Doubts concerning Buildings.Fluid.HeatPumps.EquationFitReversible model

[JordiGarciaCespedes]

Dear all,

I am using the model
Buildings.Fluid.HeatPumps.EquationFitReversible
and I would like to clarify the interpretation and internal consistency of the performance data records provided with the library.

When evaluating several example data sets (e.g. Trane_Axiom_EXW240 and EnergyPlus) exactly at the reference conditions defined in the records (i.e. inlet temperatures equal to TRef* and mass flow rates equal to the nominal values), I observe that the resulting heating/cooling capacity and electrical power do not match the corresponding Q_flow and P parameters. Instead, deviations on the order of 10–50% occur.

This appears to be caused by the fact that the EquationFit coefficients (coeQ, coeP) are not normalized such that their sum equals 1 at the reference point, meaning that Q_flow and P act as scaling parameters rather than true nominal values.

In addition, this behavior leads to very high COP values at the reference conditions. For example, in the Trane_Axiom_EXW240 heating data set (T_source = 18 °C, T_load = 53 °C), the model yields a COP of approximately 7.7, corresponding to more than 80% of the Carnot efficiency based on water-side temperatures. This seems unrealistically high for a real machine and suggests a possible inconsistency between the fitted coefficients, the reference values, and their intended physical interpretation.

Could you please clarify:

1. Whether Q_flow and P in these data records are intended to represent the actual nominal performance at the specified reference conditions, or merely scaling values for the regression.
2. Whether the EquationFit model is expected (by design) to reproduce the reference performance exactly at the reference conditions.
3. Whether such high COP values at the reference point are expected for these examples, or if this indicates an inconsistency in the data records.
4. What the recommended approach is for users who want to construct custom data records that are consistent at the reference operating point.

Thank you very much for your time and for maintaining the Buildings library.

Best regards,
Jordi

[mwetter]

The model Buildings.Fluid.HeatPumps.Validation.EquationFitReversible_EnergyPlus computes for example the available heat as

Q̇_ava = ( α₁ + α₂ T_loa,ent/T_RefHeaLoa + α₃ T_sou,ent/T_RefHeaSou + α₄ ṁ_loa,ent/(ṁ_loa,0   s) + α₅ ṁ_sou,ent/(ṁ_sou,0   s) )   Q̇₀   s.

Let's assume for simplicity that $s=1$. Then, it is the product of the $(a_1 \ldots )$ times $\dot Q_0$ that determines $\dot Q$. One could for example replace $\dot Q_0 \leftarrow 0.5 \dot Q_0$ and multiply all coefficients by a factor of $2$ and still get the same $\dot Q$.
So, the coefficients need not sum up to $1$. I think it would have been clearer to absorb $\dot Q_0$ in the coefficients, but this is not done in this model in order to have the same model as is implemented in EnergyPlus.

Regarding the high COP, this is often the case if one extrapolates using a polynomial. I did not find an information about the validity range of these coefficients that we took from EnergyPlus, but I assume they were computed for too small a range of data and hence you encounter a large extrapolation error.
Please note that we often use a model from Buildings.Fluid.HeatPumps.ModularReversible in order to not having to use polynomials.

Note that T_RefHeaLoa is not really a reference conditions for nominal operation, but rather a temperature around which the data fit is done.

[JordiGarciaCespedes]

Thank you very much for your clear explanation.

I now understand that TRefHeaLoa and TRefHeaSou are not intended to represent nominal operating conditions, and therefore there is no requirement for the coefficients to sum to one.

To better understand the behavior of the model, I evaluated both the EnergyPlus example and the Trane Axiom EXW240 data set over a range of operating conditions (not only at the reference temperatures). I attach a simple spreadsheet with these calculations.

From these checks:

The EnergyPlus coefficients seem to produce performance trends that remain physically plausible over a reasonable operating range.

In contrast, the Trane Axiom EXW240 coefficients can lead to very high COP values and unexpectedly low electrical power, even relatively close to the reference temperatures.

This makes me concerned that the Trane coefficients may not represent a physically consistent performance surface, or that their intended validity range may be quite limited.

Best regards,
Jordi

Buildings.Fluid.HEatPumps.Data.EquationFitReversible_Examples.xlsx

[mwetter]

Most likely, these coefficients were calculated based on a very narrow data range that was provided by the manufacturer.