This is a response to the post by Roger Andrews on Euan Mearns’ web site. Roger’s post can be found here.
Roger’s post describes the part 1 grid hourly simulation as “probabilistic” which it isn’t. While it is not perfect, the grid hourly simulation is “deterministic” as it is a true simulation.
What his post says Roger did with the load and generation data is not what he actually did with it.
The post says he took the data from the grid model spreadsheet linked to from this site and “…added one further adjustment by factoring wind and solar output upwards by about a third so that total generation was equal to total demand.” This implies that the average generation in the grid hourly simulation was originally less than average demand. It wasn’t. Generation was higher than load in the simulation spreadsheet by 27%.
Somehow, what he actually did was to factor the load up by a third instead, effectively removing the 27% of over-generation required in the workable 100% renewable solution design. That is why the peak hourly demand in some of his charts has become 90 GW instead of the 71 GW in my grid hourly simulation spreadsheet. The inflated 90 GW peak hourly demand is not obvious in charts containing three full years of data because demand is averaged over the whole of each calendar day, giving peak days of noticeably over 71 GW for a grid with a maximum hourly peak of 71 GW.
Equating generation to demand is not a workable solution for a 100% renewable grid, not least because of storage losses. Playing with the grid hourly simulation spreadsheet will soon convince you of this. As a result his post overestimates the total storage required by a factor of 3.5, i.e. 50 TWh compared with 14.3 TWh for a workable 100% renewable solution with 27% over-generation.
Based on the unnecessarily high total storage figure of 50 TWh his post dismisses the 100% renewables solution without costing the tier 2 renewable (synthetic) methane storage solution which provides 98% of the storage capacity at a very low cost. A two-tier storage solution is economic whereas one based on batteries alone would indeed not be.
Despite mention of 5,500 Denorwigs (a Welsh (UK) pumped storage scheme), the UK’s “Rough” depleted natural gas storage field, in use for 34 years, had more than enough capacity to store the methane required to generate 14 TWh of electricity. It had the capacity for 9 days of total UK gas supply but has now been closed. Texas has many depleted gas fields.
Renewable over-generation reduces :
- the duration of gaps during which direct renewables supply is insufficient to meet demand
- the deficit (in GW) during the times when direct renewables supply is insufficient to meet demand
- the peak storage capacity requirement
- the fraction of demand (in GWh) which must be supplied from storage
- storage losses
The following table shows the effect of different levels of over-generation for a 100% renewable solution nominal peak load of 71 GW and average load of 39.5 GW:
|Supply gap %
for no gaps
As losses are set to zero in this table to compare with Roger’s result, the above figures from the grid simulation are not directly comparable to the results for scenario 3 100% renewables.
The proposed 27% over-generation adds between $5.6 and $8.6 / MWh to the cost of supplied electricity. 14 TWh of tier 2 storage capacity (in a salt cavern or depleted gas field) itself adds only $1 to $1.2 / MWh (excluding electrolysers to produce the renewable gas). Interestingly the energy density of renewable methane stored at 200 atmospheres underground is similar to that of lithium ion batteries.
The grid hourly simulation shows 27% over-generation enables all 2010-12 scaled load to be supplied from renewable generation, either directly, or indirectly via the two tiers of storage. Electricity in Texas part II shows that the cost of doing so is not excessive. Roger’s article did not take the analysis far enough.
Requirement for two tiers of storage for Texas
Some countries, such as China, Brazil, Canada and Norway, have large quantities of hydro and/or pumped hydro generation which can be used as a single tier of storage to complement renewable wind and solar generation. For instance Norway has 84 TWh of hydro storage (around 10 days of total European electricity consumption). Some of this could be converted to pumped hydro to balance North Sea wind if countries in northern Europe are prepared to pay. However, most regions are not so lucky, and Texas has very little hydro.
The implicit assumption in Andrew Roger’s post is that, in the absence of significant hydro capacity, all storage has to be battery storage. This would be a poor solution. One workable storage solution for the Texas grid in 2030-40 is to use a combination of battery storage and renewable gas production, storage and use in gas turbines or fuel cells for back-up generation when required.
Battery storage is relatively efficient (85% for an 80% depth of discharge), but not that cheap. But more than a day of storage would be unaffordable.
The suggested second storage tier, relying on renewable methane storage, is highly inefficient (34%), but the storage capacity is very cheap indeed at $250 to $300 per MWh of storage capacity. Thus you can have huge storage capacity, but the fraction of supply which goes through this storage mechanism must be kept low – less than 10% (and 6% in the grid hourly simulation) – or you pay too much for over-generation just to support storage losses.
For Texas, a workable and affordable solution is a ratio of 2% battery storage (300 GWh) to 98% renewable gas storage (14,000 GWh). The 2% battery storage reduces the demand gap by 2/3 (down from 18% to 6% of average demand), so potential tier 2 storage losses are reduced from a nominal 36% down to 12%. Spare capacity from the suggested 28% over-generation also provides 18% of load factor for 40GW of electrolysers to produce sufficient methane to cover the 6% gap at 34% efficiency.
Other observations on Roger’s article
Roger talks about the stress imposed on gas turbine back-up with high penetrations of renewables. Providing one hour of battery storage eliminates most stress and ramp rate issues. For the 100% renewable solution 7.5 hours of battery storage is proposed, which removes any possibility of ramp rate and stress problems caused by too rapid cycling of gas turbine generation. It will also assist ERCOT in solving any remaining problems with AC grid instability caused by the last decade of major changes. Battery storage also provides required ancillary services as a by-product.
Roger’s post states that plots of gaps in variable renewable generation are essential to assessing whether an all-renewable system will work. While such charts may help visualise variability, most people understand that wind and solar PV generation has gaps. The key questions are whether the gaps can be filled using storage and how much it will cost. These can only be answered properly with a grid simulation and cost model, such as those on which this pair of articles is based.
He states that with low gas prices there is little incentive for ERCOT to increase its share of renewable energy. The Electricity in Texas part 2 article makes it clear that it is independent generators rather than ERCOT who decide what new generation to install. ERCOT is obliged to connect all new generation to the grid. Once connected, there is market competition to supply electricity to match demand. All the evidence so far is that the installation of new wind and particularly new solar is ongoing, despite in-progress reductions in the Federal PTC (Production Tax Credit) subsidies.
Some of this is driven by contractual PPAs (power purchase agreements) directly between generators and utilities or commercial entities or utilities who often value certainty of renewable pricing above a current low average gas generation price which is not guaranteed to be stable and to which an accounting risk premium is generally added. Consider the fixed price that gas generation plants would quote for a 15-year contract if they had to take on the risk of increasing fuel prices themselves instead of passing this risk to their customers as they do at present? A renewables contractual PPA is an excellent hedge against price rises for a big electricity user because, once installed, renewables prices are fixed.
In the comments section of Roger’s article there was some discussion around the maximum renewables generation which a grid could sustain. Some contributors insisted that the maximum fraction is inherently limited to the renewables capacity factor fraction – in this case 32.5% for both wind and solar.
This limit is blown apart by the grid hourly simulation of scenario 2 of part 1. Texas wind and solar, each with a capacity factor of 32.5%, can supply 68% of the load with no more than a 10% surplus/curtailment. Quite possibly this would cause significant stress on the gas back-up generation, so it makes sense to add 50 GWh of battery storage. At 2030-40 prices this quantity of storage would increase the average price of electricity supplied by about $2 / MWh, solve ramp-rate and stress problems, and also increase the coverage of renewables marginally by 2% to 70%.
For Texas there are two reasons why the capacity factor limit is not valid. Firstly Texas wind and solar are somewhat negatively correlated (-22%). That is the wind has a tendency to blow harder when the sun in not shining. Secondly both wind and solar are cheap so it costs little to have over-generation – 10% of the total renewables generation was proposed for scenario 2 of part 1. Those who doubt the conclusions of scenario 2 should download the grid hourly simulation spreadsheet and have a play.