Too much runtime for a DFN + Thermal model simulation

[abhishek-appana]

Hello everyone,

Can anyone give me an idea of how long it will take to run DFN + Thermal model simulation? Here are more details of the setup! I left the simulation running for a day and it's still going on. I noticed on this link that the airspeed velocity is very low for this simulation. I don't understand why it is taking so much time for me!

Chemistry: NCA_Kim2011
Thermal model: Pouch cell - 2D current collectors [Options: "Dimensionality": 2, "Thermal": x-lumped]

I am running this simulation on a desktop with the following specs:
RAM: 32 GB
CPU Base speed: 2.7 GHz
L1 Cache: 1.2 MB
GPU memory: 16 GB

Below is the minimal working example:

import pybamm
import os
import pandas as pd
from pybamm.expression_tree.functions import exp
os.chdir(pybamm.__path__[0]+'/..')
import numpy as np
import timeit
from scipy import interpolate


experiment = pybamm.Experiment(
    [
        (
            "Discharge at C/20 for 24 hours or until 3 V",
        )
    ]
)

options = {"dimensionality": 2, "thermal": "x-lumped"}
model = pybamm.lithium_ion.DFN(options=options)
parameter_values = pybamm.ParameterValues("NCA_Kim2011")

solver = pybamm.CasadiSolver(mode="safe",dt_max=10)
sim = pybamm.Simulation(model,parameter_values=parameter_values, solver=solver,experiment=experiment)
sim.solve()

[MarcBerliner]

Hi @abhishek-appana, I tried running the code but also did not get a solution within 15 minutes or so.

We are actively working on improving the solvers in PyBaMM. Can you try using the develop branch of PyBaMM and switching to IDAKLUSolver, i.e.

solver = pybamm.IDAKLUSolver()
sim = pybamm.Simulation(model,parameter_values=parameter_values, solver=solver,experiment=experiment)
sim.solve()

With the new solver, I got an answer in 2 minutes and 15 seconds on my machine.

[benshang1997]

Hi @abhishek-appana , thanks for your help, i will try.

[benshang1997]

Hi @abhishek-appana, this error means that you don't have a LaTeX preprocessor installed. Since the rest of the tests are passing, you can safely ignore this – PyBaMM has been installed correctly, and the IDAKLU solver should now be usable. The only functionality for which LaTeX is required is for the API that allows printing the equations for the model it is called for.model.latexify()

Hi @agriyakhetarpal thanks for your suggestion, I think I meet the same problem, Do I install it successfully too? and how should I do if i still want to install the latex? thankyou

And I can't find run-tests.py to run the test. but I have a test folder with many sample。 so I wonder if the run-test has been changed name?

[agriyakhetarpal]

Hi @benshang1997, you can skip this test safely. However, if you still wish to install a LaTeX distribution, you can try doing sudo apt-get install texlive-full (you might not need the full distribution, so you can replace it with the one you need, or a smaller one).

We remove the run-tests.py file, so the tests can be triggered entirely using pytest (and nox). Here's the relevant section in the Contributing guide: https://docs.pybamm.org/en/latest/source/user_guide/contributing.html#testing (note that this is from the latest section of the docs because it hasn't been released in the stable version yet).

[benshang1997]

嗨，您可以安全地跳过此测试。但是，如果您仍然希望安装 LaTeX 发行版，您可以尝试这样做（您可能不需要完整的发行版，因此您可以用您需要的发行版或更小的发行版替换它）。sudo apt-get install texlive-full

我们删除了该文件，因此可以完全使用 （和 ） 触发测试。以下是贡献指南中的相关部分： https://docs.pybamm.org/en/latest/source/user_guide/contributing.html#testing（请注意，这是来自文档的部分，因为它尚未在版本中发布）。`run-tests.py``pytest``nox``latest``stable`

Hi @ @agriyakhetarpal thanks for your guide I have solved this issued, but the solver maybe meet a the convergence issues as belowed when I run expriement:

At t = 279.719 and h = 1.39219e-11, the corrector convergence failed repeatedly or with |h| = hmin.

but the experiment is not stopped with this issue. so I wonder does this impact the simulation result? and could you please give me some advices on it, i also met the same issues in casadi and that is why i change to use IDA

[MarcBerliner]

At t = 279.719 and h = 1.39219e-11, the corrector convergence failed repeatedly or with |h| = hmin.

Hi @benshang1997, I'm just double-checking. Are you using the develop branch of PyBaMM? This works on my machine using the develop branch (not the latest PyBaMM release).

[benshang1997]

Hi @abhishek-appana I am install a new env and try to install IDAKLU according the guide https://docs.pybamm.org/en/stable/source/user_guide/installation/install-from-source.html.

But I meet some issues, I can clone the GitHub repository successfully but I cant run this :sudo apt install python3.X python3.X-dev libopenblas-dev gcc gfortran graphviz cmake pandoc

I cant use nox if I skip this.
(pybamm_ubun is my new env and it is empty before I clone pybamm)

could you please give me some suggestion? I am a starter on linux too.
thanks a million

[abhishek-appana]

Hi @benshang1997 ,

I don't have experience with linux either. Are you using wsl?

Try running sudo apt update && upgrade before running the install command. I don't know if it'll help but just a try! I don't have any idea as to why this is happening

[agriyakhetarpal]

You might not receive updated versions of packages such as python3.12-dev from apt. I would suggest installing via a different package manager, say, something like Linuxbrew or getting it from the official releases on https://www.python.org/downloads/. Another method would be to use conda/mamba, which is probably the easiest method.