In the first part, we presented a detailed overview of heat pump operation, highlighting the impact of anomalies on system performance and energy efficiency. We developed and validated a dynamic simulation model of a domestic air-to-water heat pump, demonstrating its ability to accurately reproduce real operating conditions. This model served as a solid foundation to explore fault scenarios without the time and resource constraints of laboratory testing.

Synthetic data generation

Since we have a reliable model available for simulations, we can use it to generate various dataset, including anomalous conditions. In fact, by tuning some components, we can replicate the fault in the real machine.
In detail, we can simulate fouling in heat exchangers, compressor inefficiencies, pressure drops, bad components control and sensor faults. We will focus on the evaporator fouling, since we can compare it with some real results acquired in the laboratory, the evaporator is a finned coil heat exchanger.
The most common fault occurring in the evaporator is the dust and dirt covering it, causing an evaporator fouling (EF). The component gives us the possibility to modify a parameter called “fouling factor”, who is low in the healthy case, while it is higher in a case of light fault and medium-high fouling. The main effects of EF on the functioning are visible on a sensible reduction of the evaporation temperature and a smaller evaporation pressure. As side effects, to mitigate the problem, the compressor increases its rps (revolutions per seconds), causing an increment of the discharge temperature and pressure, all causing an increment of the electrical energy consumption, reducing the COP.

Comparison with real data

In this chapter we will present the differences/similarities between the model and the real data. In Table 2 reports for a single operating point, all the records (measured and derived) collected using the acquisition system. 

Table 2. Comparison of real data and simulated data for healthy and evaporator fouling. the tests are all in stable conditions, with fixed thermal load (the same of the previous chapter) and fixed external temperature of 2 °C. With H we mean the Healthy case, while EF means Evaporator fouling. In the simulation columns are reported also the percentage deviation with respect to the real case.

From the results, we can notice a reduction of the performances in both real and simulation for a single operating condition. The model behaves like the real HP, except the evaporator air outlet temperature. This is due to the fact that in the fan definition we set a fixed air mass flow rate the fan has to move. With fouling, in the real scenario, we have a reduction of the air mass flow rate due to the obstruction. Instead, in the model, the fouling factor does not block the air flow, but it behaves like a less efficient exchange material. Nevertheless, the main effects are the same, as can be seen in the table 2.

Condenser fouling

The dual problem of the evaporator fouling, is the condenser fouling. In our case, the condenser refers to the water, i.e. in Heating mode, it is the responsible of the thermal load. Fouling in the condenser can occur with debris accumulating in the water or water-glycol. If the water circuit is not regularly maintained with regular filter cleaning or other rarer faults, limestone or sediments can become stuck inside the condenser, causing an inefficient heat exchange.

In the laboratory, replicating this fault is much more difficult to be replicated than the evaporator. Fortunately, we can simulate this fault very easily by tweaking the fouling factor of the condenser plate heat exchanger. As expected, the condenser fouling has a negative impact on system behaviour, the main effects are: reduced thermal efficiency, increased pressure and compressor speed and decreased COP.

The simulation was carried out by setting a sinusoidal external temperature, ranging from -5 to + 15 °C. The thermal load is constant with a water inlet and outlet temperature, respectively of 30 °C and 35 °C. Constant is also the water mass flow rate at 0.5 kg/s and constant superheat setpoint at 10 K.

Description - a: suction temperature, b: condensation temperature, c: evaporation temperature, d: discharge temperature, e: COP, f: COP

It is useful to provide insights on how much energy could be saved in a clean and efficient plant. The profile simulates how many hours per year a certain external temperature is met with the same thermal load at the condenser. Here we consider the profile of Milan.

Milan profile - Annual Electrical Energy Consumption per Temperature

Considering an average energy cost of 0,25 €/kWh, the simulated evaporator fouling will result in an extra cost of 328 € per year.

Future developments and conclusions

To conclude, we have presented a useful approach to generate semi-realistic datasets and simulate faults inside the plant easily in a short time. This is very useful to develop and test anomaly detection algorithms. Indeed, this tool helped us choosing the right algorithm to further develop, since we can test various approaches in a very limited amount of time. Rather, in the laboratory the algorithm tests are more complicated and not ever possible because they are time and resources demanding.

This blog post is based on the paper “Dynamic Modelling and Simulation of Faults in an R290-based Air-to-Water Heat Pump System”, presented on 6 June at Clima 2025, held at the Politecnico di Milano.
The paper was authored by Daniele Scapin and Mirco Rampazzo from the Department of Information Engineering (DEI), University of Padua, and Riccardo Pengo, Chiara Corazzol, and Willy Muvegi from CAREL Industries S.p.A.

References:

1. S. Capanelli (2021). Benefits and reliability of air-to-water heat pumps in residential applications, using R-290 refrigerant and an alternative design solution to guarantee high safety with standard components.
2. Siegel, Jeffrey & Walker, Iain & Sherman, Max. (2002). Dirty air conditioners: Energy implications of coil fouling. Proceedings of the 2002 ACEEE Summer Study on Energy Efficiency in Buildings. 1. Bott, T. R. (1995). Fouling of Heat Exchangers. Elsevier. 
3. Braun, J. E. (1998). "Reducing energy costs and peak electrical demand through optimal control of building thermal storage." HVAC&R Research, 4(1), 59-80. 
4. Yan, K., Huang, S., & Wang, S. (2021). "A data-driven multi-level anomaly detection method for rooftop unit HVAC systems." Energy and Buildings, 253, 110975. 
5. Tang, J., Zhang, F., Zhao, Y., Wen, T., & Zhang, X. (2021). "A hybrid fault detection and diagnosis strategy for air source heat pump heating system based on a physical model and deep learning." Applied Thermal Engineering, 197, 117965. 
6. Minsung Kim, W. Vance Payne, Piotr A. Domanski (2006). Performance of a residential Heat Pump operating in the cooling mode with single faults imposed. NIST.
7. MathWorks. Simscape Product Documentation. 
8. Cengel, Y. A., & Boles, M. A. (2015). Thermodynamics: An Engineering Approach. McGraw-Hill EducationDing.
9. IEA Heat Pump Centre. (2020). Heat Pumps: Technology and Applications.
10. ASHRAE. (2021). ASHRAE Handbook: Fundamentals. 
11. Dossat, R. J., & Horan, T. J. (2002). Principles of Refrigeration (5th ed.). Pearson. 
12. Stoecker, W. F., & Jones, J. W. (1982). Refrigeration and Air Conditioning. McGraw-Hill.
13. Granryd, E. (2011). Refrigerating Engineering. Royal Institute of Technology (KTH). 
14. Recknagel, H., Sprenger, E., & Schramek, E.-R. (2020). Handbook of Heating, Ventilation, and Air Conditioning Technology. Springer. 
15. CoolProp Developers. CoolProp Documentation. 
16. Willmott, C. J., & Matsuura, K. (2005). "Advantages of the Mean Absolute Error (MAE) over the Root Mean Square Error (RMSE) in Assessing Average Model Performance." Climate Research, 30(1), 79-82. 
17. Y. Alkurdi (2024). Anomaly Detection in Heat Pumps: Experimental Setup, Testing, and Data Analysis (Master Thesis).

Related Posts

Fault simulation in heat pumps with propane (part 1)