The `OffsetConverter` is broken in the new implementation

[fwitte]

Describe the bug
With the current formulation, the OffsetConverter can produce the outputs at the level of each respective offset without consuming anything from the input.

I am preparing a PR to solve the issue.

To Reproduce
Consider the following example of an electrolyzer producing hydrogen and heat. The electrolyzer follows a fixed hydrogen demand, a slack source is available in case the demand cannot be supplied due to part load restrictions.

from oemof import solph


es = solph.EnergySystem(timeindex=solph.create_time_index(year=2023, number=5))

eta_h2_min = 0.6                 # efficiency at minimal operation point
eta_h2_max = 0.5                 # efficiency at nominal operation point
P_in_min = 20                   # absolute minimal output power
P_in_max = 100                  # absolute nominal output power

# calculate limits of input power flow
P_out_min_h2 = P_in_min * eta_h2_min
P_out_max_h2 = P_in_max * eta_h2_max

# calculate coefficients of input-output line equation
slope_h2 = (P_out_max_h2 - P_out_min_h2) / (P_in_max - P_in_min)
offset_h2 = (P_out_max_h2 - slope_h2 * P_in_max) / P_out_max_h2


eta_heat_min = 0.3                 # efficiency at minimal operation point
eta_heat_max = 0.4                 # efficiency at nominal operation point

# calculate limits of input power flow
P_out_min = P_in_min * eta_heat_min
P_out_max = P_in_max * eta_heat_max

# calculate coefficients of input-output line equation
slope_heat = (P_out_max - P_out_min) / (P_in_max - P_in_min)
offset_heat = (P_out_max - slope_heat * P_in_max) / P_out_max

b_h2 = solph.Bus("hydrogen bus")
b_heat = solph.Bus("heat bus")
b_el = solph.Bus("electricity bus")

b_electrolyzer_h2 = solph.Bus("electrolyzer virtual input for h2")
b_electrolyzer_heat = solph.Bus("electrolyzer virtual input for heat")

source_el = solph.components.Source("electricity import", outputs={b_el: solph.Flow()})
source_slack_h2 = solph.components.Source("hydrogen slack", outputs={b_h2: solph.Flow(variable_costs=1000)})
sink_heat = solph.components.Sink("heat export", inputs={b_heat: solph.Flow()})
sink_h2 = solph.components.Sink("hydrogen export", inputs={b_h2: solph.Flow(fix=[10, 20, 30, 40, 50, 60], nominal_value=1)})

electrolyzer = solph.components.OffsetConverter(
    label="electrolyzer",
    inputs={b_el: solph.Flow(nominal_value=P_in_max)},
    outputs={
        b_heat: solph.Flow(
            nominal_value=P_out_max,
            nonconvex=solph.NonConvex(),
            min=P_out_min / P_out_max
        ),
        b_h2: solph.Flow(
            nominal_value=P_out_max_h2,
            nonconvex=solph.NonConvex()
        )
    },
    coefficients={
        b_heat: [offset_heat, slope_heat],
        b_h2: [offset_h2, slope_h2],
    }
)

es.add(
    b_el, b_heat, b_h2,
    source_el, sink_h2, sink_heat, source_slack_h2,
    electrolyzer
)

om = solph.Model(es)

om.solve("gurobi")

result = solph.views.convert_keys_to_strings(om.results())
result[("electrolyzer", "hydrogen bus")]["sequences"]

The minimum hydrogen production should be at 12, but due to the offset it is at 2.5.

	flow	status	status_nominal
2023-01-01 00:00:00	2.5	1	50
2023-01-01 01:00:00	20	1	50
2023-01-01 02:00:00	30	1	50
2023-01-01 03:00:00	40	1	50
2023-01-01 04:00:00	50	1	50
2023-01-01 05:00:00	50	1	50

The reason for that is, that the status_nominal is multiplied with the output binary variable instead of the input binary variable.

[fwitte]

Never mind, I think my example is just missing the min attribute of the H2 output Flow. Still, I would like to make the cleanup with enforcing a single NonConvex flow as the reference for all other flow's slopes and offsets.

[p-snft]

Never mind, I think my example is just missing the min attribute of the H2 output Flow. Still, I would like to make the cleanup with enforcing a single NonConvex flow as the reference for all other flow's slopes and offsets.

Does that mean this issue is closed, then?

[fwitte]

Does that mean this issue is closed, then?

Kind of yes, the linked pull requests is still valid. Now it is more of a complete rework of the OffsetConverter and might close #1010 instead?

[p-snft]

So, this was not a real bug, only the API was very inviting to put unreasonable parameters. I will close this, as it can be seen as a duplicate of the "mislieading documentation" issue.