A much better solution for solar self-consumption


1. Introduction

2. Basic Idea and Advantages

3. Details

4. Cost Calculation

5. Options, Variants, Improvements

6. Alternative for Direct DC Consumption

1. Introduction

As a late outcome of Edgeryders’ OpenVillage House prototype in Sidi Kaouki, Morocco, I discovered that there is no (!) readily available and economically meaningful, commercially made solution for self-consumption of the electricity from small (<1 kWpeak) photovoltaic installations. For larger installations, there are string inverters with a zero-export feature, but they are expensive.

Inspired by this, I developed the idea below that solves both issues. Initially for the house in Sidi Kaouki, but I think there is a worldwide market for it in all development countries where utility companies forbid or do not reward exporting locally generated electricity to the grid. If they forbid it, the idea below is the only option. If they do not reward it, the idea below is the most economical option. (See here for the different ways how utility companies treat feed-in.) In total, the product might revolutionize access to energy and utilization of rooftop solar panels – because without needing a grid-tied inverter nor a battery bank, solar energy will be the cheapest option for many grid-connected households.

2. Basic Idea and Advantages

A small-ish supercapacitor bank is recharged by photovoltaics panels and the house runs either on the grid, or from that supercap bank via a cheap, not grid-tied inverter. An automatic transfer switch switches between the two options seamlessly every few minutes, whenever the supercap bank is discharged.

This setup has a range of advantages over all alternatives, including that of a battery system in off-grid or hybrid homes:

  • Cheapest purchase price. With respect to the purchase price, it would be by far the cheapest variant for small installations <2 kWpeak, because it contains neither batteries nor a grid-tied inverter. These are the most expensive components in the alternative systems.

    Grid-tie inverters are so expensive probably due to the extra safety features and certification costs of grid-connected generation equipment. Inverters in true off-grid homes are expensive because they also need to handle the peak power draw, while in our case we can use a smaller inverter and in case of overload quickly transfer-switch to the grid. Batteries are expensive because a large capacity is needed; in cases like here not to enable high loads (which are also possible with lower capacities, e.g. automotive starter batteries) but to enable high enough charging currents by the solar panels even when the batteries are 80-95% full (which is needed as deep cycling reduces their lifetime).

  • Cheapest to maintain and repair. With respect to maintenance costs, the system will also be much cheaper than current alternatives. It does not contain a battery, so does not incur regular major replacement costs; instead it uses supercapacitors, which do not degrade noticeably. Also the whole system consists of individual modules that can be exchanged individually, and each such module / part is a very standard off-the-shelf component available from multiple manufacturers. The biggest cost factor will be the not grid-tied inverter, and even that is <50 EUR for the smallest size.

  • Easily made in many sizes. Grid-tied inverters are not available in sizes less than 250 W, probably because the fixed costs of safety features and certification would make smaller sizes unreasonably expensive. The idea outlined below can be built economically in smaller sizes. (Even though smaller inverters than 300 W are hard to find and not that meaningful as even a 300 W inverter is only 20 EUR).

  • Scales on-site. People can start with a small system size and expand it later, as they have money. This is very suitable for low-income markets. Technically, this is possible because the system needs no charge controller. Simply add more photovoltaics panels, supercaps and (for installation sizes >2 kWpeak) more inverters in parallel to the existing one. Transfer switches are sized for the inverter and added as inverters are added, so so not have to be exchanged. Only the solar panel disconnect relay should be oversized in the beginning, to match the possible peak load from future additions of solar panels. It should be made for 30 A, thus able to switch all panels that can connect to a single MC4 connector (which are rated either 20 A or 30 A depending on manufacturer). If even more panels are added, simply a second disconnect relay will be added as well.

  • No vendor lock-in. It only uses readily available off-the-shelf components available from many manufacturers. This is way better than vendor lock-in for a highly specialized “micro inverter with zero export” or “zero export device”, as these products might be discontinued, their manufacturer might go out of business or provide poor support in cases of issues or warranty.

  • No permission needed. It can be used even where all grid-tied inverters are forbidden by the utility company, including those with a zero-export function. Because this system only connects electrical consumers to the grid, and that is what the grid is for. (Usually, when everything is forbidden the reason given is the safety of line workers, even though that should be covered by the anti-islanding function of the grid-tie inverter.)

  • DIY friendly. No permission from the utility company is needed, all components are off-the-shelf products and the only work is connecting them and putting them into a case. This makes development and production very do-it-yourself, enabling anything from micro-enterprises producing these items to people building them for themselves in their villages.

3. Details

Loads are by default connected to an inverter that is fed by the supercaps bank, which is fed in turn by both the solar panels.

A description of the main components of this system:

  • Microcontroller. An off-the shelf microcontroller (e.g. Arduino) or single board computer (e.g. Raspberry Pi) that controls all other components.

  • Inverter. An inverter that accepts 24 V DC input (usually meaning a 20-30 V DC range) and outputs 230 V AC. By not being certified for grid-tied operation, it is much cheaper. A model with pure sine wave output is recommended (ca. 150 EUR for a model with 1000 W continuous power output), but those cheaper ones with modified sine wave output can also be used (see chapter “Options, Variants, Improvements”).

  • Supercapacitors. The accumulate solar energy until the loads are switched over to use it up again. They can endure about one million charge cycles, so frequent cycling is ok. This allows to make them quite small: the supercap bank can be for example 10 Wh of usable capacity, and that should be enough for a small house. At the max. load the inverter can handle (say 1000 W, somewhat above baseload) and negligible PV input, this would equate to a worst-case discharge time of (10 Wh * 3600 s/h) / 1000 W = 36 s. Since inverters usually only work within the limits of 10-15 V or 20-30 V input, we cannot use the full energy in the supercaps but only about half. Still, that would mean 20 Wh capacity, or about 200 EUR of costs.

  • Transfer switch. A standalone, programmable automatic transfer switch (ATS) with open transfer, which means it first disconnects the load from the inverter before connecting it to the grid, or vice versa. Since our system contains a microcontroller that can do the automation part, the ATS would be a simple changeover relay with open transfer, with its default (de-energized) position connecting the loads to the grid. (Instead of a changeover relay, combining two single relays and switching them with a short delay would also be possible, but a bad idea – it is not fail safe in case of a controller software bug that would connect both relays, causing unsynchronized and anyway forbidden backfeeding to the grid.) The switchover time of an ATS is so fast (10-100 ms) that consumers are not affected – even computers will not be switched off by the transfer.

  • No solar charge controller needed! The supercaps bank needs a balancer but is otherwise very simple to charge when not desiring MPPT to run the solar panels: it only needs to stop when the supercaps are full, to not damage them. Due to this, we do not need a charge controller but can use a simple relay, controlled by our microcontroller. Since we will cycle the supercaps within the voltage range of lead-acid batteries to match the input requirements of the inverter, the driving efficiency of the solar panels would be equivalent to what a PWM charge controller achieves – which is generally acceptable. And since there is no real charge controller, a major cost esp. of scaling is elimited, and there are no limits for oversizing the photovoltaics array either – in contrast to a charge controller, the supercaps will handle more or less any inrush current an array of PV panels can throw at them.

A description of the control algorithm in the microcontroller:

  • While more solar electricity can be produced than needed by the house, the house is permanently connected to the supercaps bank via the inverter.

  • When less energy than needed can be produced, the house will discharge the supercaps bank. When reaching the lower limit of the acceptable voltage range (20 V), the microcontroller will switch all loads in the house over to the grid.

  • At that point, the supercaps will be charged again by the solar panels as there is zero consumption now. When they are full, the loads are switched over to the supercaps bank and the cycle starts anew. One full cycle should last at least 20 s under peak load, perhaps longer, depending on the expectable lifetime of the supercaps and on how much stress switchover puts on electrical equipment.

  • If the inverter becomes overloaded by its loads, the microcontroller switches it immediately over to the grid, to protect the inverter from damage.

4. Cost Calculation

Rough estimate for a size with 500 Wpeak photovoltaics:

  • Photovoltaics panels, 500 Wpeak: 195 EUR (used from Germany, where they are available for 30-40% their original price in large quantities; no problem here as enough identical modules can be purchased for a small 500 Wpeak plant)
  • Changeover relay, 240 V, 16 A: 20 EUR
  • Relay for solar panel disconnect, 50 V, 30 A: 5 EUR
  • Microcontroller: 20 EUR
  • Relay control board, AC current sensor, DC current sensor, DC voltage sensor: 40 EUR
  • inverter, 20-30 V to 240 W, 1000 W pure sine wave: 140 EUR
  • supercaps, 20 Wh: 200 EUR
  • supercaps balancer: 20 EUR
  • cables, connectors, case etc.: 60 EUR
  • sum: 700 EUR (with industrial mass production maybe 450 EUR)

Rough estimate for a size with 50 Wpeak photovoltaics and a modified sine wave inverter, all in all the smallest reasonable variant for households with very low electricity needs:

  • Photovoltaics panels, 50 Wpeak: 50 EUR
  • Changeover relay, 240 V, 16 A: 20 EUR
  • Relay for solar panel disconnect, 50 V, 30 A: 5 EUR
  • Microcontroller: 20 EUR
  • Relay control board, AC current sensor, DC current sensor, DC voltage sensor: 40 EUR
  • inverter, 20-30 V to 240 W, 1000 W modified sine wave: 50 EUR
  • supercaps, 10 Wh: 100 EUR
  • supercaps balancer: 20 EUR
  • cables, connectors, case etc.: 50 EUR
  • sum: 355 EUR (with industrial mass production maybe 200 EUR)

5. Options, Variants, Improvements

  • Without supercaps but with charge controller. This is probably a major improvement because it replaces the rare and expensive components (supercaps) with standard ones (a charge controller and a lead-acid battery or iron-nickel battery). The battery will not (!) by cycled, it is just used to provide a few seconds of power to the inverter when direct input from the sun via the charge controller is not sufficient. This can be a transient condition of high power draw (for example when a fridge starts) but when it lasts longer than a few seconds then the system would switch over to the grid because grid electricity is still cheaper than local battery electricity.

    The charge controller can be small, as it can be dimensioned according to the base load of the house (typically 100-200 W, so a 24 V / 15 A MPPT charge controller is fully sufficient, and quite cheap). Likewise, the lead-acid battery can be small because it will not be cycled. An automotive battery is the cheapest option and fine for this, as it is meant to provide high currents in short peaks, without being deep-cycled. Even a used automotive battery (without much current carrying capacity left) can be utilized here for some time as the system can be configured to also switch over to the grid when the battery cannot supply enough current (means, when its output voltage buckles), even before the normal 3-7 seconds of running loads on battery have passed. In this use of being properly charged more or less all the time, a lead-acid battery can last for 10-20 years, and an iron-nickel battery lasts 30-50 years anyway. The only case when the battery would be cycled is when the grid is down and there is not enough sun to power the loads.

  • Multiple transfer switches. The proposal just above of using a lead-acid battery instead of supercaps and not cycling the battery has the disadvantage that locally produced electricity is not used at all when the total electrical consumption of the house exceeds current photovoltaic electricity generation. That can be fixed by having multiple transfer switches, one for each of the different circuits in the house, mounted inside the house’s switchpanel. Then, circuits that together do not exceed the current power production will still be powered by solar, and only the remaining circuits are switched over to the grid. Since a house usually has 1-2 circuits per room, this easily allows to single out the large loads (vacuum cleaner, powertools etc.) and power them from the grid, while the base load of the house (fridge, freezer, lighting, electronics etc.) can still be powered by the solar panels during the day.

  • Modified sine wave inverters. For low-income markets, significant cost savings are possible by using a modified sine wave (“square wave”) inverter instead of a pure sine wave version, which costs about three times as much. Some but few equipment items may be damaged over time by the high peak voltages (“HF noise”) in square waves, and motors may make annoying sounds (details here and here). Both can be averted by simple and cheap filter circuits. Examples: 1, 2, 3, 4, 5. These circuits have to be adapted to the power consumption of the individual consumers, so it makes sense to sell them as a range of plug adapters that will be kept connected to the devices. Anyway, a “household standard” that requires all devices to somehow accept square wave input is on the long run cheaper and more friendly for DIY-generated AC power.

  • Microgrids. It is posisble to connect the supercap banks of neighboring houses into a SELV microgrid, further reducing dependence on the grid by leveling out excess supply and demand of electricity in the neighborhood. 24 V DC cabling of sufficient diameter can be run over 50-100 m easily, and a DC-DC converter will be used to accept any input voltage for charging ones own supercaps bank.

  • Demand management to utilize the solar electricity fully. The controller would simply switch non-essential 24 V DC devices on and off, powered directly by the supercaps bank as long as on average more electricity is produced than consumed. On average, because the supercaps can average out some differences: it would be ok to run a washing machine for 30 s at 500 W if 250 W excess production recharges the supercaps in the next 60 s again, at which point the cycle would repeat. For the basic options for demand management, see this question).

  • UPS and off-grid functionality. Adding a backup system to prepare for grid outage and even off-rid operation is easily possible. A small generator (fueled by locally produced biogas maybe) that runs occasionally and outputs 30 V DC directly to the supercaps seems better than a battery, as there is no battery degradation and it provides independence from solar conditions. Even batteries would hardly degrade, as they would be hardly in use. But a generator will probably be cheaper than a battery bank covering 3-6 days of low solar irradiation and the increased amount of panels necessary to cover off-grid use in winter.

  • Small MPPT charge controller for low-sun days. As detailed above, charging supercaps from solar panels does not need a charge controller, just a relay to stop the charging at a target voltage. However, it may make economic sense to use a small (undersized) and thus cheap charge controller that can be matched to arbitrarily oversized solar panel arrays (for example some of the Victron BlueEnergy MPPT charge controllers). In high-yield conditions, the charge controller is not used, and not needed b/c there is an excess of PV energy anyway. In low-yield conditions, it is used, and most useful to extract as much as possible from the now-scarce sunlight. It can also have a double use as the DC-DC converter to import energy from neighbors by a cable.

6. Alternative for Direct DC Consumption

Most of the base load of a house, esp. in “developing countries”, consists of LED lighting, computers, TVs and other electronics, all of which internally use DC electricity. So it is possible to reduce ones energy consumption by avoiding the DC → AC → DC conversion losses of the supercaps-and-inverter design from above and instead use the DC-DC converters of the devices, directly connected to the supercap bank. So a wall outlet would provide the 20-30 V DC voltage range of the supercap, and a DC-DC converter is connected to that which provides the constant voltage as needed by the device. All AC devices with an external power adapter can be converted to this system by replacing the power adapter with a DC-DC converter with the same output voltage and at least the same output ampacity.

Now it would be impractical if one would have to change ones DC power adapter for an AC power adapter every evening or when too may people use the DC power during a low-sun day. A practical solution would be to always use 24 V DC for the low-powered devices in the house, and 230 V AC directly from the grid for all high-powered devices and for those that cannot be obtained in 24 V DC variants. When avoiding electric cooking, the latter are rarely used and do not make up a large share of the energy consumption, so trying to power them from solar is probably not worth it.

In this setup, we do not really need the inverter, as the supercaps bank or battery will feed DC consumers directly. But we will need a large AC-DC converter (for example a usual 24 V battery charger) to charge the supercaps bank or battery in parallel if the photovoltaics panels are not bringing in enough power (esp. at night of course). It will be switched on and off by our controller electronics, and it will be ok to do this several times per minute to keep a supercaps bank within its voltage range, or simply to keep it connected in case of a smart charger for a lead-acid battery. So compared to the variant where the supercaps power an inverter, we need 50-75% less supercaps in this case, again decreasing the costs considerably. A 500 W AC-DC converter should be enough, equivalent to a 20 A battery charger for 24 V batteries. Larger households can go to 1000 W.

When selling this variant as a product, a range of DC-DC converters and also good sockets and plugs should be offered in parallel. Still, people are used to AC electricity now and all their devices use that, so the AC based variant from above will be the much more marketable variant. (The DC variant as described here can still be promoted as an add-on or extension option, switching solar panel input over to the direct DC circuit as needed.)


Wow this is amazing. Can this be done in Nepal?

It is amazing. Coming up at the new Reef!

1 Like

Signing you up as customers two and three then :smiley: (Number one is the house in Sidi Kaouki.)

Yes @anu, it can be done in Nepal, all components except supercaps are available locally. In Nepal and everywhere else where feeding into the grid is not rewarded, nearly nobody with access to the electric grid has rooftop solar due to the cost issues of current self-consumption solutions. With this new system, that can change. Would be amazing to see solar panels on every roof in Gorkha (and Kathmandu).


I’d like to run this by my friend Doug Livingston, who consults and teaches modern solar installation to see what he thinks. Using super capacitors in this manner is not something I know about. Is this in use someplace, or did you just invent it?

1 Like

Not done anywhere like this … own idea. Would be great to get some professional feedback, thank you!

Did you get feedback from your solar pro friend already, John? Would be great to have a second opinion about the commercial potential of this design!

I contacted him, he asked me to send him the design paper, I sent it, have not yet heard back.

1 Like

where can I get this? … could I use as a replacement for the battery - or would I need a different technical setup?


You cannot use supercaps as a drop-in replacement for lead-acid batteries as they have very different charging characteristics. If you were lucky enough to own a charge controller that can be configured to charge both lead-acid and supercaps, then yes. But that would be super rare.

If you replace both battery and charge controller, it will work though. Also take care your inverter can accept the input voltage range of the supercaps (which will be larger than what typical batteries output, for example the range could be 24 - 48 V).

As for getting supercaps, you can buy them from electronics suppliers. A good and typical spec would be 2.7 V / 3000 F in size 60×125 mm. A good brand is Maxwell. I’m not sure about the Chinese brands yet – I’d only buy these if I know a reliable supplier and have good independent tests.

If you haven’t worked with supercaps before, figure out how to work with them safely. While their voltage is low and safe, they can put out 2000 A … means, lots of sparks, potentially igniting things etc…