Plots look downsampled compared to the original data

[cyasing]

I simulated the WLTC drive cycle (using my own data), with a power function input. The current, voltage and power plots look really smoothed out compared to the original data that I have (also plotted for comparison). What could be the possible reasons?

Plots:

1. Original drive cycle:
   

2. Simulation Result:
   

3. Power Function:
   

You can see that some of the trends in the data are captured, but yeah, not completely accurate.

[YannickNoelStephanKuhn]

Power inputs are handled via an algebraic function; that is, an extra equation gets solved for the current that reproduces the requested power input. So you probably just need to increase the temporal resolution of your simulation or lower the absolute or relative tolerances. That way, the current needed to mimic the power input gets computed more accurately.

[cyasing]

Sorry about that, I meant "all it takes" in perspective of setting up the power to current conversion using CasADi. I was aware of CasADi solver from our earlier discussion, just wasn't sure how P = V * I was enough information to get current values. So these DAE's are the SPM/DFN/etc. models that we choose right?

[YannickNoelStephanKuhn]

The DAEs are the SPM/SPMe/DFN models, yes. P = V * I is enough information because CasADi solves for the current that satisfies P_applied - V * I = 0. In terms of internal equation structure, V is a function of current. So it's not that different from solving for the current that satisfies V_applied - V = 0.

[cyasing]

Nice, that makes sense. Thank you, it's been bugging me since forever!

[valentinsulzer]

We never wrote a paper on this, but luckily @MarcBerliner did https://iopscience.iop.org/article/10.1149/1945-7111/ac9a80/pdf (he calls it "General Operating Modes")

[cyasing]

@tinosulzer I see. That's neat!

[valentinsulzer]

Looks like a time scaling issue. How are you defining the interpolant? It should be something like this, but for power:

# create interpolant - must be a function of *dimensional* time
timescale = param.evaluate(model.timescale)
current_interpolant = pybamm.Interpolant(drive_cycle[:, 0], drive_cycle[:, 1], timescale * pybamm.t)

You won't need to specify the timescale any more once we remove the non-dimensionalization. But for now you do.

[cyasing]

Oh no, here's what I got:

[cyasing]

Apparently, all I needed to do was to eliminate the extra columns of data in my drive cycle file! I also had a bit of calculation on that data in the script, I moved it into excel. Then, it simulated well! That's so strange to me.

The 'power_interpolant.evaluate(t_eval)' did not get better, though.

[cyasing]

power_interpolant.evaluate(t_eval)

Is my result something to be worried about? It does look like it to me.

[valentinsulzer]

Do power_interpolant.evaluate(t_eval/timescale). The power interpolant is stored as a function of dimensionless time

[cyasing]

That makes sense. Here, I plotted it:

Looks like it's fine now. Thanks for all your help!