Wednesday, July 6, 2016

Wind Water and Sunlight: Jacobson's sorcery

Dreams of a future powered by wind, water and sunlight (WWS) have been spread by Stanford professor Mark Jacobson. Jacobson's work is more optimistic about the contributions to come from WWS than other projections, which may explain an infectious political appeal. In my country the allure of a WWS power supply is impacting Canada's most left political party, the NDP.

SK Premier Wall and Lead Leaper Klein exchange tweets
I've looked at the the WWS work and will demonstrate its shortcomings in Ontario, with the expectation that can be extended to all northern jurisdications. One inspiration for looking at the viability of WWS is opportunistic: Ontario, with a new Minister of Energy, is initiating another Long-Term Energy Plan (LTEP) process. The WWS work provides a test case to run through a data model I created for the previous LTEP planning, and refine my work as necessary.

I won't concentrate on Jacobson's work to nearly the extent of Jani-Petri Martikainen's convincing and entertaining Why does Mark Jacobson hate Finland? or multiple insightful articles at the "A Chemist in Langley" site, including the most recent, Debunking the Leap Manifesto’s 100% Wind, Water and Sunlight Annual Energy, Health, and Climate Cost Savings. I'll attempt to limit myself to briefly discussing how the WWS paper arrived at its supply mix for Canada, and testing that mix within an hourly model - the only way a mix can be rigorously tested.

The April 2016 draft paper from Jacobson and others, 100% Clean and Renewable Wind, Water, and Sunlight (WWS) AllSector Energy Roadmaps for 139 Countries of the World, concluded, "The study finds that the conversion to WWS is technically and economically feasible. The main barriers are still social and political." It's a grand claim for a draft exercise that claims only to,
"[calculate] the number of generators of each type needed to power each country based on the 2050 power demand in the country after all sectors have been electrified but before considering grid reliability and neglecting energy imports and exports."
WWS does not present itself as attempting any hourly modelling of an electricity system, building requirements solely based on total energy requirements for a calendar year.

WWS establishes a mix by the awkward metric of gigawatt-year. That's a steady stream of 1 million watts each second throughout an entire year. I'll bring all measurements back to metric variants of the watt-hour (MWh and TWh).[1] For generation capacity I'll use simply megawatts (MW). The annual capacity is the starting point in the WWS exercise because it builds the energy requirements for its base year, 2012, from International Energy Agency (IEA) annual data presented in yet another energy metric (barrels of oil equivalent).

Some points, quoted from the paper, with some emphasis added, demonstrate how WWS builds from the initial annual data:
  • All energy-consuming processes in each sector are then electrified...
  • The WWS electricity generation technologies include wind, concentrated solar power (CSP), geothermal, solar PV, tidal, wave, and hydropower...
  • Technologies for ground transportation include battery electric vehicles (BEVs) and hydrogen fuel cell (HFC) vehicles...HFCs will not be used to generate electricity because of the relative inefficiency and cost of this process.
  • ...In this study, ~9.0% of all 2050 WWS electricity (44.5% of the transportation load and 4.8% of the industrial load) is for producing, storing, and using hydrogen...
  • Under the roadmaps, the 2050 nameplate capacity of hydropower in each country is assumed to be exactly the same as in 2014. However, existing dams in most countries are assumed to run more efficiently for producing peaking power, thus the capacity factor of dams is assumed to increase (Section 5.4).
There are highly questionable assertions here, but in trying to select a WWS mix for scenarios in Ontario, I did work through the Jacobson, and others, WWS spreadsheet (xlsx) to determine a WWS supply mix for Ontario with which to test an updated model of the supply mix calculator I first developed during the planning period for Ontario's 2013 Long Term Energy Plan (LTEP).
Conversely, the model will be testing both the WWS mix and the process system operators in North America use to ensure supply adequacy. The second test is the more important aspect of this work for any electricity system where wind and solar generation are growing to be substantial shares of the generation mix,

The WWS mix, for Canada, is set at:
  • 58.5% wind (37.5% onshore and 21% offshore),
  • "Wave Device" is set at 2%, none of which I assumed would be in Ontario,
  • Geothermal and "tidal turbine" are set by available capacity, and I've assumed none of that capacity is in Ontario,
  • CSP (concentrated solar power) is not given a share in Canada
  • roof solar photovoltaic (pv), both residential and commercial, are assigned a fixed capacity based on a share of all available roof locations
  • any unmet need for supply is to come from solar pv plant.
Remembering that WWS is set by annual consumption, the amount of generation capacity required demands a capacity factor for each of the generation types. The WWS spreadsheet has 3 WWS scenarios for Canada - 2012 (the year of the data they've converted to electricity needs), 2030 and 2050. I've considered Ontario to be 40% of the total annual needs in each scenario, to determine supply mixes for each. Ignoring the distraction of separating wind into on and offshore, and various types of solar, the WWS supply mix has only 3 entries:

Each of Wind, Water (hydroelectric) and Sun (solar) requires some assumptions to model hourly. I'm better prepared than most having gone through the process in 2011 and again in 2013. I've been actively tracking the system operator's hourly solar forecasts (starting 2014), and capacity, in Ontario.

There are modeling issues for each of W W and S, as well as some comments I'll make to be clear on my attempts to follow the WWS framework.

Industrial wind turbines are, I have no doubt, capable of the WWS data's 43% capacity factors in Ontario, and in 2005 the Ontario Power Authority (OPA) reported, for wind, "the theoretical potential is vast – in the order of 600,000 MW."  Problematically, the WWS specifies "before considering grid reliability" whereas the OPA, which did consider such things, recommended little of that theoretical wind potential be tapped. While 43% capacity factors are possible, the average annual capacity factor of actual turbines has been closer to 30% historically, which presents a challenge for hourly modeling. The data I've collected is mostly from conventional turbines and the advanced designs that deliver 40+% capacity factors have a different production profile, generating more at lower wind speeds, but not at high wind speeds. I've inelegantly programmed the hourly capacity factors to adjust to produce an annualized 43% (or whatever percent is entered in the calculator) without changing the highest production hours (approximately top 25th percentile).

Solar generating facilities in Ontario are producing above 16.2% capacity factors in the WWS scenarios, which seem plausible. This is not a simple comparison because Ontario solar is contracted allowing an overbuild of solar panels (DC) generation capability behind the connection point (AC). In reality individual panel capacity factors in Ontario are probably lower than the 16.2% average from the WWS planning, but not significantly so.

The WWS paper's premise that hydro-power will be more efficient if used for balancing wind and solar is obviously unsubstantiated - existing generators cannot run differently without extensive reconfiguration of power plants and transmission facilities, and peaking plants do not have the 50% capacity factor WWS assigns to hydro (which is close to today's factor for hydro in Ontario). I could not entirely overlook the irrationality of WWS for hydro. My 2013 LTEP calculator separated hdyro into 2 categories: baseload as many installations do have a "use it or lose it" aspect to their supply at all times, and "other". The "other" I have programmed, for this exercise to allow storage - which is not practical, but also doesn't really end up impacting the model's outcomes much. I have limited the hydro output at any one point a "firm capability" share, which limits hourly output values, but each scenario has hours of hydro production that exceed any value I've collected in nearly 6 years of capturing IESO data.

I back-tested changes to the calculator tool with Ontario's current supply mix (here).[2] In addition to testing the performance of the calculator, it also provides a benchmark comparison for the outcomes as the WWS mixes are fed into the model.

The first test I'll show inputs a supply mix using the WWS methodology to determine the supply mix appropriate to meet the demand levels for only electricity in recent years. WWS is intended to electrify all sectors, but the simplest test of the WWS method is to see how it does in electrifying the electricity sector. WWS specifies 9% of electricity generated will be used to produce and use hydrogen. I've set the supply mix to produce 158 TWh, both because WWS recognizes additional need and it's likely the minimum Ontario's current mix could produce (without curtailing generators), but the demand in Ontario is 142.5 TWh.


This is a daunting set of numbers, but the most important messages are highlighted in shades of red - these are cells which go red based on conditional formatting indicating supply cannot meet demand. An operator working the calculator might keep working until red was eliminated. The dark red cells tested the mix based on ability to meet the expected peak demand hours in each of winter and summer - the light red cells query 8760 records in each of 3 annual models to determine the total energy demand not capable of being supplied (shown in terawatt-hours) and the maximum shortage for a single hour (in MW)

There's an important difference in the quantification of the failure indicated by the dark red and the megawatt shortfalls of the hourly testing.

The dark red cells are based on my modeling of how Ontario's IESO meets the North American Electric Reliability Corporation (NERC) tests for sufficiency in generation capacity: they look to the anticipated peak demand time for a period, and then to each generator's expected capability to produce during that peak. I call this latter figure the "capacity value." The figures in the spreadsheet are from my 2013 work in which I took values for winter, and summer, peaks from the IESO documents.
There are multiple methods employed by system operators to estimate capacity values, but few looking at multiple years data to model the ability to meed demand at all demands -not only the one hour of peak demand.

The WWS scenario tested above fails the NERC test most significantly is being 11,328 megawatts short of acceptable winter firm capacity requirements. The simplest way to address this would be to throw up 12,313 megawatts of natural gas-fired capacity (if you are at the IESO you'd call this "flexible capacity" - which you'd estimate with the same cost and other attributes of a simple-cycle gas turbine power plant). Some WWS proponents would call for imports to fill the shortfalls, but Jacobson's WWS plan excludes such consideration.

Regardless of how one argued around the failure of NERC reliability requirements, this WWS scenario has greater hourly shortfalls in each of the three test years - ranging from 14,227 MW in the 2014 based scenario to 15,432 in my original LTEP 2013 hourly annual model.

The hours where generation is not substantial enough to meet demand are not rare.

Looking at the greatest generation shortfall on each day of the 2014 hourly scenario, 8 months have at least one day where the shortfall is greater than 12000 MW:


Comparing this first WWS supply mix to the current supply mix run through the same model, the current mix  has no shortfalls in any test (NERC or hourly). The WWS scenario drops 13,000 MW of nuclear capacity, and 10,658 MW of gas and biomass, and replaces it with 38,100 megawatts of wind and solar but causes a need for another 12,300 - 15,400 megawatts of firm capacity to meet demand.

Functionally, it appears 13,000 MW of nuclear would be replaced with an additional 2000-5000 MW of gas plus 38,000 MW of wind and solar.

Perhaps not.
Jacobson, like Amory Lovins, seems fond of the rhetoric around solving problems by expanding them. The next test scenario takes the 2012 WWS scenario's mix for meeting all energy requirements (displacing gasoline and natural gas). It may be possible that the great excesses of electricity can be used to accomplish other goals, but I simply want to test if it can accomplish the one goal of meeting today's level of electricity demand.
I've rerun the calculator with the WWS energy mix calculated to produce ~570 TWh a year (40% of Canada's 1,428 in Jacobson's scenario), although I've left electricity demand unchanged at 142.5 TWh.


With this heavily skewed annual overproduction the model does show NERC reserve margins met, but the hourly modelling tells a different story. Again there are hours where supply is not met: this time the best performance is in the 2015 test model, with only a 12,052 MW shortfall, with the 2013LTEP scenario the worst at a 14,636 high hourly supply shortage.[3]

If I was lazy and just plugged in gas capacity to cover the hourly shortfalls, this scenario would meet demand at all times by effectively replacing 13,000 MW of nuclear, but no natural gas-fired generators, with 220,000 MW of wind and solar.

One reason I ran the numbers through 3 annual data sets was to see how the heavy variability of wind and solar impact the ability to meet demand each year. My conclusion is very little.

Jacobson's WWS work covers other scenarios and explores other components including costs and job impacts, but I see no point in continuing after the comprehensive failure in testing against hourly models for 3 distinct years. 100% WWS starts with a nonsensical premise in breaking down annual totals to capacity needs. The important message here is that is exactly wrong. The test of an electricity system is its ability to meet demand at all times, and WWS fails that test.

Having demonstrated the error there's no need to evaluate the WWS paper further.

This does not claim much about wind and solar beyond dismissing the 100% WWS scenario. The first scenario I displayed above would increase emissions in Ontario if natural gas-fired generators were built to avoid the shortfalls, but firm dependable generators are required.

Naomi Klein's description of Mark Jacobson as a "wiz" is unintentionally appropriate. The curtain doesn't need to be pulled back very far to expose an emptiness behind the glorious presentation of a world powered by only wind, water and sun.

______________________________________


postscript

Producing this post provided an excuse to revisit my calculator as my province moves into a Long-Term Energy Plan exercise.

I hesitantly provide the link for the calculator, but note this version is revised specifically for this post, and I have not tested the implications on all aspects of the calculator, nor do I think the hydro changes made for this post reflective of what is operationally possible.
If one unhides all cells in the spreadsheet, they'll see it is also intended to estimate costs - but those cells are untouched since 2013, and likely outdated.

One further note, for skeptics: the 2014 and 2015 data should be similar to what you'd see from the IESO. For instance, the greatest shortfall in the final scenario came during the evening of February 21st, 2015. You may know it was dark, and much of the historic wind data (the IESO transmission connected sites) is available in the Hourly Wind Generator Output, 2006-present file regularly updated on the IESO's Power Data page. To be clear, my claims of hourly inadequacy during non-daylight hours in the 2014 and 2015 models can be supported using IESO data. However, the data values will differ as my demand, and generation figures, are adjusted to include estimated distribution-connected generation, and further adjusted to the total annual demand requirements, capacity factors, and generation totals entered into the [SupplyMix] tab of the calculator/spreadsheet.

End-notes

1. One excellent source on energy units, and many other things, is the Energy and Power section of the late David Mckay's Sustainable Energy Without the Hot Air.
While units are generally force over a second, in my general discussions a MW (megawatt) is often used interchangeably with a MWh (a megawatt exerted for an hour), but it's strange to see it used interchangeably with a MWy (megawatt-year), which would more often referred to the 8760 MWh it also is.
Summarily:
one watt-hour is 3600 joules
one watt is a joule per second
one joule is a Newton exerted for one metre
one Newton is "the force needed to accelerate one kilogram of mass at the rate of one metre per second squared in direction of the applied force."

2. I show this, and comment, here. The supply mix is from the 2016 Q1 Ontario Energy Report, inclusive of both transmission (Tx) connected resources and distribution (Dx) connected. Because Dx is included, actual consumption/demand is higher than reported by the IESO, Ontario's system operator - which ignores Dx in its reporting, therefore showing demand lower than it actually is.

3. Showing the maximum daily shortfalls throughout the year in the same format as for the earlier scenario (this time for the 2015 model) shows the shortfalls are again not rare:

No comments:

Post a Comment