I am considering a home solar photovoltaic installation, probably 5 kW or so. The usual configuration around here is with a grid-tie inverter and net metering. But when the grid goes down, the inverter shuts off (to avoid back-feeding into the power grid and electrocuting linemen).

We are interested in being able to use the solar when the power goes down (not infrequent in this rural area). According to the solar installer, the only way it is possible to do this is to also install a battery system like Tesla PowerWall. The idea is that when the inverter loses the 60 Hz sinewave from the grid it shuts off; the inverter in the PowerWall supplies this sinewave, so the inverter doesn't shut off, and there's also a mechanism to disconnect from the grid.

That sounds reasonable. But the cost is not. Two PowerWalls would roughy double the cost of the installation. During outages, we would be satisfied to only have power when the sun shines; this would keep food safe, allow us to draw water, etc. Electrically, there is no reason this can't be done without the expensive battery system. All that is necessary is to generate the 60Hz reference for the solar system's inverter, and to disconnect from the grid.

Has anyone been able to manage such an installation? Would it be strictly DIY (perhaps feasible as I'm a electrical engineer)?

  • 2
    Have you considered a small generator? That would also provide a signal for your inverters to lock on to, at a much lower cost.
    – Nate S.
    Commented Jun 19, 2019 at 20:24
  • The only way possible in their catalog of products... And your expectations of cost are highly influenced by the "selling power back to the power company" part of the equation and all the capitalization tricks they play around that, which are off topic for this forum. Commented Jun 19, 2019 at 20:36
  • The batteries are needed for when there is no sun, ie night-time or cloudy days...
    – Solar Mike
    Commented Jun 19, 2019 at 20:51
  • @Harper, all I'm talking about are the installation costs. I understand there's a lot of specsmanship in how quickly you may (or may not) recoup those costs. Commented Jun 19, 2019 at 21:02
  • 1
    @RustyShackleford, agreed, but i mentioned a generator because they're cheap commodity hardware, and it's therefore probably the least expensive way to go in terms of installation costs. There's fuel costs to consider, but since you're only using this occasionally when the grid goes down, and even then you're not drawing much power from it due to the solar system providing the bulk of the power you need, it's probably pretty negligible.
    – Nate S.
    Commented Jun 19, 2019 at 21:08

4 Answers 4


This is an XY problem at the industry level. Imagine some guy says "Design me a dimmer" and you design him an excellent triac dimmer, no big. And then he goes "Cool, can you just tack on a VFD onto this thing?" Facepalm, not in any elegant or cost efficient way.

So it is with solar panels. Everybody and their dog is selling 2007 tech, a solely grid-tied system because it's established, built at scale, and financial models are well established for sharing profits with the finance company, etc. etc. They can build, sell and finance that all day all night. It's the cheapest way to get you in the door and talking about solar. But they are selling system X.

Now, post-Sandy, there's the hue and cry to "bolt on" off-grid capability, in a manner compatible with dumb consumers. They did it, but it's inelegant and pricey, like bolting a VFD on a triac. The interlocking-for-linemen takes extra resources. I'd call it System XX.

For what you want, you want system Y.

And since you presumably don’t want to design every inch of it from scratch, we should grab out of the traditional "off-grid solar" parts bin. That means a few minor concessions.

The inverters are simply too different

A grid-tie inverter needs to follow grid frequencies, take all its input power, and force it all onto the grid by pushing voltage as hard as necessary.

An off-grid inverter needs to generate its own AC frequency, take only enough input power to do the job, and provide only as much power as the loads are drawing.

These are dramatically different roles and one inverter doing both will be a costly engineering challenge. They do exist; there are "hybrid inverters" on the market capable of both. Of course they are pricey and have their own system design requirements, not least batteries. You are probably better off getting cheap commodity COTS inverters for each application.

That said, your "No batteries whatsoever" plan is flawed. There will be cases where you have loads simply too big for the panels themselves to carry, e.g. motors starting up.

An off-grid system, "plus"

This is where I want to raid the parts bins of the off-grid solar people. The only thing that isn't out of their wheelhouse is the grid-tie inverter, but that's no big.

Their systems are violently battery-centric, for obvious reasons. Because of this, there is no viable way to avoid it, but nobody says your battery can't be tiny. Last I checked, they sell used car batteries for $20, there ya go.

Now in the house, we have a main panel with the main breaker. It has 2 breakers in it: The breaker to the grid-tie inverter, and the breaker to the subpanel.

The subpanel is where the magic happens. It has a $23 Siemens or Square D style "generator interlock" which backfeeds two breakers (one at a time). One side it's on utility, the other it's on "generator" (read: off-grid inverter). Every load you ever foresee running off solar goes in this panel. If you put every single load in your house in this subpanel, I won't tell :)

enter image description here

"Normal" operation involves the subpanel interlock in "Utility" mode, the off-grid inverter driving into nothing (shut it off), the charge controller instantly topping up the tiny battery, and diverting all solar power to "dump". Dump then goes into the grid-tie inverter and is sold back to the power company.

"Power out" operation, you throw the interlock to "Gen", power up the off-grid inverter. As the off-grid inverter draws to supply household loads, the charge controller diverts power to it as needed, blocking the rest of available current from the panels (since "dump" is a dead-end). If a motor starts and pulls more than panels can deliver, then we see how good or big your battery is.

And, if experience shows you it's advantageous to have a real battery, you can get one.

Note that in all cases, power flow is one direction only, and nothing ever reverses flow. That is a huge advantage over the "System XX" type setups, which flow power every which way, and need to "bolt on" elaborate, expensive interlocking systems to avoid backfeeding. Here, backfeed protection is a plain interlock, except for the grid-tie inverter which is UL 1741 compliant; you can't find a grid-tie inverter that isn't.

The whole system is cheap, scungy, simple, bottom-tier tech. That's not to say "don't use good stuff" - yes, of course use Morningstar or Midnight Solar charge controllers. But we're using their standard boilerplate gear, not special-purpose exotica.

Of course, if you want to phone up one of the finance-- I mean solar installation companies, they aren't going to be conversant in a simple system like this. They won't want to build a thing like I describe here; they've got a standard package they sell over and over that they know how to build.

As far as panels

Many panels sold today have built-in grid-tie "micro-inverters", so their output is 120/240VAC and they drop out if the grid fails. Cheap way to comply with NEC 690.12 Rapid Shutdown, but totally useless off-line. There are other panels with smarts on board, too. You want plain Jane run-of-the-mill DC output panels, again, the cheapest thing on the market.

There's also the law to contend with. Roofs have a super important job, and I'm no fan of compromising that with a bunch of holes. Many other places on your property would benefit from shade/snow sheltering, so I say "build freestanding racks over your parking spot etc." But if you do roof-mount, you have to comply with 690.12:

  • If your panels are <30V (i.e. common "12 volt"/19V float panels), you need do nothing.
  • If you add a fireman's switch and relays, you can series-stack them to 80V per segment, and series-stack those segments to anything you please, but much over 80V starts putting you out of bounds for cheap, common off-grid charge controllers. Note the relays can be powered by a <30V solar panel :)

All in, or all out

Panels aren't just panels anymore. Many have embedded electronics (like microinverters) to "help out" the installer. As such, there's no halfway anymore. You can't just let these people install what they please and expect to hack it later: you might, but you're relying on luck. So you really are at a courageous crossroads; you have to jump all-in, one way or the other.

  • Accept exactly what they install, as a turnkey system that you won't modify ever, and pay the bill.
  • Take control of the process, do the research, and spec everything yourself soup to nuts. You're responsible for the result, but you also have the flexibility to do it your way, and you get the best prices by far. And no financing.
  • Comments are not for extended discussion; this conversation has been moved to chat.
    – Michael Karas
    Commented Jun 21, 2019 at 3:34
  • One other divot re: 690.12: you can't use any old relays to meet the requirements, you need gear (other than the initiating device) that's listed and labeled specifically for rapid shutdown Commented Jun 22, 2019 at 1:41
  • @ThreePhaseEel Yeah, that's to be expected... 110.3b and all that. rolls eyes I swear they're trying to kill solar, but I just see it as yet another reason not to put solar panels on roofs... Commented Jun 22, 2019 at 1:44
  • @Harper -- it's not even 110.3(b), it's actually an explicit and specific requirement written into 690.12(D) itself. Commented Jun 22, 2019 at 1:52
  • Well, the chat room has been "frozen for inactivity". Anyhow, looking into gear, I came across this for the Morningstar unit 3PE mentions at the beginning of the chat: youtube.com/watch?time_continue=177&v=BNXerUeHk30 The system they describe is VERY similar to what Harper details here. Commented Aug 22, 2019 at 21:32

The batteries are a "buffer"

Solar panel output inherently varies with a variety of factors, even when the sun is out: temperature, shading from obstacles, incidence angle, and more. As a result, any solar power system that can't count on the grid to soak up fluctuations in power output requires some other means to do the same, especially during low-output conditions, and this practically means that you need a battery bank to do the job, since other small-scale energy storage system technologies simply haven't had the time to mature yet.

Inverters aren't all the same, either

Another complication in your plans is that inverters are not 100% interchangeable. In particular, there are three different inverter technologies out there, namely transformer-type, transformerless, and micro, each with its own benefits and drawbacks.

We start at the historical beginning, which is the transformer-type inverter. The first solar inverters in off-grid work were repurposed vehicle (recreational/marine) inverters, designed to run off of deep-cycle lead-acid batteries at a battery bus voltage of 12 or 24 VDC. This required them to use a step-up transformer to both isolate the AC and DC sides from each other and to convert the switched (PWMed or MSWed) battery voltage to the correct mains voltage. As a result of this, these inverters can be used with grounded (earthed) DC systems as well as floating DC systems, and also do not require expensive high-voltage switching components, but make up for that with the cost and weight of the inverter magnetics. This also meant that the early grid-tie inverters were transformer-based as well, as they were battery inverters with UL 1741-compliant anti-islanding controls (and perhaps maximum power point tracking as well) "bolted on". Eventually, functionality such as power switching, battery charging, and multimode operation was added to these, yielding today's hybrid inverters that are the mainstay of modern off-grid work.

As grid-tied systems became more popular, though, folks wanted more efficiency, which required higher string voltages, and high-voltage power switches also became cheaper, which weighed against the use of big, heavy transformers in the system. Enter transformerless string inverters; these, which represent the bulk of purely grid-tie string inverters out there these days, take high-voltage DC (~400V) from the solar panel string or MPPT stage and switch it directly through to the output to produce the mains AC waveform. This gets rid of the transformer, at the cost of requiring high-voltage switching semiconductors, and also at the cost of requiring a floating, high-voltage solar array, since it's not isolated from the AC mains any longer. Furthermore, transformerless inverters require specialized energy storage system hardware instead of being able to integrate with inexpensive COTS batteries, due to the challenges of managing high-voltage battery strings and the DC distribution issues that a field-integrated 400VDC bus poses.

The transformer-type inverter architecture wasn't done, though. Some bright sparks thought to shrink it down to a box that could manage the output of one or two panels and convert it to AC while fitting on the roof, behind the panel. This became known as a microinverter, and became popular as it minimizes the need for specialized DC distribution hardware while allowing all the wiring to run at efficient voltages (often 240VAC), keeping wire size and cost down. These also can perform efficient MPPT on a panel-by-panel basis (there are other ways to do this, using MPPT DC-DC electronics, but they're beyond the scope of this discussion), but have the cost that they, like transformerless inverters, are more difficult to use in off-grid and multimode applications.

What you're after is a multimode solar system

The functionality you are after (export when grid is present, run standby loads when the grid goes down) is called a "multimode" solar system in the Code (and parts of the trade), and there are two basic ways to achieve this. (Think of it as a hybrid of a grid-tied and an off-grid system, by the way.)

One is what is called AC coupling, where the solar panel output is inverted to AC which is then rectified back to DC and stored by a multifuction inverter/charger hooked to a battery bank (Outback Radian, Schneider Conext, Victron MultiPlus) or integrated energy storage system (such as a Powerwall), with a single-throw transfer means integrated into the system at a suitable point to prevent backfeeding of the grid. This has the advantage that it's easier to integrate with existing grid-tied PV hardware, and also allows for a wider variety of inverter architectures in order to avoid some of the difficulties involved with high-voltage DC panel strings, such as PV arc fault protection and 2017 NEC in-array rapid shutdown. The downside, though, is that you get reduced solar->battery efficiency due to multiple conversions, and it also can limit you as to how much solar you can provision without overwhelming the battery charging/ESS system, depending on the details of the chosen topology.

There is an alternative, though, and that is called DC coupling your system. In this case, the solar panels feed a charge controller that feeds a low voltage (<60VDC) DC bus, with batteries, inverters, and DC-DC converters connected to that DC bus. This is more in line with how off-grid systems are designed, and has the downside that it restricts you in your system topology somewhat with regard to what lives at the panels, forcing you to confront issues like rapid shutdown, DC distribution, and PV arc fault protection more or less head-on. However, it offers much more flexibility and efficiency in some ways; you don't have redundant DC-AC-DC conversion losses, and you also can handle a mismatch between grid-tied solar desires and power-outage demands more readily here, provided you're using good charge control hardware. It also is better for driving low-voltage DC loads such as low-voltage lighting and electronics, which while not a factor in a multimode system, leads to this being the predominant way a fully off-grid system is designed.

Not all solar installers are up to this task

Given what we now know about the multiple types of inverters out there and what type of system you're looking at, it's clear what your problem is now. Your installers are "box pushers" with regards to everything downstream of the solar panels, trained to sell and install a given set of boxes (SolarEdge inverters+optimizers and Powerwalls). As a result of this, they don't have the system integration capabilities and know-how to tackle more sophisticated multimode (or off-grid, for that matter!) systems. Given your situation, I recommend at this point that you find an installer that's familiar with off-grid/multimode work; someone who's an Outback or Victron dealer would be a good starting point, as they are the main makers of the inverters used in these types of systems.

What I'd go with if I was being cheap is the automatic version of Harper's approach

If I was trying to do a multimode solar system inexpensively, what would come out would be similar to Harper's diagram, save for two things:

  1. I would not put the whole house on the standby inverter system. This keeps most of the truly colossal loads off the standby system, and allows it to be kept reasonable-sized, especially if you can manage heating your house without having to throw a bunch of electricity at the job. In fact, I would restrict the standby inverter to (perhaps a subset of) dedicated lighting circuits, as well as a few key fixed loads such as the well pump, refrigerator, and (standby/emergency) heating system. Hot water is perhaps an option as well, but you will want to maximize your efficiency here if you do this, which means that a cheapo electric tank heater is not an option, and tankless is out as well.
  2. Instead of using the manual transfer interlock Harper describes, I would use a combination ATS/subpanel to feed the AC standby loads, with the standby inverter taking the place of the "generator" input on the transfer switch. (A Cummins RA112L1 is suitable for this provided you feed it with 12VDC from a DC-DC box off the battery bank so that the transfer controls can be rigged to work right; Kohler makes such a thing as well, but in a NEMA 3R box, which works outdoors, but is hard to use indoors.) Like Harper's approach, this allows the use of any old sine-wave inverter for the standby inverter at the cost of perhaps limiting what you can use for a grid-tie inverter based on your charge controller selection, but it provides automatic changeover functionality as well, for less cost than a hybrid inverter runs for.

Of course, if you aren't crunched for budget, I would look instead at a system based on a hybrid/multimode inverter, as that combines all the inversion and transfer duty into one box, giving you the net effect of Harper's System XX without the inelegant, "bolt on" design. Either way, this allows you to use less-expensive battery technology (such as marine deep cycle, golf cart, or AGM lead-acid batteries), which is another cost benefit over what you are being proposed with the Powerwalls, depending on how much battery cost factors into the equation of course.

  • Thanks @ThreePhaseEel. The powers that be seem to think we should be discussing this in chat, so let's go there. Commented Jun 21, 2019 at 21:22
  • I've adapted both of your suggestions into my diagram. And by the way, for your 12V SCADA controls, you either use battery if you have one, or just use a 12V solar panel for SCADA (since you only care when the sun shines). I, however, am happy to use low voltage DC for everything it can possibly be used for. (Rapid Shutdown exempts panels <30V). Commented Jun 22, 2019 at 1:05

Because even when it isn't dark, there may be clouds.

The amount of power you can get from solar panels depends enormously on the amount of sunlight hitting them. My rough rule-of-thumb is that light clouds reduce the power output to a tenth of the maximum, and heavy clouds reduce it to a hundredth, or less.

If you only have panels pointing one way, there is also the problem that the sun may not be shining on them at any given time, and you are only getting power from the ambient daylight. In such cases, light clouds can even increase the generation.

If you tried to take more power than the panels are generating at any given moment, the inverter would have to shut down. This could be really inconvenient if you were in the middle of doing something.

Adding a battery allows the inverter to ride through such problems. You can keep operating at full power until the battery is flat.

  • Sure. But I thought I was clear that we would be satisfied to have power only when the sun shines. Your ancillary issue is well-taken though, that it could be inconvenient to have the available power fluctuate. Commented Jun 19, 2019 at 21:05
  • Yeah, while I am not downvoting, this does misalign with the question. Commented Jun 19, 2019 at 21:16
  • @RustyShackleford I have made a quick edit.
    – Simon B
    Commented Jun 19, 2019 at 21:44

Not to be a cynic: Any solar powered system that requires batteries does not pay (where grid power is available).

That being said, this seems odd to me. I've always learned that the batteries power the inverter. The batteries are charged by the Solar panels (when available) or by the grid (when solar isn't available).

In a generator installation (where grid power is available), the generator only powers certain "critical" circuits, such as the refrigerator, furnace, well pump, etc. This will reduce your batter power requirements.

  • 2
    There are many variations - e.g., for a generator it can be "just key circuits" but it can also be "whole house including air conditioning". Similarly, solar can be solar->inverter->grid or it can be solar->integrated battery/inverter system -> house and/or grid depending on battery capacity, grid status, house usage, etc. Commented Jun 19, 2019 at 21:26
  • 1
    Depends what you mean by "pay". For instance there's paying the hotel because their house is uninhabitable because too many systems are dependent on mains AC. Commented Jun 19, 2019 at 23:14
  • "pay" in this case, means "I want to save money by using solar and not paying the electric company or a generator"
    – Scottie H
    Commented Jun 20, 2019 at 14:27
  • @ScottieH -- while power from the utility might be cheap, generator power sure isn't! Commented Jun 21, 2019 at 0:12
  • Grid power isnt cheap everywhere. Try living on a sunny island like Hawaii where the rates are 3x higher and you will see solar makes sense. Commented Jun 22, 2019 at 0:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.