My Home Photovoltaic (PV) Power System

Latest addition: A day in the life of my PV system

I've long been fascinated by electric power systems, especially ones that individuals can own and operate. During the annual Field Day ham radio contest that stresses independent sources of electric power, I always seemed to have more fun playing with the generators than actually operating the radios.

So when a friend and colleague, Mike Brock (WB6HHV), started a photovoltaic (PV) power system on his house, I got interested in doing one on mine. It helps that California has abundant sunlight and some of the most PV-friendly laws and regulations in the country. These include a net-metering law that requires the electric company to buy my surplus electricity at the full retail price, and a state-funded buy-down program that subsidizes part of the cost of such a "grid tied" PV system.

A "grid tied" PV system operates in conjunction with a conventional electric utility feed. If the PV system produces more power than needed to operate local loads, the excess is "sold" to the utility, running the electric meter backwards and generating a billing credit. When the local loads are greater than the power generated by the PV system, e.g., at night or when large loads are being operated, the meter runs forward.

PV technology is maturing rapidly, though the cost is still not low enough to compete directly with utility electric rates, even with the government subsidies. So I maintain no illusions that I'm doing this to save money. Nor am I one of those loony Y2K survivalists. I'm doing it because of a long-standing interest in the technology. I also get some side benefits, such as a UPS (uninterruptable power supply) that can run my computers for hours in the event of a power failure.

Solar Panels

Here you can see the twelve Astropower AP1206 solar panels on my roof. They were installed by Carlson's Solar of Hemet, California. They are configured as three strings of four panels each. Each panel consists of 36 cells producing a nominal 12VDC under load, so each string produces a nominal 48V (the no-load voltage is considerably higher, about 80V).

The roof on which the panels are mounted slopes to the south-southwest at about 20 degrees. Ideally they would face due south, but that would have required more complex mounting brackets. Besides, it is frequently cloudy in the early mornings here in San Diego, with the clouds burning off by late morning. A more westerly orientation favors the afternoon when it is more likely to be sunny. It also favors production when electric rates are at their highest (more about this later).

The panel cabling is 10-gauge 2-conductor Type TC cable, moisture and sunlight resistant. The cables penetrate the roof in a conventional weatherhead used for utility service entrances. The cabling goes in 1" flex conduit to a Trace TCB10 PV Combiner Box in the attic. Three #6 wires (DC +, DC- and ground) run from the combiner box in 1" flex to the garage.

System Electronics

The array power enters the back wall of my garage, where I installed a Trace SW4048 sinewave inverter, a Trace DC disconnect, a 48 volt 220 amp-hour battery bank using Trojan T-105 commodity golf cart batteries, and a Crusing Equipment Company E-meter battery monitor. A conventional AC subpanel on the inverter output feeds branch circuits around the house with various "protected" loads, such as my computer and networking gear and my VCR (so I won't lose my programs in the event of a power failure!)

Here you can see the back wall of my garage. The conduit from the arrays is at the right side of the photo; it enters the DC breaker panel (the upright rectangular white box with a horizontal green stripe). After passing through circuit breakers, the array power flows to a Trace Photovoltaic Ground Fault Protector , an overpriced and largely useless device that is nonetheless required by section 690-5 of the National Electrical Code. The PVGFP is mounted in the grey box at the lower right.

After passing through the PVGFP, the array power flows through a relay in the PVGFP mounting box. This relay opens whenever the battery voltage exceeds a programmed level.

Under normal operation, the relay is always closed as any excess power from the arrays is automatically sold back to the utility. This keeps the battery voltage from ever reaching the point that would open the relay. But this relay is important to protect my batteries in the event the grid is not available as a "diversion load". E.g., the inverter could fail, a circuit breaker could open, or the grid could go down.

The relay does lack two features of a solid state controller: a multi-stage battery charging program, and a nighttime cutout. Because my system is grid tied, the multistage battery charger program is not very useful to me. My batteries are normally fully charged, and if they aren't (e.g., after a power failure) the Trace inverter already has a multistage battery charger. The lack of a nighttime dropout means that unless I open the array breaker at sundown, I'll get about 200 mA (about 10 watts) of backfeed from the battery into the panels at night. I could stop this with diodes on the panel feeds, but the power loss is so small that I would probably lose more during the day in the voltage drop across the diodes than I'd save at night. The relay coil itself draws another 2-3 watts.

The batteries are in the brown box at the bottom center of the photo. I made the box out of 3/4" MDF (medium density fiberboard). The top slopes up toward the back to encourage hydrogen gas produced by the batteries to flow into the 3" vent line connected to the left rear corner of the top of the box. This line carries the hydrogen to just below a turbine vent in the roof of my garage. (Personally, I believe hydrogen in a garage is a lot less dangerous than ordinary gasoline; while gasoline vapors collect and linger near the floor, hydrogen rapidly dissipates. But the inspector made me do it anyway.)

The battery is connected back through the DC breaker box through a big 175A circuit breaker to the Trace inverter, the big white box just above the DC breaker panel. Because the inverter can carry substantial power, the cables here are quite heavy (2/0).

The two grey boxes at the top center of the picture are conventional AC subpanels. The one on the left is an existing garage subpanel I put in when I got my EV1 and I installed the Magnecharger EV charger (on the left side of the picture). The right-hand subpanel is connected to the output of the inverter and supports the special branch circuits for the "protected" loads (protected against power failure, that is).

The black cord hanging from the left side of the inverter is for a generator. In the event of an extended power outage, I can set up a portable generator and use it to supplement the power from the solar arrays to operate the protected loads and to recharge batteries.


Here's a view looking down into the battery box.

The interconnects are 2/0 building wire from Home Depot. It was relatively cheap ($0.85/foot) but a real pain to work with; very stiff and unforgiving. The cables leading out of the box on the right side are 2/0 flexible "boat cable". It is much easier to work with, but it is very expensive ($7.50/foot at Boat/US, a local marine store). The yellow cable is for a battery temperature sensor affixed to the battery on the right rear. The blue wire taps the battery at the mid-point to supply 24VDC for the e-meter; note the in-line fuseholder near the battery terminal. Although tapping a battery like this can theoretically cause the pack to become unbalanced, in practice the drain low enough that the imbalance is is quickly corrected by routine battery equalization.

The bottom of the battery box consists of three 2x6s with gaps between to allow air to circulate up from ventilation holes drilled in the sides of the box. Under the box is a sheet of polyethylene to catch any acid spills. Before putting in the batteries I sprinkled a box of baking soda into the box to help neutralize any spilled acid.

DC wiring

Here's a closer look at the DC breaker panel:

The E-meter is mounted in a 2" knockout on the upper left side. The 175A inverter breaker is in the center of the box. The array comes in on the right side through 1" flexible conduit, and the 2/0 cables to the inverter leave in 2" PVC on the upper right. Note the white tape around one of the inverter cables indicating that it is a grounded neutral. It is connected to a 500A 50mV battery current shunt for the E-meter. To the right of the same taped cable is a negative bus bar. The array negative leads connect here.

The three array hot leads (red) go to 15A DC circuit breakers mounted at the upper right, upper left and lower left. The combined output of these three breakers passes down to the PVGFP mounted in the grey box at the lower right.

PV Ground Fault Protector

This is the ground fault detector. It is nothing more than two mechanically ganged DC circuit breakers. The one on the right trips at 1 amp; the one on the left is actually a switch rated at 100 amps. The 1-amp breaker is shunted by a 50 kohm resistor. When closed, the 1-amp breaker connects the negative DC bus (the white wire, signifying that it is the system neutral) to earth ground (the green wire with a yellow stripe); this is the only place where the DC neutral is grounded. The 100A switch connects the combined array output to the input of the charge controller.

The black device hanging off one of the red wires is a clamp-on DC ammeter measuring array output current.

In the event of a fault between a positive lead (battery or array) to ground in excess of 1 amp, the 1 amp breaker will trip and open the 100A switch with it. This will interrupt the ground fault except for a small amount of current that will continue to flow through the resistor. And when the 100A switch opens, the arrays are isolated from the charge controller.

Why is this pretty much useless? Because one could block any ground fault currents from flowing by merely not grounding the negative lead in the first place! A resistor (such as the 50 kohm unit here) could serve to reference the negative lead to earth ground. Under normal conditions no current would flow through the resistor, so there would be no voltage drop across it. An indicator light could serve to warn of a ground fault if desired, but the system could continue to operate normally.

As far as I can tell, the only function of the 100A switch is to get your attention by disabling the solar panels. The most credible ground fault scenario involves shorts within a panel string to the grounded panel frame, or perhaps a ground fault within the array wiring. Opening the positive array lead at this downstream point does nothing to interrupt the fault current.

The Charge Controller

A conventional PV charge controller is nothing more than a fancy semiconductor power switch between the arrays and the battery. When the battery voltage falls below a predetermined setting, the switch closes to allow the arrays to charge the battery. When the voltage rises above this point, the switch opens to protect the battery against overcharging.

In normal grid-tied operation, any excess power from the arrays is automatically sold to the utility so this switch should be closed at all times. But if the system is for whatever reason unable to sell power, the charge controller will disconnect the arrays. This could happen if the inverter fails, the DC breaker opens, the AC disconnect or circuit breaker opens, or the grid loses power.

For a time I considered getting a peak-power-tracking controller such as that sold by Fire, Wind & Rain. A peak power tracker operates the arrays at whatever voltage produces maximum power, rather than always operating them at the battery voltage. Depending on the battery condition, solar illumination and panel temperature, this could produce up to 20% additional power.

But while I waited for these units to become available, I did some measurements and calculations that told me that a peak power tracker wouldn't do much for me. Here are some notes that explain why.

So I eventually decided on a simple electromechanical relay, controlled by an auxiliary relay in the Trace inverter.

I have a TINI (Tiny InterNet Interface) single-board computer from Dallas Semiconductor that I plan to integrate into my PV system. One of its functions will be to control this relay, closing it at dawn and opening it at dusk.

The Utility Interface

Here we see the main electrical panel on the side of my house. The red signs and their wording were a SDG&E requirement, as was the disconnect switch (grey box in the lower right). They want to be able to disconnect my inverter from the grid and lock it in that state to protect their workers during a power outage when they could otherwise assume that the lines are dead. The Trace inverter has a half dozen different ways to detect a grid outage and disconnect automatically from the grid, but I can't fault SDG&E for being conservative.

The watthour meter is a GE kVS solid-state time-of-use meter that is actually quite sophisticated. It is programmed to implement SDG&E's Time of Use Rate for Households With Electric Vehicles.

Net Metering with a TOU Tariff

With time-of-use metering, the amount I am credited for each kilowatt-hour I sell back to SDG&E is a function of when I produce it. Electric rates are normally highest on sunny summer afternoons when air conditioners are running at full tilt, and this is precisely when photovoltaic systems produce their maximum output.

While I already had a TOU tariff thanks to my EV1, I learned to my dismay that only SDG&E's standard domestic tariff had net metering provisions; their existing TOU tariffs did not provide for net metering. This was disappointing, but not too surprising. After all, the utilities have never been all that enthusiastic about buying (as opposed to selling) electricity at the regular retail rate, so I could see how they'd be even less happy about buying at the even higher peak afternoon TOU rate.

Nevertheless, I read the California net metering statute as requiring the utilities to provide for net metering on all of their tariffs, including TOU. Vince Schwent at the California Energy Commission, who drafted the original net metering statute, agreed once I pointed out that it had been recently amended. I stood my ground with SDG&E, and after some deliberation they agreed to amend their TOU tariffs to provide for net metering. This new tariff went into effect into late August 1999. Finally, the sauce for the goose could become the sauce for the gander!

The pre-July 1999 EV-TOU-2 rate had a very high differential between the on-peak (noon-6pm) and super-off-peak (midnight-5am) summer rates: 32 cents/kWh and 4.2 cents/kWh, respectively. Not only did this very high afternoon rate make the economics of a PV system look better than ever before, it would have made economic sense to charge up my battery bank from the grid at cheap nighttime rates and sell it back to them during the afternoon at the premium peak rate! (This is not actually such a crazy idea; many utilities do the same thing with "pumped storage" plants.)

Electricity Deregulation - My System Pays Off!

On July 1, 1999, SDG&E finished selling off its conventional power plants as required by California's move to electricity deregulation. SDG&E is now a "Regulated Utility Distribution Company" (UDC) that no longer generates its own electricity. They buy it wholesale on the California Power Exchange (PX) at the going market price and tack on a per-kWh markup, regulated by tariff, to cover transmission, local distribution, energy losses and other miscellaneous costs. With a TOU meter, this UDC markup depends on time of day, but the difference is only a few cents/kWh.

The TOU meter is also used to determine the PX component of the total price. For example, the EV-TOU-2 on-peak period is noon to 6pm, so my per kWh energy charge for that period is based on SDG&E's average cost of buying electricity from the PX during the afternoon.

During the summer of 1999, a SDG&E price cap was in effect that hid the wide daily and seasonal swings in PX prices, thwarting my "battery sellback" scheme and decreasing the payback rate of my PV system.

The price cap was lifted in the summer of 2000, so the consumer now sees the PX prices directly. And, just as many people had predicted, the PX prices went through the roof. This happened thanks to hot weather, a shortage of generators and transmission lines, and a lack of meaningful competition in the now-deregulated generating market. The resulting dramatic rise in electric bills quickly became a hot local issue, with many angry protests and avowed refusals to pay bills. As this is written (late July) here are the prices under the EV-TOU-2 tariff:
PX (energy) SDG&E markup Total (cents/kWh)
On-peak (1200-1800) 23.1 8.438 31.5
Off-peak (1800-0000,0500-1200) 11.1 6.183 17.3
super-off-peak (0000-0500) 4.7 5.879 10.6

So not only do the economics of PV power generation suddenly look attractive again, but so does on-peak battery selling!

But the plot continues to thicken. SDG&E has proposed a new set of tariffs that would temporarily lower the UDC component, apparently to comply with a PUC mandate to refund some excess revenues. For the regular (non-TOU) domestic rate, the decrease is about 2.6 cents/kWh. Here are the current and proposed summer numbers for the EV-TOU-2 tariff:
Current Proposed
On-peak 8.438 -2.538
Off-peak 6.183 4.425
Super-off-peak 5.879 5.364

That's right, not only are SDG&E's proposed on-peak rates less than the off-peak rates (exactly backwards from the usual), but the proposed on-peak rate is actually negative! Not only does this again act to level out the day/night differential that encourages peak period battery selling (though this may not be enough to overcome sufficiently high peak PX prices), it also penalizes me for being a net energy producer during the peak period, when every kilowatt counts.

There are, of course, two interpretations of this proposal. One, SDG&E is trying to give us EV drivers a little break when we're forced to charge our cars in the afternoon. Two, SDG&E has been reading my web page and has finally figured out a way to retaliate for my "sauce for the goose" remarks. You decide.

Phil Karn
Original version 9/27/99
last updated 25 April 2006