One of the constraints of solar power is that it is not always available: it is dependent on daylight hours and clear skies. In order to fill these gaps, a storage solution or a backup infrastructure of fossil fuel power plants is required -- a factor that is often ignored when scientists investigate the sustainability of PV systems.
Whether or not to include storage is no longer just an academic question. Driven by better battery technology and the disincentivization of grid-connected solar panels, off-grid solar is about to make a comeback. How sustainable is a solar PV system if energy storage is taken into account?
Picture: Tesla's lithium-ion home storage system.
In the previous article, we have seen that many life cycle analyses (LCAs) of solar PV systems have a positive bias. Most LCAs base their studies on the manufacturing of solar cells in Europe or the USA. However, most panels are now produced in China, where the electric grid is about twice as carbon-intensive and about 50% less energy efficient.  Likewise, most LCAs investigate solar PV systems in regions with a solar insolation typical of the Mediterranean region, while the majority of solar panels have been installed in places with only half as much sunshine.
As a consequence, the embodied greenhouse gas emissions of a kWh of electricity generated by solar PV is two to four times higher than most LCAs indicate. Instead of the oft-cited 30-50 grams of CO2-equivalents per kilowatt-hour of generated electricity (gCO2e/kWh), we calculated that the typical solar PV system installed between 2008 and 2014 produces close to 120 gCO2e/kWh. This makes solar PV only four times less carbon-intensive than conventional grid electricity in most western countries.
However, even this result is overly optimistic. In the previous article, we didn't take into account "one of the potentially largest missing components"  of the usual life cycle analysis of PV systems: the embodied energy of the infrastructure that deals with the intermittency of solar power. Solar insolation varies throughout the day and throughout the season, and of course solar energy is not available after sunset.
Off-grid Solar Power is Back
Until the end of the 1990s, most solar installations were off-grid systems. Excess power during the day was stored in an on-site bank of lead-acid batteries for use during the night and on cloudy days. Today, almost all solar systems are grid-connected. These installations use the grid as if it was a battery, "storing" excess energy during the day for use at night and on cloudy days.
Obviously, this strategy requires a backup of fossil fuel or nuclear power plants that step in when the supply of solar energy is low or nonexistent. To make a fair comparison with conventional grid electricity, including electricity generated by biomass, this "hidden" part of the solar PV system should also be taken into account. However, every single life cycle analyse of a solar PV ignores it. [3, 2].
Until now, whether or not to include backup power or storage systems was mainly an academic question. This might change soon, because off-grid solar is about to make a comeback. Several manufacturers have presented storage systems based on lithium-ion batteries, the technology that also powers our gadgets and electric cars. [4, 5, 6, 7] Lithium-ion batteries are a superior technology compared to the lead-acid batteries commonly used in off-grid solar PV systems: they last longer, are more compact, more efficient, easier to maintain, and comparatively more sustainable.
Lithium-ion batteries are more expensive than lead-acid batteries, but Morgan Stanley's 2014 report on solar energy predicts that the price of storage will come down to $125-$150 per kWh by 2020.  According to the report, this would make solar PV plus battery storage commercially viable in some European countries (Germany, Italy, Portugal, Spain) and across most of the United States. Morgan Stanley expects a lot from electric vehicle manufacturer Tesla, who announced a home storage system for solar power a few days ago (costing $350 per kWh).  Tesla is building a factory in Arizona that will produce as many lithium-ion batteries as there are currently produced by all manufacturers in the world, introducing economies of scale that can push costs further down.
Morgan Stanley expects off-grid solar PV to be commercially viable in some European countries and across most of the USA by 2020
Other factors also come into play when it comes to home storage for PV power. Solar panels have become so much cheaper in recent years that government subsidies and tax credits for grid-connected systems have come under pressure. In many countries, owners of a grid-connected solar PV system have received a fixed price for the surplus electricity they provide to the grid, without having to pay fixed grid rates. These so-called "net metering rules" or "feed-in rates" were recently abolished in several European countries, and are now under pressure in some US states. In its report, Morgan Stanley predicts that, in the coming years, net metering rules and solar tax credits will disappear altogether. 
A 5 kWh lithium-ion battery pack from Powertech Systems.
Utility companies are fighting the incentivisation of PV power succesfully with the argument that solar customers make use of the grid but don't pay for it, raising the costs for non-solar customers.  The irony is that the disincentivization of grid-connected solar panels makes off-grid systems more attractive, and that utilities might be chasing away their customers. If a grid-connected solar customer has to pay fixed grid fees and doesn't receive a good price for his or her excess power, it might become more financially savvy to install a bank of batteries. The more customers do this, the higher the costs will become for the remaining consumers, encouraging more people to adopt off-grid systems. 
Lead-Acid Battery Storage
Being totally independent of the grid might sound attractive to many, but how sustainable is a solar PV system when battery storage is taken into account? Because a life cycle analysis of an off-grid solar system with lithium-ion batteries has not yet been done, we made one ourselves, based on some LCAs of stand-alone solar PV systems with lead-acid battery storage.
One of the most complete studies to date is a 2009 LCA of a 4.2 kW off-grid system in Murcia, Spain. The 35 m2 PV solar array is mounted on a building rooftop and supplies a programmed lighting system with a daily constant load pattern of 13.8 kWh. The solar panels are connected to 24 open lead-acid batteries with a storage capacity of 110.4 kWh, offering three days of autonomy.  The study found an energy payback time of 9.08 years and specific greenhouse gas emissions of 131 gCO2e/kWh, which makes the system twice as energy efficient and 2.5 times less carbon-intensive than conventional grid electricity in Spain (337 gCO2/kWh). Manufacturing the batteries accounts for 45% of the embodied CO2, and 49% of the life cycle energy use of the solar system.
Lead-acid batteries easily double the energy and CO2 payback times of a solar PV system
This doesn't sound too bad, but unfortunately the researchers made some pretty optimistic assumptions. First of all, the results are valid for a solar insolation of 1,932 kWh/m2/yr -- Murcia is one of the sunniest places in Spain. At lower solar insolation, more solar panels would be needed to produce as much electricity, so the embodied energy of the total system will increase. . If we assume a solar insolation of 1,700 kWh/m2/yr, the average in Southern Europe, GHG emissions would increase to 139 gCO2e/kWh. If we assume a solar insolation of 1,000 kWh/m2/yr, the average in Germany, emissions amount to 174 gCO2/kWh.
Secondly, the researchers assume the lifespan of the lead-acid batteries to be 10 years. For the solar panels, they assume a lifetime of 20 years, which means that they included double the amount of batteries in the life cycle analysis. A lifespan of ten years is very optimistic for a lead-acid battery -- a fact that the scientists admit.  Most other LCA's looking at off-grid systems assume a battery life of 3 or 5 years [14, 15]. However, the lifetime of a lead-acid battery depends strongly on use and maintenance. Because of the low load of the system under discussion, a battery lifespan of 10 years is not completely unrealistic.
On the other hand, if the batteries are used for higher loads -- for example, in a common household -- their lifetime would shorten considerably. Because almost 50% of embodied CO2 and life cycle energy use of a PV solar system is due to the batteries alone, the expected lifespan of the 2.4 ton battery pack has a profound effect on the sustainability of the system.
A lead-acid battery system. SuperiorSolar.
If we assume a battery lifespan of 5 instead of 10 years, and keep the other parameters the same, the GHG emissions increase to 198 and 233 gCO2e/kWh for a solar insolation of 1,700 and 1,000 kWh/m2/yr, respectively. In grid-connected solar PV systems, assuming a longer life expectancy for the solar panels improves the sustainability of the system: the embodied energy and CO2 can be spread over a longer period of time. With off-grid systems, this effect is countered by the need for one or more replacements of the batteries.
If we increase the life expectancy of the solar panels from 20 to 30 years, and keep the battery lifespan at 10 years, CO2e emissions per kWh remain more or less the same. However, if we assume a battery lifespan of only 5 years and extend the lifespan of the solar panels to 30 years, GHG emissions would increase to 206 gCO2e/kWh for a solar insolation of 1,700 kWh/m2/yr, and decrease to 232 gCO2e/kWh for a solar insolation of 1,000 kWh/m2/yr.
Made in China
Thirdly, the researchers assume that all components -- PV cells, batteries, electronics -- are made in Spain, while we have seen in the previous article that manufacturing of solar PV systems has moved to China. Spain's electricity grid is 2.7 times less carbon-intensive (337 gCO2/kWh) than China's electric infrastructure (900 gCO2e/kWh), which means that the GHG emissions of all components of our system can be multiplied by 2.7. This results in specific carbon emissions of 353 and 471 gCO2e/kWh for a solar insolation of 1,700 and 1,000 kWh/m2/yr, respectively, which is higher than the carbon-intensity of the Spanish grid. Considering a battery lifespan of 5 instead of 10 years, emissions would rise to 513 and 631 gCO2e/kWh for a solar insolation of 1,700 and 1,000 kWh/m2/yr, respectively.
If solar panels and batteries are produced in China, the CO2-emissions are double those of conventional grid electricity
Although there are some assumptions by the researchers that are less optimistic -- such as a battery recycling rate of only 50% instead of the more commonly assumed +90% -- it's obvious that an off-grid system with lead-acid batteries is not sustainable, and definitely not when the components are manufactured in China. That doesn't make off-grid solar with lead-acid batteries pointless: compared to a diesel generator, a solar PV system with lead-acid batteries is often the better choice, which makes it a good solution for remote areas without access to the power grid. As an alternative for the centralized electricity infrastructure in western countries, however, it makes little sense.
Lithium-ion Battery Storage System
When we replace the lead-acid batteries by lithium-ion batteries, the sustainability of a stand-alone solar PV system improves considerably. At first glance this may seem counter-productive, because it takes more energy to produce 1 kWh of lithium-ion battery storage than it takes to manufacure 1 kWh of lead-acid battery storage. According to the latest LCA's, aimed at electric vehicle storage, the making of a lithium-ion battery requires between 1.4 and 1.87 MJ/wh, [16, 17, 18] while the energy requirements for the manufacture of a lead-acid battery are between 0.87 and 1.19 MJ/Wh. [18, 12]
6.6 kWh Energy management and lithium-ion storage solution in one. Bosch Power Tec.
Despite this, the higher overall performance of the lithium-ion battery means that considerably less storage is required. For a prolonged lifetime, lead-acid batteries demand a limited "Depth of Discharge" (DoD). If a lead-acid battery is fully discharged (DoD of 100%) its lifespan becomes very short (300 to 800 cycles, or roughly one to two years, depending on battery chemistry). The lifespan increases to between 400 and 1,000 cycles (1-3 years, assuming 365 cycles per year) at a DoD of 80%, and to between 900 and 2,000 cycles (2.5-5.5 years) at a DoD of 33%. . This means that, in order to get a decent lifespan, a lead-acid battery system should be oversized. For example, three times more battery capacity is needed at a DoD of 33%, because two thirds of the battery capacity cannot be used.
Although the lifespan of a lithium-ion battery also decreases when the depth of discharge increases, this effect is less pronounced than with its lead-acid counterpart. A lithium-ion battery lasts 3,000 to 5,000 cycles (8-14 years) at a DoD of 100%, 5,000 to 7,000 cycles (14-19 years) at a DoD of 80%, and 7,000 to 10,000 cycles (19-27 years) at a DoD of 33%.  As a consequence, lithium-ion storage usually has a DoD of 80%, while lead-acid storage usually has a DoD of 33 or 50%. In the LCA of the Spanish off-grid system discussed above, the assumption of three days of autonomy implies that 41 kWh of storage is required (3 x 13.8 kWh per day). Because the DoD is 33%, total storage capacity should be multiplied by three, which results in 123 kWh of batteries. If we would replace these by lithium-ion batteries with a DoD of 80%, only 50 kWh of storage is needed, or 2.5 times less.
6 x Less Batteries Needed
For utmost accuracy, we should mention that the lifespan of a battery isn't necessarily limited by the cycle life. When batteries are used in applications with shallow cycling, their service life will normally be limited by float life. In this case, the difference between lead-acid and lithium-ion is less pronounced: at no-cycling (float charge), lithium-ion lasts 14-16 years and lead-acid 8-12 years. Battery life will be limited by either the life cycle or the float service life, depending on which condition will be achieved first.  Nevertheless, if we focus on off-grid systems for households, the assumption of deep daily cycling better reflects reality, although there will be periods of float charge, for example during holidays.
The total storage capacity to be manufactured over the complete lifetime of a solar PV system is 6 times lower for lithium-ion than for lead-acid
If we also factor in the lifespan of the batteries, the advantage of lithium-ion becomes even larger. Assuming a lifespan of 20 years for the solar PV system and a DoD of 80%, the lithium-ion batteries will last as long as the PV panels. On the other hand, the lead-acid batteries have to be replaced at least 2-4 times over a period of 20 years. This further widens the gap in energy use for manufacturing when comparing lead-acid and lithium-ion batteries.  In the original LCA, a total storage capacity of about 240 kWh is needed over a lifespan of 20 years. On the other hand, the cycle life of the lithium-ion battery is 19-27 years, meaning that no replacement may be needed. Consequently, the total storage capacity to be manufactured over the complete lifetime of the system is 6 times lower for lithium-ion than for lead-acid. 
E3DC lithium-ion battery system. Picture: Thomas Salzmann.
If we take the most optimistic values for energy during manufacturing, being 0.87 MJ/Wh for lead-acid and 1.4 MJ/Wh for lithium-ion, and multiply them by total battery capacity over a lifetime of 20 years (248,000 Wh for lead-acid and 42,000 Wh for lithium-ion), this results in an embodied energy of 60 MWh for lead-acid (the value in the original LCA) and only 16.5 MWh for lithium. In conclusion, energy requirements for the manufacturing of the batteries is 3.6 times lower for lithium-ion than for lead-acid.
Another advantage of lithium-ion batteries is that they have a higher efficiency than lead-acid batteries: 85-95% for lithium-ion, compared to 70-85% for lead-acid. Because losses in the battery must be compensated with higher energy input, a higher battery efficiency results in a smaller PV array, lowering the energy requirements to manufacture the solar cells. In the original LCA, 4.2 kW of solar panels (35 m2) are needed to produce 13.8 kWh per day. If we assume the lead-acid batteries to be 77% efficient, and the lithium-ion batteries to be 90% efficient, the choice for lithium-ion would resize the solar PV array from 4.2 kW to 3.55 kW. We now have all the data to calculate the greenhouse gas emissions per kWh of electricity produced by an off-grid solar PV system using lithium-ion batteries.
GHG Emissions of the Off-grid System with Lithium-ion Batteries
In the original LCA, the batteries and the solar panels (including frames and supports) account for 59 and 62 gCO2e/kWh, respectively. The rest of the components add another 10 gCO2e/kWh, resulting in a total of 131 gCO2e/kWh. If we switch to lithium-ion battery storage, the greenhouse gas emissions for the batteries come down from 59 to 20 gCO2e/kWh. Because of the higher efficiency of the lithium-ion batteries, the greenhouse gas emissions for the solar panels come down from 62 to 55 gCO2e/kWh. This brings the total greenhouse gas emissions of the off-grid system using lithium-ion batteries to 85 gCO2e/kWh, compared to 131 gCO2e/kWh for a similar system with lead-acid storage.
While this result is an improvement, it's dependent on the assumptions of the researchers; most notably, a solar insolation of 1,932 kWh/m2/yr, and that all manufacturing of components occurs in Spain. If we adjust the value for a solar insolation of 1,700 kWh/m2/yr in order to compare with the other results, total GHG emissions become 92.5 gCO2e/kWh (assuming battery capacity remains the same). If we correct for a solar insolation of 1,000 kWh/m2/yr, the average in Germany, GHG emissions become 123.5 gCO2e/kWh. Furthermore, if we assume that the solar panels (but not the batteries or the other components) are manufactured in China, which is most likely the case, GHG emissions rise to 155 and 217 gCO2e/kWh for a solar insolation of 1,700 and 1,000 kWh/m2/yr, respectively.
Testing a lithium-ion battery. Picture: A123 Systems.
In conclusion, lithium-ion battery storage makes off-grid solar PV less carbon-intensive than conventional grid electricity in most western countries, even if the manufacturing of solar panels in China is taken into account. However, the advantage is rather small, which effects the speed at which solar PV systems can be deployed in a sustainable way. In the previous article, we have seen that the energy and CO2 savings made by the cumulative installed capacity of solar PV systems are cancelled out to some extent by the energy use and CO2 emissions from the production of new installed capacity. For the deployment of solar systems to grow while remaining net greenhouse gas mitigators, they must grow at a rate slower than the inverse of their CO2 payback time. [20, 21, 22]
Off-grid solar systems with lithium-ion battery storage can have GHG emissions below 30 gCO2e/kWh if they are produced in countries with clean electricity grids, and installed in countries with high solar insolation and carbon-intensive grids.
For solar panels manufactured in China and installed in countries like Germany, the maximum sustainable growth rate is only 16-23% (depending on solar insolation), roughly 3 times lower than the actual annual growth of the industry between 2008 and 2014. If we also take lithium-ion battery storage into account, the maximum sustainable growth rate comes down to 4-14%. In other words, including energy storage further limits the maximum sustainable growth rate of the solar PV industry.
On the other hand, if we would produce solar panels in countries with very clean electricity grids (France, Canada, etc.) and install them in countries with carbon-intensive grids and high solar insolation (China, Australia, etc.), even off-grid systems with lithium-ion batteries would have GHG emissions of only 26-29 gCO2/kWh, which would allow solar PV to grow sustainably by almost 60% per year. This result is remarkable and shows the importance of location if we want solar PV to be a solution instead of a problem. Of course, whether or not there's enough lithium available to deploy battery storage on a large scale, is another question.
Battery Production Powered by Renewable Energy?
Another way to improve the sustainability of battery storage is to produce the batteries using renewable energy. For example, Tesla announced that its "GigaFactory", which will produce lithium-ion batteries for vehicles and home storage, will be powered by renewable energy. [23, 24] To support their claim, Tesla published an illustration of the factory with the roof covered in solar panels and a few dozen windmills in the distance.
However, the final manufacturing process in the factory consumes only a small portion of the total energy cost of the entire production cycle -- much more energy is used during material extraction (mining). It's stated that the GigaFactory will produce 50 GWh of battery capacity per year by 2020. Because the making of 1 kWh of lithium-ion battery storage requires 400 kWh of energy [16, 17, 18], producing 50 GWh of batteries would require 20,000 GWh of energy per year.
If we assume an average solar insolation of 2,000 kWh/m2/yr and a solar PV efficiency of 15%, one m2 of solar panels would generate at most 295 kWh per year. This means that it would take 6,800 hectares (ha) of solar panels to run the complete production process of the batteries on solar power, while the solar panels on the roof cover an area of only 1 to 40 ha (there is some controversy over the actual surface area of the factory under construction). Tesla's claim, though potentially factually accurate, is an obvious example of greenwashing -- and everyone seems to buy it.
There are other ways to improve the sustainability of solar PV when storage is taken into account. Most of these solutions require that solar systems remain connected to the grid, even if they have a (more limited) local storage system. In this scenario, chemical batteries could help to balance the grid system, acting as peak-shaving and load-shifting devices. The electric grid has to be sized to meet peak demand, and battery storage could mean that less power plants are needed for that. Decentralized, grid-connected energy storage could also increase the share of renewables that the electricity infrastructure can handle. Of course, this "smart grid" approach should also be subjected to a life cycle analysis, including all electronic components.
Kris De Decker (edited by Jenna Collett)
EDIT: the paragraph about Tesla's GigaFactory was rewritten to reflect the fact that most energy is consumed during material extraction.