A while back, I wrote about a house with 14 kilowatts (kW) of solar panels installed on the southern exposure roof. Voltage and current monitoring at the breaker panel prior to the solar panel installation had shown 10–12-volt (V) deep sags occurring as frequently as every 22 minutes. The source of the sags was traced to a newly installed, 5-ton heating, ventilating and air conditioning (HVAC) unit located on the opposite end of the house from the utility meter and breaker panel. At the time, the concern was the effect of the solar-panel-supplied power to these voltage sags. Would each 0.25-second sag cause the islanding protection circuitry in the panels inverter section to trip offline? Or, would the local power from the solar panels negate part of the sag and reduce the effect from the inrush current when the HVAC unit kicked on? Or neither of the above?

After negotiating with the electric utility company on the rates for excess power return to the grid, the panels were finally energized. A couple months passed before it was warm enough to turn the HVAC system on. We reconnected the same power-quality-monitoring system to find the answer. Turns out, it was “none of the above,” but the results don’t seem consistent with Ohm’s or Kirchhoff’s Laws so far.

Each time the HVAC unit turns on, the 10–12V sag is from additional losses in the wiring to the source from the 200-ampere (A) peak followed by the unit’s 50A running current. The strange part is the sag depth does not change in any observable way if the solar panels are producing power. During peak output, the panels supply the house loads and push more than 5 kW back into the grid. The average root mean square (rms) voltage is more a factor of current loading than any solar factors. It tends to drop as the solar output decreases after 4 p.m., but so does the average current increase.

Figure 1 shows the line-to-line voltage versus current and the repetitive voltage sags. The bottom graph shows the power flow at the panel; positive is power flowing back into the grid; negative is grid-supplied power. Around 3 p.m. on June 12, a thunderstorm rolled through the area. As the cloud cover curtailed the solar-panel output, the switch in the direction of power flow is obvious, but the voltage sags continue at the same depth. The average voltage decreases prior to this time and correlates to the increase in average current levels.

If you look closely at the waveforms and rms plots of the current in figures 2 and 3, you can see something is different. The total peak current drops 50–60A, from 230A to around 170A. In addition, the current imbalance and current waveform asymmetrical shape with respect to the zero axis changes visibly, especially on Phase A current.

So, the quandary comes from Ohm’s Law, which states the voltage is proportional to current times impedance, and Kirchhoff’s Law, which states the sum of the voltage drops around a closed loop should equal zero. If the current level during the 0.25-second inrush during startup is reduced, so should the voltage drop in the source impedance, resulting in a less severe sag. But scouring through the data for the last month and comparing it to data recorded last year shows the same 10–12V sag, regardless of the solar panel output.

If the HVAC is provided with the same voltage level and has the same impedance during startup and steady-state, the current levels should be the same. Somehow, in this case, the HVAC impedance and the solar panel source impedance are in cahoots, trying to disprove those trusted laws. Hopefully, more monitoring will provide more clues and help solve the mystery.

As a side note, the solar panels offset approximately 30 percent of the power over the two-week monitoring period. While the panels are capable of providing 80 percent of the peak power loading of the residence, they only do that starting at 9 a.m. and tapering off around 7 p.m. The 11 a.m.–6 p.m. peak solar output doesn’t correspond with the peak power consumption in the residence, which occurs from 5 p.m.–11 p.m. While the savings are considerable, they are nowhere near break-even.