Experiment setup with a custom model

[isaacbasil]

I am having some trouble running experiments in Pybamm with a custom model (which is basically the DFN). My goal is to discharge an NMC half-cell until a voltage constraint is reached, followed by a relaxation period for a prescribed length of time. Also, I need to be able to provide t_eval. When I try the following,

        discharge_step = pybamm.step.current(current_value, termination="2.7 V")
        relaxation_step = pybamm.step.current(0, duration="200 seconds")
        experiment = pybamm.Experiment([discharge_step, relaxation_step])
        self.simulation = pybamm.Simulation(self.model, experiment=experiment)

I get the following error:
"Cannot update parameter 'Current function [A]' as it does not have a default value. ('Current function [A]' not found. Best matches are []). If you are sure you want to update this parameter, use param.update({{name: value}}, check_already_exists=False)"

However, I am not sure how to set a default value for 'Current function [A]'. Or alternatively, I am not sure where to place param.update({{name: value}}, check_already_exists=False).

Any advice appreciated :)

[martinjrobins]

does your model have a default_parameter_values property set, if you don't provide one to pybamm.Simulation this is what it uses as the parameter values object. Also, does your model have a Current function [A] parameter in it somewhere?

If you are able to provide a minimal script that can reproduce the error this would be helpful

[rtimms]

Why do you need to provide t_eval as well as an experiment? You can interpolate the solution onto any times you like.

[isaacbasil]

Thanks for the comments. That's a good point, interpolating would get around the t_eval issue. Is it possible to use the experiment feature without using the Simulation class? Since I am processing the model step by step, as in this tutorial, and I guess calling the simulation class would repeat these steps?

[rtimms]

You need to use a simulation to use an experiment, but you should be able to just pass your model, mesh etc. to the simulation instead of processing the model step by step.

[isaacbasil]

ok cheers, I'll come back to this on Monday.

[isaacbasil]

I was already half way through doing this so I thought I'd try it - I solved the model for the discharge until 2.7V as normal, and then solved the model again using set_initial_conditions_from(previous_solution) for the relaxation period. That seems to work for now. Perhaps I'll switch to doing it with the Simulation class in the future. Thanks for your help!