Monitoring the Performance of an Air-to-Air Heat Pump System

Author

John Ewbank

Introduction

There has been considerable debate about how well air-to-air heatpumps perform. How efficient are they?

To gather empirical data on the performance of an air-to-air heat pump, Trystan Lea from OpenEnergyMonitor kindly provided me with monitoring equipment comprising six temperature probes and two CT sensors (for measuring power).

Monitoring Setup

Sensor Configuration

Temperature probes:

Sensor #	Location	Measurement Purpose
1	Above internal wall unit	Air intake temperature
2	Below internal wall unit	Air outlet temperature
3	Hot gas line (under insulation)	Refrigerant discharge temp
4	Liquid line (under insulation)	Refrigerant return temp
5	External unit front	Ambient air intake
6	External unit rear	Conditioned air exhaust

CT sensors:

Sensor #	Location	Measurement Purpose
1	Wire from RCB to outdoor unit	Total Heatpump Power
1	Wire from outdoor unit to internal unit	Internal wall unit Power

Installing these sensors proved challenging. Time constraints meant completion took place in darkness on a cold, rainy February evening. Splitting electrical cables with cold hands in the dark presented a particular challenge, and the initial installation required some refinement.

Retrofitting monitoring equipment into a space that wasn’t designed to accommodate additional components was problematic. I would recommend that if you are considering installing an air-to-air unit and wish to incorporate monitoring, you should plan for this in advance.

Data Collection and Limitations

Some important notes about the data:

Temperature data for the internal wall unit outlet before 7 February is inaccurate. The sensor became detached on 30 January, and I was unable to reattach it for approximately a week.
Between 1-3 March, the Wi-Fi connection to the monitoring equipment failed due to a router replacement.

There are several constraints to this monitoring project:

As I am not the homeowner and don’t frequently visit the property, making adjustments is difficult.
Since it’s not my heat pump or house, I’m reluctant to implement changes that might affect the unit’s aesthetics or the house exterior.
Limited access means I cannot simply arrange to examine the installation whenever adjustments are required.

Methodology

Calibrating Air Flow

Since I don’t have an anemometer and cannot measure flow in the refrigerant lines directly, I used a simple air mass flow calculation based on the manufacturer’s data and interpolated values.

The process involved first recording the heating airflow figures from the manufacturer’s documentation.

I then cycled through the internal wall unit's fan speeds (silent, 1, 2, 3, 4) whilst monitoring the power draw. The peak power corresponds to fan speed 4 with the air purifier mode active, whilst the minimum is observed at silent mode with the air purifier deactivated. There is another fan speed above 4 - which is only achieved by using a special powerful mode. This doesn’t occur in normal operation.

There is an air purifier fitted to this unit; the homeowner keeps it on and it uses 2W. It activates automatically each day when the unit starts. I’ve included this in the airflow model.

The graph above shows the manufacturer’s airflow data points plotted against the power draw for each wall unit speed. When the unit runs in “AUTO” fan mode, it adjusts in smaller increments than the five standard fan speeds. To estimate airflow versus input power, I fitted a linear model to the data points.

The gradient of the line is used to calculate the airflow at different fan powers. While not 100% accurate, this approach provides a reasonable approximation with the available equipment.

Conversion from Airflow to Mass Flow

I’ve kept the conversion from airflow to mass flow straightforward. Although a humidity sensor monitors the room where the unit is installed, humidity’s effect on mass flow is relatively minor. The conversion is calculated by dividing the airflow in litres per second by 1000, then multiplying by the mass of 1 cubic metre of air at room temperature (approximately 1.2 kg).

Calculating Heat Output and COP

The heat delivered by the heat pump is calculated by:

Finding the temperature difference between the intake and outlet air of the internal wall unit
Multiplying this by the specific heat capacity of air (approximately 1005 J/kg/K)
Multiplying by the mass flow of air

The Coefficient of Performance (COP) is then calculated by dividing the heat output (in watts) by the electrical input (in watts).

While I have allowed for negative COPs in the analysis, this might not fully represent the average COP of the heat pump on cold days. This is because when the unit reverses during the defrost cycles, the air-temperature sensors above and below the wall unit are both influenced by the cold air being pushed into the room. Consequently, both fall, meaning the delta between them might not represent the heat being taken from the room. This is further compounded by the unit closing up during defrost cycles. I have an idea of how I could account for this, but I haven’t implemented it in this analysis.

Data Analysis

The COP (Coefficient of Performance) measurement applies only to specific operational conditions of the heat pump. The data is filtered using several constraints:

The heat pump must be actively running, with total power consumption of at least 60 watts
The internal wall unit power must be 50 watts or less
The COP value must be within a reasonable range between -15 and 10
COP values that are missing (NaN), zero, or otherwise invalid are excluded
The external air temperature (measured after the heat pump) must be 15°C or lower

This last temperature constraint specifically helps filter out periods when the unit was operating in air conditioning mode rather than heating mode, as these operational states would skew the COP analysis for heating performance.

The heat pump at minimum modulation operates at around 90W, though when cycling it can settle just above 60W. There was a period where I installed a CT sensor in reverse briefly, so that was filtered out, as was the period where I had the CT sensors plugged into the incorrect ports.

Results

Average COP: 3.95
Average external temp: 5.28°C
Average indoor temp: 22.12°C

After filtering the data for periods where the heat pump is running and excluding outlier COP values (below -15 or above 10), I calculate a COP between 3.9 and 4.0 from 7 February to 4 March. As mentioned previously, this does not fully account for defrost cycles; however, these cycles are unlikely to substantially lower the average COP as they are relatively infrequent. I would suggest an average COP of around 3.7 to 3.8 when accounting for the defrosts.

It should be noted that the average internal temperature was over 22 degrees while the heat pump was operating. This is higher than many people maintain in their homes. If the internal temperature were set lower, it would be reasonable to assume the heat pump would achieve a higher COP.

Performance Visualisation

The graphs below are interactive, you can click and drag your cursor to zoom in on specific sections.

The heat pump is rated to deliver 4kW of heat to the property. I was pleased to see that this heat input was achieved multiple times. It was encouraging to observe that during cold weather this heat was delivered while the heat pump was drawing around its maximum operational power - while not in the special ‘powerful’ mode. As it means the output roughly conforms to the figures from the specification sheet, which provides some credence to the monitoring accuracy.

This max heat delivered generally occurred when heating up the house in the morning; it typically inputs around half that amount of heat.

During the latter half of the monitoring period, the operation of the heat pump was changed from turning off at night to remaining on. The goal was to take advantage of a lower overnight tariff, while also trying to maximise heat pump COP by running at steady state.

Conclusion

The monitoring data shows that this air-to-air heat pump achieves a respectable COP of approximately 3.9-4.0 (not fully accounting for defrosts) during the monitored winter period. This suggests that the system is delivering heat at roughly four times the efficiency of direct electric heating.

It is unfortunate that installation errors on my part eliminated one week’s worth of data at the beginning of February; however, several cold snaps during the month demonstrated the heat pump performing with a COP of around 3 at minus two degrees Celsius, which is respectable.

While this single data point cannot represent the entire air-to-air heat pump market - as it was a premium unit - it provides encouraging evidence that air-to-air heat pumps are reasonably efficient, and installers can achieve these efficiencies with minimal training.

It appears that the current generation of air-to-air heat pumps may not reach the efficiency of quality air-to-water installs; though there is reason to believe the gap between air-to-air and air-to-water could narrow as residential units are further optimised for heating rather than cooling.

Limitations of the Study

While this monitoring setup provides valuable insights, it has several limitations:

Airflow estimation: Without direct measurement of airflow, manufacturer’s data and power-to-flow relationships are used for modelling.
Temperature sensor placement: The exact positioning of sensors can influence readings, particularly for air temperatures.
Limited monitoring period: Data covers only part of the heating season.
Discontinuities in data: There are gaps in the dataset due to connectivity issues and sensor displacements.
Defrosts not fully accounted for: I have not implemented a full method to account for heat losses during defrost cycles.

Future Improvements

For a more comprehensive assessment, future modelling may include:

Ultrasonic flow sensors: Non-invasive monitoring of refrigerant flow would enable more accurate heat transfer calculations.
Specific Pipe Clamp Thermistors: Thermistors designed to clamp around and measure the temperature of a pipe - rather than the general area.
Manufacturer Data: There are a number of pre-installed temperature sensors on heat pumps, having access to these would provide valuable data.
Extended monitoring period: Collecting data across multiple seasons would provide insight into year-round performance.
Comparative monitoring: Simultaneous monitoring of multiple heat pump types would allow direct comparison.
Coriolis mass flow: To provide the most accurate monitoring of refrigerant, a coriolis mass flow meter could be used.

I hope this study provides a valuable real-world performance benchmark for air-to-air heat pumps in a UK climate. I would like to see monitoring of a low-wall air-to-air unit wall unit, which takes cold air from around the floor area and directs it upwards. This contrasts with the more traditional high-wall wall unit analysed in this work.