The context

A timber processing mill located in NSW Australia
Upgrade/repairs made to the PFC in late June 2024 (Bills report PFC in June as 0.708, July as 0.980)
New solar inverter installed late January 2025
kW = active power (the good kind)
kVAR = reactive power (the bad kind)
kVA = apparent power that is used to calculate demand (the combination of kW and kVAR)

The summary

Approx electrical cost across FY23-34 was ~$166,000
Of the total cost, total demand charges are ~$40,000
The max demand for FY23-24 was 261 kVA on 3rd June 2024, when the PFC was broken
~24% of the cost is in demand charges: a good ratio particularly considering the PFC was broken in June 2024; however this figure is distorted due to the constant base load of the kilns.
Power factor at peak demand is good but not perfect
The PFC unit has a coarse resolution for switching, resulting in saw-toothing of reactive power
Negative reactive power (leading power factor) is common

The load, disaggregated

Daily summary shows significant base load consumption (~1 MWH/day) across the previous FY, with the constant load in orange and the mill operations in blue (in this analysis, variable load):

Demand

Daily peak demand clearly shows weekdays vs weekends. Summer holidays still has demand because the solar PV is exporting power.

The PFC Unit

The PFC unit has been exhibiting the same behaviour regardless of the presence of the solar PV, except for the brief period when it failed. The failure window provides us with a baseline without the PFC to compare with, noting that the kVAR and kW figures are roughly equivalent, showing a PF of ~0.45, which is typical of mills without PFC. Also note the reactive power is always positive with the PFC offline.

Looking at May 2024:

These two charts track the PFC failure during June 2024, from not working on the 17th & 18th, to switching on all the capacitors in the PFC on the 27th June. Then, it was working for the first few days in July, then 3 days of failure before it is fixed on the 8th of July. You can see the impact this has upon kVA.

With the PFC operating, the kVAR figures are roughly 30% that of the KW as can be seen in subsequent data (green line above yellow). However, you can also see that the values of the capacitors are too coarse, indicating minimum 50 kVAR capacitors (5 steps, total 250 kVAR). This results in the PFC generating a negative (leading) power factor and switching back to positive (lagging) power factor continually.

The result of the coarse PFC switching (aside from negative kVAR) is that the max demand event isn't caught by 100% the PFC, as you can see below. There is also a delay with PFC in switching capacitors on and off. This is why the power factor at peak demand was 0.95 rather than the ideal number of 1. Reactive power should be at 0... in an ideal world. This is what the PFC should be doing, within some margin of error.

Prior to 2022, utility data was measured in 15 minute intervals: now it is in 5 minute intervals. Where speed of the PFC wasn't a big factor before, it is now. Looking at the max demand event in May 2025. Also, the kVAR should never be negative (this is against the supply rules), where we see long durations of negative KVAR at the mill. This and the coarse resolution of capacitors impacts the power factor:

However, all this said, the important thing is that peak demand is constrained, and the PFC is doing this quite well. Could it do better? Yes. But what would it save? Somewhere in the region of $400/month for demand charges: not a huge number in the grand scheme of things; but it could be an easy fix, potentially, by using the PV inverter (see below).

That said, the PFC is a single point of failure as you experienced roughly a year ago. And when there is a single failure (for even 5 minutes) within a month, you end up paying ~$1,500 more for that month as a penalty: typically the mill's monthly demand charges are circa $3,500, but in June of 2024 you paid a little less than $5,000. A more decentralised approach could mitigate this risk and provide better overall power quality.

The other consideration is that a negative power factor can have negative impacts on motors, leading to reduced efficiency. A negative power factor is as bad, if not worse, than a positive power factor.

The Solar PV

Strangely we couldn't find any evidence that the solar stopped operating in the last 2 years. The vast quantity of generated electricity is consumed by the constant load present on the site, except when (presumably) the kilns are not operating over summer holidays.

There are periods of time when the solar PV is sending KW to the grid, and pulling KVAR from the grid, for example on the 6th, 9th (see above) or 23 & 24th May (see below). This is simultaneous import and export of power to the grid, which shouldn't really be happening, and is likely is a conflicting interaction between the coarse switching PFC and the solar PV: when the PFC sees generation, it turns off capacitors to compensate (as it sees more active power). Wrong move!

Ideally, the solar system should be reading the power factor at the MSB and doing something pro-active about it. The PFC should be 'aware' of what the solar PV is doing. Also, the PV inverters have the capability to be reprogrammed to 'fill-in' the gap between the 50KVAR switching of the PFC unit while generating power. Or, additional electronics can be added to do a similar task on a 24/7 basis.

Given the mill is not being paid to put electricity back to the grid, it would probably make sense to export limit to zero, and reserve the inverter for self-consumption only. Why strain the inverter electronics (during summer, in particular) for no financial benefit?

Conclusion

This was a longer report than we usually do, but there is more to talk about with respects the PFC than usual (because, typically, they aren't working in the first place). What is interesting is the PFC and solar inverter being at odds with one another as you suspected. Definitely some room for improvement, but the mill is far above the benchmark out there.

We'd recommend:

check if there are sensors (CTs) for the PFC to detect solar generation
explore other (more efficient) fan options for the kilns (or reserve this for an audit)
log the mill for a week using a detailed power quality meter to establish a baseline (see photo below)
ensure there are no dangerous system harmonics present, and if there are, what mitigation should be taken
explore reprogramming of the solar inverter for zero export and KVAR compensation
based on power quality logging, determine if there is a case for an energy audit, focusing on overall savings rather than purely demand (particularly the kilns
validate any savings, and see if ESCs can be generated, or other subsidies can help pay for upgrades

We can assist you with the above, independently assessing the state of the electrical & control systems, and working with your electricians and solar installer to improve the overall state of the infrastructure that underpins your operation.

What we found:

Amongst other things, with respects the solar PV, we found

100kW Solar PV inverter with 180A breaker instead of recommended 250A
PV inverter connected to existing distribution board, 100m from MSB
Existing DB fed workshop (including welding equipment) and kiln fans
Sub-main to this board was 4x 25mm2 XLPE
Recommended cable size for solar at that distance and size should be 120mm2
By connecting via the existing distribution board, the solar PV was connected to the load side of the PFC CT
The PFC controller was not aware of reverse-flow of electricity
No gate meter or data connection to the utility meter was installed for export limitation capability

Recent Posts