Heat-pump performance: voltage dip/sag, under-voltage and over-voltage

Reverse cycle air-source heat-pumps are an increasingly significant load in New Zealand and in many other countries. This has raised concern over the impact wide-spread use of heat-pumps may have on the grid. The characteristics of the loads connected to the power system are changing because of heat-pumps. Their performance during under-voltage events such as voltage dips has the potential to compound the event and possibly cause voltage collapse. In this study, results from testing six heat-pumps are presented to assess their performance at various voltages and hence their impact on voltage stability.


SS voltage test
The applied voltage was slowly varied from 1.1 per unit (pu) down to the cut-out voltage while the real, reactive and apparent powers, current total harmonic distortion (THDi), power factor (PF) and displacement PF (DPF) were recorded. The results of these tests are tabulated in Tables 2-7 for heating mode and Tables 8-13 for cooling mode (Appendix 1). Figs. 1-4 display this information graphically. Positive reactive power implies a lagging/inductive current. Care must be taken interpreting these results as the electrical power drawn by a heat-pump is determined by many factors. Factors include the indoor and outdoor temperatures, the refrigerant temperatures and phases, as well as the heat/cool setting and temperature set-point. For this reason, it is very difficult to exactly reproduce any given operating condition and there is some variability in these parameters when tests have been repeated.
For instance, a measured drop in input power as the applied voltage is reduced may be because of the reduced voltage, but may be compounded because of changes in refrigerant temperature, phase etc. that are approximately coincident. The cut-out voltages are also influenced to some extent by the power drawn by the unit at that particular instant, but are generally within a few volts of the figures recorded. A marker showing the nominal input power ratings of each unit is displayed in Fig. 1. This shows that these devices may operate at levels significantly different from nominal, even at nominal voltage. Moreover, a marker showing the nominal current at nominal voltage, assuming unity PF is displayed in Fig. 3, demonstrating that nominal parameters are a very poor guide to predicting actual heat-pump behaviour. Despite the variability, overall patterns emerge for each unit. Note that all units continued to operate down to a voltage of 0.7 pu or below.
All units except C are inverter drive types, although their rectifier circuits differ significantly, as reported in [7]. Unit C is a direct-on-line induction motor type. It was noted in [7] that the units classified as types 1 and 2 (A25, B, E and A50), have two distinct modes of operation: at reasonably high input power levels they draw leading current with reasonably low distortion and acceptable PF. At lower input power levels, they draw current with high distortion and poor PF. The change-over point varies between about 400 and 800 W, with some hysteresis, depending on unit type. This is because of a switchable active PF correction (PFC) circuit, which may be trying to maximise efficiency by staying out of circuit at lower power levels. In practice, this makes such units particularly hard to test in a coherent manner.
Running below its nominal rated input power, the A25 unit operated without its PFC circuit being activated during the tests, the power level being 400-500 W in both cycles, accounting for the poor PF results at all voltage levels. The unit's current never exceeded 4 A during these tests, although its current can approach 6 A as shown in [7]. This unit cuts out below about 162 V (0.7 pu) in heating mode and 155 V (0.67 pu) in cooling mode.
Unit A50 ran with approximately constant input power over the heating mode test, with reasonable leading PF and DPF at this power level, down to about 120 V (0.52 pu), at which point the input current was 9.6 A. This is 160% of the measured current drawn at nominal voltage (and 138% of rated current at nominal voltage). This is the highest current recorded for this unit in any of the SS tests. In cooling mode, A50 again continued to operate down to about 120 V (0.52 pu). The PF was reasonable for this unit, although falling off at higher voltages. Unit B behaved in a fairly similar fashion to A50 down to about 127 V (0.55 pu) in heating mode, at which point the input current of 7.4 A was 154% of the current at nominal voltage (and 98% of nominal current at nominal voltage). In cooling mode, unit B also ran down to 127 V (0.55 pu), and on this cycle appeared to be very well behaved, apparently being controlled to a current limit (CL) of about 7.5-8 A (causing power to fall off with falling supply voltage below 0.9 pu, well aligned with nominal current of 7.48 A at nominal voltage) with good (leading) PF under all recorded conditions. In heating mode, unit C operated down to about 156 V (0.68 pu), at which point it was drawing 285% of the current at nominal voltage (and 269% of nominal current at nominal voltage), being about 22 A, and presenting an increasingly inductive load. Between 0.8 and 1.1 pu, it was reasonably well behaved, although a worsening of PF and DPF is noted at high supply voltage. In cooling mode, unit C operated down to about 138 V (0.6 pu), at which point it was drawing 185% of the current at nominal voltage, being about 11 A. Between 0.7 and 1.1 pu, it was reasonably well behaved, although a slight worsening of PF and DPF is noted at high supply voltages. Under different operating conditions (e.g. higher power), it may again show the same highly inductive low supply voltage behaviour as in the heating case.
In heating mode, unit D continued to operate down to an extremely low voltage of 56 V (0.24 pu), at which point the input current of 16.9 A was 163% of the measured current at nominal voltage (and 229% of nominal current at nominal voltage). The PF and DPF are reasonable down to about 0.5 pu, but get worst below this and change from leading to lagging below about 0.75 pu. In cooling mode, unit D again operated down to 56 V (0.24 pu). At 0.3 pu, the input current, at 16.5 A, was 330% of the current at nominal voltage (again 229% of nominal current at nominal voltage). The PF and DPF are reasonable between 0.45 and 1 pu, but they get worst outside these limits. (Repeated tests in Figs. 5 and 7 show that this unit is probably controlled to a CL of about 16.5 A. This, in conjunction with its ability to operate below 0.3 pu, indicates that its control circuitry may be set up for 115 V supply.) In both heating and cooling modes, unit E appears to behave well, apparently operating with a CL of about 6.5 A, well aligned with its nominal current of 6-7 A at nominal voltage. (Again this is very difficult to state categorically without very extensive testing in controlled ambient conditions, to obtain repeatability.) Nevertheless its PF and DPF deteriorate at high supply voltage.
It was found that the cut-in voltage, as the supply voltage was increased again, was approximately the same as the cut-out voltage for all models, but that there was a time delay in starting.
The variability in operating point is demonstrated in Figs. 5-8 where the heat-pumps are tested on different days with whatever the ambient outside temperature was on that day (uncontrolled). Hence, the operating state will be different because of the differing temperatures. The most dramatic difference is the peaks at 0.95 and 0.6 pu on two different runs with the A50 and the absence of such peaks on the third (see Fig. 6). Figs. 6-8 clearly demonstrate a sample of the many different operating regions the heat-pumps can be working in (when the outside and inside temperatures are not controlled).

Switch-on inrush current
The typical inrush current was measured for each of the six heatpumps when switched on at nominal supply voltage. Note that a slightly different result will be obtained each time as the transients are never absolutely identical. The root mean square (RMS) current magnitude, in amperes, is plotted against historical time, in minutes, in Figs. 9-14. Although the details are slightly different, the inrush currents for heating and cooling modes are similar. Note that the electrical supply was already connected to each heat-pump, the switch on transient being activated by the remote control unit.
All of the units draw a spike of current at the instant of switch on, with the exception of unit D. These spikes are of no greater amplitude than the normal running current, except for unit C, the direct-on-line induction motor model, which draws a major spike of nearly 40 A. Some of the inverter drive units draw a second current spike a little later, in some instances shortly before the main motor current starts. Analysis of the heat-pump electrical circuits [7] shows that, for the inverter drive models, some of the spikes are almost certainly caused by charging of the inverter's direct current (DC) bus capacitors before the inverter starts switching. In the case of units displaying two spikes, the earlier one is likely to be because of charging of DC bus capacitors on power rails supplying some of the ancillary circuitry. However, some of the units may keep one or more ancillary rail powered up during standby operation, in which case inrush current to such a rail would only be observed on supply voltage connection. Unit D seems to avoid an initial spike altogether, probably by means of a relay and its boost converter rectifier, which is capable of charging the inverter bus capacitors with a controlled current. In general, any type 2 or 3 heat-pump [7] should be capable of avoiding inrush current into the inverter DC bus. In all inverter drive cases, the magnitude of inrush into the DC bus capacitors is in any case limited by the PFC inductance [7]. The magnitude of the inrush current spikes will also vary with the instantaneous voltage at turn-on, unless a zero-crossing detection scheme is used. There is perhaps room for further investigation in this area.
In the case of unit C, the much larger current spike observed is a combination of stalled rotor current and magnetising inrush current for the direct-on-line induction motor [11].
After a variable delay time, typically about 2-3 min, depending on model, indoor/outdoor temperatures, refrigerant state etc., the inverter-based heat-pumps ramp up to operating power in a more or less controlled fashion. Unit C's outdoor unit (in which the compressor motor resides) begins operation immediately. As can be seen, units D and E have the most benign start-up behaviour.

Voltage dip/sag performance, low-impedance system fault (heating)
To assess the transient performance of the heat-pumps, they were subjected to the voltage dip/sag shown in Fig. 15. This is a representative voltage profile for a low-impedance fault on the New Zealand system and was supplied by the national grid company, Transpower NZ Ltd. This voltage profile was emulated by programming a Chroma AC voltage source to produce the piecewise linear approximation to this waveform shown in Fig. 16 (with expanded view in Fig. 17), in which actual voltage (230 V nominal system voltage) is shown against historical time in seconds. All six heatpumps, while running on the heating cycle, were subjected to this voltage transient and their current waveforms captured (the voltage waveform is the same in all cases, although the trigger point varies in time). Note that only the expanded current traces Inrush current for A25 a Heating mode (initial peak: 5 A; run: 3.2 A) b Cooling mode (initial peak: 5 A; run: 0.2 A (zero cooling power))

Fig. 10
Inrush current for A50 a Heating mode (initial peak: 3.2 A; run: 5.2 A) b Cooling mode (initial peak: 3.3 A; run: 3 A, then 2.3 A) Fig. 11 Unit B inrush current a Heating mode (initial peak: 7 A; run: 5.5 A, then 2.4 A) b Cooling mode (initial peak: 5.5 A, then 7 A; run: 2.2 A) Fig. 12 Unit C inrush current a Heating mode (initial peak: 38.6 A, duration 100 ms; run: 9.4 A) b Cooling mode (initial peak: 38.9 A, duration 102 ms; run: 5.2 A) Fig. 13 Inrush current for unit D a Heating mode (initial peak: N/A; run: 8.2 A) b Cooling mode (initial peak: N/A, then 5.5 A; run: 3.2 A) Fig. 14 Inrush current for unit E a Heating mode (initial peak: 4.2 A; run: 6.2 A) b Cooling mode (initial peak: 6.2 A (not shown); run: 3.9 A) are shown, as there is nothing worthy of note in the trailing part of the transient current waveforms.
The recorded waveforms are displayed in Figs. 18-22, in which the heat-pump's RMS current in amperes is shown against historical time in seconds. Also shown below each figure are the nominal rated current, I, at nominal rated voltage, V (230 V) and the approximate SS CL (I_lim) observed. Units A25 and E behave in the most benign fashion, as they do not draw a spike of current as the voltage falls, despite the fact that they are both operating below nominal current at nominal voltage and certainly below any SS CL observed earlier. All the other models tested make some attempt to maintain their input power during the voltage dip, resulting in a significant current spike higher than their nominal current at nominal voltage, although only unit B exceeds its observed SS CL. Note that all the inverter-based units return to a waiting state when the supply voltage comes back up and will not draw significant current again for typically 2-3 min. The non-inverter model (Unit C) could not be tested as the Chroma source could not provide the inrush current necessary to start it (nor could two Chroma sources in parallel), indicating the unit was trying to draw an extremely large current.

Voltage dip/sag performance, low-impedance system fault (cooling)
The voltage dip/sag tests were repeated with the heat-pumps in cooling mode. The recorded waveforms are displayed in Figs. 23-27, with the same format as in the heating mode.
Recorded behaviour is very similar to the heating case, although this time inrush spikes for both units A50 and B exceed their observed SS CLs. Unit C's large starting current again prevented this test from being carried out with the Chroma sources.

Voltage dip/sag performance, high-impedance system fault (heating)
One concern was whether the heat-pump D, with its low cut-out voltage, would ride through a high-impedance fault event and try to maintain constant power (CP). The representative high-impedance

Discussion
It is not possible to control input power directly, as this is determined by the indoor and outdoor temperatures (and humidity/ thermal capacity of the air) and the state of the refrigerant. Hence, it requires many repeated tests to build up an overall picture of each heat-pump's full range of operating behaviour.

SS system voltage tests
In the SS tests, all the units generally met the harmonic distortion requirements of AS/NZS 61000-3-2 Class A, at nominal voltage [7].
As the system voltage deviates from nominal, both above and below 230 V, the harmonic content of the current drawn varies significantly, resulting in the changes of PF shown in Fig. 4. Harmonic content also varies with input power and is affected by the existing system voltage distortion [7].
Analysis of results obtained indicates that all units have a reasonably well-defined under-voltage cut-out (and cut-in) level. However, for many of the models this is set at a surprisingly low voltage. This can be of concern, because, depending on actual power drawn in the present operating conditions, the units typically operate in CP mode; hence, lower system voltage leads to higher operating current. To illustrate this, Fig. 36 shows the approximate aggregate effect on a system in which one of each of the six units is running in heating mode. In practice, the balance of different units would lead to different results, but in all cases the current-voltage relationship is far from linear and generally leads to increased load current as system voltage falls to 0.7 pu. Fortunately, maximum current in SS operation appears to be limited for all inverter-driven units. However, for at least one model this is set at a surprisingly high level. An investigation into the effects on the system of the aggregated current spikes drawn during a system transient has not been carried out, but may be worthy of further study.

Switch-on inrush current
As expected, the non-inverter model draws a large inrush current of nearly five times its nominal value on start-up in either cycle. All the inverter-driven units draw inrush currents of the same order of magnitude as their nominal currents in either cycle.

Voltage dip/sag transient tests
The non-inverter model could not be subjected to the system transient tests because of its large inrush current, which could not be supplied by the electronic AC voltage source, even with a second voltage source connected in parallel.
During the low-impedance fault transient, three of the five inverter-driven models drew substantial spikes of current during the collapse of the system voltage, as they tried to keep up their input power, in both heat and cool cycles. These current spikes could be of greater magnitude than the observed SS CL, albeit of relatively short duration (typically < 3 cycles). The A25 and the E unit were notable in that they did not draw a current spike, but retired gracefully in both heating and cooling cycles. All five units fell to low power operation after the transient and resumed their previous operating conditions after their normal start-up delay (typically 1-3 min).
A voltage dip/sag transient for a high-impedance fault was applied to heat-pump D, as SS tests had indicated that its cut-out voltage was only slightly above the minimum voltage of the lowimpedance fault transient. This confirmed the unit would stay operational provided the peak current did not exceed the heat-pump's instantaneous over-CL, which appears to be about 25 A (distinct from the SS CL of 16.5 A).

Conclusions
Six heat-pumps available on the NZ market have been tested to determine their electrical behaviour under both prolonged (SS) voltage sag and swell and transient conditions in both the heating and cooling cycles. Five of these units have inverter-driven threephase compressor motors. The inverters run from a DC bus, supplied by a single-phase rectifier, as described in [7]. The sixth unit has a direct-on-line single-phase induction motor (with a capacitor-run auxiliary winding).
Overall, each of the heat-pumps has certain shortcomings with respect to its electrical behaviour, with each exhibiting some good and some poor features, as summarised in Table 1.

Acknowledgments
The financial support for this research from Transpower New Zealand Ltd, the New Zealand Electricity Engineers' Association and the New Zealand Foundation for Research in Science and Technology (now MBIE) is gratefully acknowledged. The authors would also like to thank Ken Smart (University of Canterbury) Fig. 35          Unit E could not be made to draw rated power on the cooling cycle (possibly because of an internal fault) although on heating mode rated power could be achieved.