with dispatch set up to replicate the results in your screenshot?

Thanks,
Paul.

Hi Paul

My mistake, I removed the wrong sub models when I went it to you.

I have included both the weather file and updated model. Scenarios 1B, 2A and 2B all show the same counting mechanism.

Regards
Michael

Hi Michael,

Thanks for sending the files.

The following graph from case "2B" shows an example where the cycle counting algorithm appears to miss a charge/discharge cycle:



SAM uses a rainflow cycle count, which takes some time to wrap one’s head around: en.wikipedia.org/wiki/Rainflow-counting_algorithm .

The graph above shows a total of 13 cycles over the period of 11 days. The rainflow algorithm was designed for stress-strain testing based on analysis at the end of an experiment. Our our real-time adaptation for SAM sometimes has “delayed” cycle counts. Cycles being delayed by 2-3 days shouldn’t have a big impact on SAM degradation calculations.

There is a debate in the battery literature about the best method for cycle counting. SAM uses rainflow counting, but full equivalent cycle counting and other methods are under discussion: www.sciencedirect.com/science/article/pii/S2352484722022946 .

The rainflow method was popular when we first implemented SAM's battery model, and we have not had a chance to re-evaluate the latest literature and make updates.

As of SAM 2024.12.12, the battery models line up well with results from github.com/NREL/BLAST-Lite . For a high-level overview of that comparison see Mirletz, B.; Guittet, D. (2022) Recent Improvements in PV+Battery Modeling in NREL's System Advisor Model. 2022 PV Performance Modeling Workshop. NREL/PO-7A40-83677 ( PDF 1 MB ).

If you are concerned about SAM underestimating the number of battery charge/discharge cycles, you could either adjuste the battery bank replacement threshold on the Battery Life page, or use the "replace at specified schedule" option to force a battery replacement so they are accounted for in the project costs.

Best regards,
Paul.