How much backup storage is really required for alternative energies? How does this compare to existing storage on the grid?
Studies on Wind Intermittency and Required Backup
A comprehensive study was done in 2006 by the UK Energy Research Centre. Formally established and overseen by a body of experts, it assessed 200 studies and reports from around the world about the actual and theoretical impact and mitigation of intermittency in power grids.
For penetrations of intermittent renewables up to 20% of electricity supply, additional system balancing reserves due to short term (hourly) fluctuations in wind generation amount to about 5-10% of installed wind capacity.
The risk of demand being unmet can be characterised statistically, and the measure commonly used to quantify this risk is called Loss of Load Probability (LOLP). This measures the likelihood that any load (demand) is not met, and it is usually a requirement of electricity systems that LOLP is kept small.
Capacity credit is a measure of the contribution that intermittent generation can make to reliability. It is usually expressed as a percentage of the installed capacity of the intermittent generators. There is a range of estimates for capacity credits in the literature and the reasons for there being a range are well understood. The range of findings relevant to British conditions is approximately 20 – 30% of installed capacity when up to 20% of electricity is sourced from intermittent supplies (usually assumed to be wind power). Capacity credit as a percentage of installed intermittent capacity declines as the share of electricity supplied by intermittent sources increases.
(Note that this is nameplate capacity, not averaged contribution.)
Current costs are much lower; indeed there is little or no impact on reliability at existing levels of wind power penetration. The cost of maintaining reliability will increase as the market share of intermittent generation rises.
A Finnish study  concurs:
From the investigated studies it follows that at wind penetrations of up to 20% of gross demand (energy), system operating cost increases arising from wind variability and uncertainty amounted to about 1ñ4 Ä/MWh. This is 10% or less of the wholesale value of the wind energy.
It contains an excellent chart showing the value of widespread wind energy in reducing intermittency of supply garnered from data from multiple countries historical experiences:
The Finnish study had the following participating countries and bodies:
• Denmark: Ris National Laboratories; TSO Energinet.dk
• EWEA (European Wind Energy Association)
• Finland: VTT Technical Research Centre of Finland (Operating Agent)
• Germany: ISET; TSOs RWE and E.ON Netz
• Ireland: SEI; UCD; TSO Eirgrid
• Norway: SINTEF; Statkraft
• Netherlands: ECN
• Portugal: INETI; TSO REN
• Spain: University Castilla La Mancha
• Sweden: KTH
• UK: Centre for Distributed Generation & Sustainable Electrical Energy
• USA: NREL; UWIG
Backup required for other generation sources
There’s an interesting example of the odd way that some people look at this in the moderately famous Ardrossan wind turbine fire of December 2011. One of a dozen 1.2 MW wind turbines caught fire in a massive wind storm that swept Scotland, taking its 1.2 MW out of generation. The same wind storm knocked down transmission lines from the nearby Hunterston nuclear plant. It was offline for 54 hours for a loss of 17,000 MWh to the grid. That’s about six years of generation capacity of the wind turbine. 
Similarly, when an Australian 800 MW coal plant stopped delivering electricity to the grid recently, the wholesale price of power increased by a factor of 200 in minutes before returning to normal. This graph is leveled over 30 minutes so the peak price is masked, but the dramatic loss of power is readily apparent. As the linked article shows, loss of major generating assets is common and unpredictable, while loss of wind generation is common, but typically only a percentage of capacity and very predictable.
Grid managers have to maintain hot backup contingencies for failure of their largest single generation plants, typically coal, hydro or nuclear in the 1 GW range. Wind energy doesn’t rank as a grid management issue until you get into > 20% ranges, and even then it isn’t a particularly hard or sudden problem compared to dealing with a nuclear plant that suddenly isn’t there. 
Ontario, as another example, gets 55% of its energy from its fleet of nuclear plants, which average around 850 MW per reactor. Ontario’s nuclear fleet has experienced many unforeseen shutdowns. One of these plants having a failure which takes it offline requires 100% backup for that contingency, or 850 MW.
Ontario has has a fleet of large hydro facilities, one of which generates 1500 MW by itself. While Ontario is only slightly geologically active, earthquakes registering 3 on the Richter scale have occurred and could cause a hydroelectric dam to be taken out of service.
Ontario has 1500 MW of wind capacity at present and is much lower than 20% of generation from wind (and does not currently plan to go anywhere near 20%).  At the 20% backup, this would currently require 300 MW of backup generation, or about the size of a single large gas turbine generator.
Ontario is ahead of schedule to eliminate coal generation entirely, with all of the attendant global warming and health advantages that moving from that dirtiest of fuels entails.
 “GHG [Green House Gas] and Cost Implications of Spinning Reserve for High Penetration Renewables, Technical Assessment Report 73 – March 2008” from the CRC for Coal in Sustainable Development.