Running repeated part load simulation often fails

[fwitte]

Likely there needs to be a reset to design point before running next part load simulation...

[fwitte]

Reproduction

"""Reproduce convergence issues when running offdesign_simulation twice.

Mirrors the dashboard scenario:
- Model: HeatPumpIHX (simple cycle with IHX)
- Refrigerant: Butane (R600)
- Design: default values from params_hp_ihx.json
- Part load: 55 % .. 100 % nominal mass flow
- Heat sink feed flow: 80 .. 100 °C (variable)
- Heat source: constant (design temperature)
"""
import numpy as np
import contextlib
import io

from heatpumps.models import HeatPumpIHX
from heatpumps.parameters import get_params

params = get_params('HeatPumpIHX')

# Refrigerant selection as done by the dashboard for "Butane"
params['setup']['refrig'] = 'R600'
params['fluids']['wf'] = 'R600'

# Sidebar offdesign settings, replicated from hp_dashboard.py
params['offdesign']['partload_min'] = 0.55
params['offdesign']['partload_max'] = 1.00
params['offdesign']['partload_steps'] = int(np.ceil((1.00 - 0.55) / 0.1) + 1)

# constant heat source
params['offdesign']['T_hs_ff_start'] = params['B1']['T']
params['offdesign']['T_hs_ff_end'] = params['B1']['T'] + 1
params['offdesign']['T_hs_ff_steps'] = 1

# variable heat sink 80..100 °C
params['offdesign']['T_cons_ff_start'] = 80
params['offdesign']['T_cons_ff_end'] = 100
params['offdesign']['T_cons_ff_steps'] = int(np.ceil((100 - 80) / 1) + 1)

params['offdesign']['save_results'] = False

hp = HeatPumpIHX(params)
hp.run_model()
print(f'Design solved, COP = {hp.cop:.3f}, m_design = {hp.m_design:.3f}')


def run_and_count(tag):
    buf = io.StringIO()
    with contextlib.redirect_stdout(buf):
        hp.offdesign_simulation()
    q = np.asarray(hp.Q_array, dtype=float)
    n_nan = int(np.isnan(q).sum())
    n_tot = q.size
    print(f'[{tag}] failed points: {n_nan}/{n_tot}')
    return n_nan


n1 = run_and_count('run 1')
n2 = run_and_count('run 2')
print()
print(f'RESULT: run1 failures = {n1}, run2 failures = {n2}')

[jfreissmann]

Hey @fwitte,

thank you for sharing your issue! I tried to reproduce the problem locally by running it with the current dev state of heatpumps (soon to be published as v1.5.0), which yielded no fails in both runs. Running the script on main (currently stable version v1.4.1, based on TESPy < v.0.10) also leads to 0 runs failing. What version of heatpumps and TESPy did you use?

[fwitte]

TESPy current dev and heatpumps current dev

[fwitte]

Actually, I might have used dev of my fork...