Time-series simulations: How to update an already built model?

[nil-co]

I am using PyBaMM as part of a larger simulation framework that gives constant power targets for 60s to the PyBaMM simulation. Afterward, the results of PyBaMM are used to calculate the power target for the next 60s outside of PyBaMM.

I managed to implement a solution that creates new experiment-based simulations every 60s. However, building/discretizing the model seems to take most of the time, which makes this approach slow for long simulations.

Is there a way to use a pre-built model and only update the power target of that model for each simulation timestep?

Here is a minimum example showcasing what I am trying to do:

# new simulation each step (this approach is quite slow for long simulations)
import pybamm
pybamm.set_logging_level("NOTICE")
import numpy as np

model = pybamm.lithium_ion.SPMe()
param = pybamm.ParameterValues("OKane2022")

max_V = 4.2
min_V = 2.5
delta_t = 900
power_precision = 9

powers = [-5, -10, -15, -20]
model = pybamm.lithium_ion.SPMe()
sol_ts = None


for power in powers:
    exp_str = f"Discharge at {abs(power):.{power_precision}f} W for {delta_t} seconds or until {min_V} V"
    exp = pybamm.Experiment([exp_str])
    # how do I only initialize and build the simulation/model once and then update the power targets for each timestep?
    sim = pybamm.Simulation(model, experiment=exp,parameter_values=param)
    sol_ts = sim.solve(calc_esoh=False, starting_solution=sol_ts)

sol_ts.plot(["Voltage [V]", "Current [A]", "Power [W]"])

Below is as far as I have gotten. It gives me a solver-related error:

# attempting to keep the same simulation
import pybamm
pybamm.set_logging_level("NOTICE")

model = pybamm.lithium_ion.SPMe()
param = pybamm.ParameterValues("OKane2022")

max_V = 4.2
min_V = 2.5
delta_t = 900
power_precision = 9
powers = [-5, -10, -15, -20]

model = pybamm.lithium_ion.SPMe()
sol_ts = None
exp_str = f"Discharge at 1 W for {delta_t} seconds or until {min_V} V"
exp = pybamm.Experiment([exp_str])
sim = pybamm.Simulation(model, experiment=exp,parameter_values=param)

for power in powers:
    exp_str = f"Discharge at {abs(power):.{power_precision}f} W for {delta_t} seconds or until {min_V} V"
    exp = pybamm.Experiment([exp_str])
    sim.experiment = exp
    if sol_ts != None:
        # replace experiment keys with new experiment conditions:
        key_old = next(iter(sim.experiment_unique_steps_to_model))
        sim.set_up_and_parameterise_experiment()
        key_new = next(iter(sim.experiment_unique_steps_to_model))
        sim.op_conds_to_built_models[key_new] = sim.op_conds_to_built_models[key_old]
        sim.op_conds_to_built_models.pop(key_old)
        
        # update model parameter values
        op_number, op  = list(enumerate(sim.experiment.unique_steps))[0]
        experiment_parameter_values = sim.get_experiment_parameter_values(op, op_number)
        new_param = param.copy()
        new_param.update(experiment_parameter_values, check_already_exists=False)
        old_model = next(iter(sim.experiment_unique_steps_to_model.items()))[1]
        new_param_model = new_param.process_model(old_model, inplace=False)
        built_model = sim._disc.process_model(new_param_model, inplace=True, check_model=True)
        sim.op_conds_to_built_models[key_new] = built_model

        sim.op_conds_to_built_solvers[key_new] = sim.op_conds_to_built_solvers[key_old]
        sim.op_conds_to_built_solvers.pop(key_old)

    sol_ts = sim.solve(calc_esoh=False, starting_solution=sol_ts)

sol_ts.plot(["Voltage [V]", "Current [A]", "Power [W]"])

[TomTranter]

You can use simulation.step and use an input for the current function. Not exactly what you're after maybe as you would have to calculate power from last voltage and fix current for the next step to a constant value. Something similar might be possible with power directly but I've never tried it