For places between latitudes 30 or 35 north and south, we only need to provide enough storage for one day (diurnal storage). Although there are seasonal fluctuations, the difference between winter and summer insolation is small or at least tolerable. Averaged out over thousands of locations, the supply of power is predictable. One day's worth of battery storage will cater for energy demands most of the time, with the batteries filled up between 9 and 3 and drained during the morning and afternoon peak.
And even one day's storage would be pricey. The Tesla Powerpack probably costs about $115/MWh now [Edit, 10/Dec 2016: I have probably overestimated the cost of Tesla batteries in my initial stab at estimating their LCOE because Tesla understates the actual capacity, in which case costs after this new decline are more like $70/MWh] , which means that if solar costs say $30 per MWh, adding just one day of storage increases the cost by 4 to 5 times, to the top end of coal. Battery costs are falling, so in 5 years, costs might have fallen by 2/3rds (let's be conservative!) and solar costs may well have halved so solar plus batteries would be about $65/MWh. Those costs assume an "island grid", a real island like Hawaii or a virtual island like Western Australia, where no power can be "borrowed" from other regions. If a region is part of a wider grid much less storage is needed. And the calcs also ignore that 2/3rds of power is consumed during the day. (The only reason the net demand peak has moved into the afternoon is because rooftop solar panels pump so much power into the grid at midday.) It also ignores that the big battery installers will initially be households with "behind the meter" storage, where the cost consideration is different, because they pay the (much higher) retail rates for electricity than the lower wholesale rates big users pay. But still, let that calc stand: $145/MWh now, probably falling to $65/MWh in 5 years.
Obviously, the further away you go from the equator, the less the insolation and the bigger the seasonal variance. At the extremes, within the arctic circle, there is no sun for 3 months of winter, while for 3 months over May/June/July there is sun for nearly 24 hours. In the high latitudes, solar can still be used for power in summer, but not for winter. And one day's solar storage wouldn't really be much use.
The good news is that at high latitudes wind is stronger and blows for longer. Hence the prevalence of wind rather than solar as a generating source in Scotland/Denmark/Northern Germany. And the wind blows all winter. Although from year to year the wind averages the same, it's possible (though rare) in high latitudes that there might be no wind for a week (the longest "no wind period" recorded in Denmark is 7 days.) And it is possible that there could be a cold spell, which could even--oh horror!--coincide with that wind-free period.
For this we need seasonal storage: storing power when the sun is strong and the wind high to provide for the winter dark. This isn't a problem now: many high latitude wind-dependent regions have reached 50% renewables without encountering problems with electricity supply in winter or with grid stability. But as the percentage from renewables rises this is going to become more difficult. Currently, we use gas peaking power plants to provide for times of high demand and low supply, but that means we are still using fossil fuels and are still emitting CO2.
Battery storage for such contingencies would be prodigiously expensive. For an "island grid", there would need to be at least one week of storage. The storage would cost 20 times the cost of the wind power alone.
But while we wait for battery costs to halve and halve again, there are two cheaper alternatives. One is concentrated solar power with storage. As the molten salts storage is integral to the system, the total cost of power production plus storage is cheap: below $80/MWh. But even here, building tanks to store molten salts for a week would add to the cost, though I have no idea how much. Producing power 24/7 is one thing; storing enough power for a week or more would be more costly. And anyway, for an "island grid" in high latitudes there just isn't enough sun in winter.
The other is power-to-gas. This is a shorthand phrase for producing hydrogen by electrolysis of water and either feeding the hydrogen into the gas grid directly or converting the hydrogen to methane using the Sabatier process and storing that. This synthetic methane is indistinguishable from natural gas. The sources I looked at proposed various limits on the percentage of hydrogen you could mix with methane from a conservative 5% to a generous 20%. Electrolysing just to hydrogen and not producing methane as an additional step means a slightly lower (8%) energy conversion loss and less capital expenditure, which would reduce costs, so the higher the hydrogen percentage feasible, the better.
Most developed countries have about 10 weeks of gas storage already, and a gas reticulation system. Adding power-to-gas to our energy system to provide for seasonal storage would be cheaper (for now) than using batteries. How much would it cost? According to this article, the energy loss going from electricity to gas and back again is about 65%. With co-generation (using the heat produced during the burning of the gas to heat buildings as well as produce electricity), it's only 50%. That means that the electricity produced from this power to gas would cost between 2 and 3 times the cost of wind electricity on its own. On the other hand, it would be needed only for say 4 weeks a year, with a four to six weeks reserve. That would increase the cost of wind by about 15%. However, you need a concentrated source of CO2 to produce the methane, and the best way to get that would be to capture and store the CO2 emitted when the methane is burnt. This would produce a closed cycle with no net CO2 emissions. The best estimate of costs I could find was a piece by the Global CCS institute (chart on page 9), which calculates the additional cost of carbon capture and storage for a gas combined-cycle generating plant as somewhere between $6 and $34/MWh. This increases the additional cost of power-to-gas to about 23% above the cost of wind alone (using the highest CCS estimate), for three weeks' storage. In other words, wind alone $30/MWh, but with 4 weeks power to gas storage, $30 +$7.1.
These are my guesstimates. I would be fascinated if any of you know of better data.
One final point. None of this is to be taken as arguing that we can't have renewables in our electricity grid. We can, and many countries have reached 50% from renewables without grid instability or power outages. But if we want to increase that percentage, storage will become essential. It will take us 15 or 20 years to get to 100% renewables, so we have time to plan just how we'll do it, while all the time, the costs of these new processes will fall. In the end, I suspect we'll use all of the techniques at our disposal: battery storage, CSP, power-to-gas, HVDC lines, and demand management.