Modeling Disparity - LifeCycle Analysis vs Specific Carbon Emissions from Combustion

[HansHyde]

Please consider the following Charts based on Real-Time Data for US-NY (NYISO) over the course of 24-hours as reported for January 5th 2016.

Data was compiled from 3 sources available on the NYISO website - http://www.nyiso.com/public/index.jsp

• Interface Limits & Flows - http://mis.nyiso.com/public/csv/ExternalLimitsFlows/20180105ExternalLimitsFlows.csv
• Integrated Real-Time - http://mis.nyiso.com/public/csv/palIntegrated/20180105palIntegrated.csv
• Real-Time Fuel Mix - http://mis.nyiso.com/public/csv/rtfuelmix/20180105rtfuelmix.csv

Data from all 3 was compiled to verify the system was in balance; Supply, Load, Exchanges. The one irregularity, where System Load drops below System Supply offers a clue to NYISO not reporting Pumped Hydro operations distinctly from Hydro Generation, and can be discounted during the analysis.

The following 3 Charts were scaled horizontally to roughly the same dimensions for equal comparisons.

Chart 1 - NYISO Generation, Imports/Exports & Load

Chart 2 - CO2 Emissions Expected - SECF Modeling

Modeling: Specific CO2 Emissions from Combustion of Fuels, including estimated emissions resulting from Extraction, Processing & Delivery of Fuels to the Power Plant. For thermal generation, industry accepted heat rates (plant Prime Mover basis - not yearly "operational efficiency" aka Capacity Factor) was used.

Chart 3 - CO2 Emissions Expected - EM/LCA Modeling

Modeling: Based on the IPCC 2014 LifeCycle Analysis CO2 Emissions per kilowatt-hr. Values used, grams per kilowatt-hr, were as would be applied to a real-time data stream. In this model, Coal & Dual Fuel were included within "Unknown" at 700g, "Other REs" at 230g for Biomass. For nuclear, hydro, wind, etc., if they do not show on the graph, it is only because carbon-based fuel generators are flooding them out.

Imports & Exports were equally not included in either model.

Carbon Emissions Rates for Model 1 & Model 2 were plotted on both Charts for reference purposes.

What strikes me most, is not that LCA modeled Carbon Intensity Rate is less than Specific Carbon modeling Rate, but that the LCA is almost flat-line, with little variance as the generation varies or peaks throughout the day. Carbon Emissions (real) are increasing/decreasing dramatically, while the LCA Rate is "buffered" for no explainable reason.

Thoughts...

I will let all look at the data, comparisons, before offering any additional thoughts on what may be at play.

[jarek]

So, am I getting it right - they are peaking mostly with coal? Isn't this mostly down to not having a breakdown between coal and natural gas in "dual fuel" real-time data?

[HansHyde]

@jarek

No, coal or oil-only plants are reported distinctly as "Other Fossil Fuels".

In the SECF (Specific Emissions for Combustion of Fuels) model,

• "Other Fossils Fuels" = Coal or Oil only plants with SCEF - COAL.
• "Natural Gas" = NG only plants with SCEF - NATURAL GAS - This is named "NATURAL GAS" on chart
• "Dual Fuel" = NG &/or Oil capable plants with SCEF - NATURAL GAS x 1.2 (to factor some oil use) - This is named "Natural Gas" on the chart.
• "Other REs" = biomass/biogas/solar ( <400 MW Total Installed Cap) with SCEF - SOLAR

To find SECF parity between carbon emitting & non-carbon emitting fuels, I have modeled a carbon emissions factor for a Fuels-basis calculation as such, where Natural Gas = 1.0;

• Coal = 1 / 0.4875 NG (coal is ~2x more carbon emitting than NG)
• Oil = 1 / 0.652 NG
• Hydro = 1 / 52 NG (water is 1/52th as carbon emitting as NG)
• Solar = 1 / 52 NG
• Wind = 1 / 52 NG
• Nuclear = 1 / 5,216,308 NG (nuclear is 1/5,216,308th as carbon emitting as NG)

In the EM- LCA model,

• "Other Fossils Fuels" - "Unknown" with EM/LCA - 700
• "Dual Fuel" - "Unknown" with EM/LCA - 700
• "Natural Gas" - "Unknown" with EM/LCA - 700
• "Other REs" - "Biomass" with EM/LCA - 230 - If EM-LCA used Solar = 45 AND SCEF used Biomass (wood), the differences in CO2 emissions would only increase.

To answer your question specifically, please check that you are not misinterpreting the data in Chart 1 with the data in Charts 2 & 3. They are not reporting the same things.

Source of SECF - https://www.engineeringtoolbox.com/co2-emission-fuels-d_1085.html

[HansHyde]

for clarity of Modeling Methodology,

• SECF does NOT include generation plant commissioning or decommissioning OR delivery of electricity.
• EM/LCA does include plant commissioning (comment @brunolajoie) AND delivery of electricity (comment @corradio).

"EM COUNTRY RATE" equals the Carbon Intensity that would be calculated for a Country "Parser", using EM/LCA modeling.

"GRAMS PER KILOWATT-SEC" equals the Carbon Intensity calculated for a Country "Parser", using SECF modeling.

"Rate inc. (Import/Exports)" represents fluctuations that might be expected if imports/exports were included in the EM/LCA Carbon Intensity Rate. They include real data (volumes) for 4 exchanges;

• NE-ISO using EM/LCA - 700
• PJM using EM/LCA - 700
• Ontario using EM/LCA - 24
• Quebec using EM/LCA - 24

Imports/Exports were NOT used in either the SECF or EM/LCA models for calculation purposes. It is only displayed for reference purposes.

[jarek]

What is the source of the data in Chart 2 "Specific Carbon Emissions from Fuel Combustion" where it shows separate information for Coal and Natural Gas? How are the emissions from burning coal calculated? How does it know how much coal was burned?

[HansHyde]

@jarek

1) SECF info is linked above. 2) based on SECF & Fuel-Type Heat Rates (not yearly capacity factors, often misrepresented at "plant efficiency"). 3) reversed from real-time output (MW) against No. 2

[jarek]

If I understand correctly, you put "other fossil fuels" as coal and "dual fuel" as NG*1.2.

This causes the discrepancy. EM puts all of "other fossil fuels", "dual fuel", and "other renewables" as "unknown" at 700. You have "dual fuel" at 490*1.2=588 and "other fossil fuel" at 820. The critical thing is that "dual fuel" is a lot bigger than "other fossil fuels", so it will drown out and "dampen out" changes in "unknown" for EM. You show "other fossil fuels" separately, and it is much more peaky - production varying from 300 to 700 MW while "dual fuel" goes from ~5000 to ~7000 MW. Your assumptions allow you to pick up more of the peaking effect.

[HansHyde]

@jarek - no, your assumptions are not correct.

Any "peaky-ness" has nothing to do with differences in FUEL-TYPE PLANT efficiency. The dataset is based on a 5 minute frequency of reporting. Any increases in emissions from "ramping" without electricity generation are balanced/factored accordingly for all 1 hour or less ramp speed plants at a 1/12 delay between emissions - uncalculatable (no reported MW output) and emissions - calculated (reported MW output).

Even if this was a factor, SECF's rate would increase, increasing the difference between EM/LCA's rate.

Look at the charts again. What value is not responding proportionally to other data compiled in the charts?

[jarek]

By "peaky" I meant difference in production between middle of night and evening. "Other fossil fuel" production goes up by 100%, while "dual fuel" production goes up by 40%.

Unfortunately I don't understand beyond what I posted above. Hopefully others do.

[HansHyde]

@jarek in the EM/LCA model, "dual fuel" is not at 490*1.2=588, it is at "unknown" = 700

In the SECF model, "dual fuel" is at 1.2x the expected emissions of natural gas per MMBTU required to produce 1 MW-sec of electricity.

[HansHyde]

@jarek Chart 1 is electricity output (MW). Charts 2 & 3 are CO2 emissions.

Are all Rates responding equally to CO2 emissions shown for both models? (PS I fought with what the data was showing me, as a potential calculation error. But the data & calculations are not "lying".)

[corradio]

Wait - if we are using different assumptions as to what the underlying mix of electricity is, how can we even conclude anything about CO2? Seems also to be what https://github.com/tmrowco/electricitymap/issues/990#issuecomment-356255027 is concluding. I think the thing that @HansHyde is questioning here is first and foremost how the mix of electricity is calculated no?

[jarek]

Hm, my latest guess is that the difference is from use of grams-per-kWh vs grams-per-kW-second. Electricity Map takes in instantaneous production (kW, MW) and presents output as g/kWh (even when data is sub-hour resolution). Could the conversion be wrong?

[HansHyde]

@corradio the mix of electricity is known. It is our real-time data reported in MW to a precision of 0.1 MW (or 100kW) per Fuel-Plant Type.

NYISO data from the 3 sources identified above (CSV files) contained 36 rows of data, each row consists of 288 columns, or 10,368 unique data points. This took ~3 hours to reformat in Excel to use in the 2 models, and parse out unneeded data points from the CSV timestamp strings.

I am willing to run the models against other "country's" data - but I don't have time to parse & reformat from raw CSV file data to make compatible for incorporate into the models.

If someone wants to supply me with formatted data in Excel from different countries, I'm happy to run it.

Fuel-Plant Type Output as rows, Timestamp (5min increments) as columns is what I am looking for.

[corradio]

@HansHyde what I meant is that you're not using the same electricity mix in your analysis as in the parser (as fuel types are defined and broken down differently) - and that is highly likely be the root of the discrepancy. @jarek in principle it should be taken into account. If you can find an example that fails I'm happy to debug.

[HansHyde]

@jarek CO2 volumes are accounted for accordingly in both models. Rates are rates.

Originally, I wanted to run a baseline CO2 Emissions for the overall Installed Capacity Fleet (commissioning, operational years & decommissioning) weighted accordingly and divided across time to give either a Rate or a CO2 emissions expected "volume", but too much reseach needed to find values. Although, the LCA graphs @brunolajoie discussed re: his PhD thesis might work, and may provide a good reference.

But facts are, this will only INCREASE the SECF Emissions & Rates, not the EM/LCA - further increasing the disparity.

I have to run to German lessons (yeah - not!)

[HansHyde]

@corradio something is "buffering" the EM/LCA calculated Rate, it is not responding as it should be to the CO2 emissions data present.

The disparities are NOT coming from differences in fuel type reporting..

[corradio]

I don't understand that. What rate are we talking about? What does buffering mean? If the differences doesn't come from fuel type definitions, can we please use the same in order.to isolate the problem?

On Tue 9 Jan 2018 at 13:43, Hans Hyde notifications@github.com wrote:

@corradio https://github.com/corradio something is "buffering" the EM/LCA calculated Rate, it is not responding as it should be to the CO2 emissions data present.

The disparities are NOT coming from differences in fuel type reporting..

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/tmrowco/electricitymap/issues/990#issuecomment-356273716, or mute the thread https://github.com/notifications/unsubscribe-auth/ABlEKMg6HtcE9SG41xL01Dx7TQxbwqsDks5tI17vgaJpZM4RXFMh .

[alixunderplatz]

@HansHyde Hello Hans, as a reply to your request for data, please find attached an Excel spreadsheet with some half hourly/hourly data for a few days of February 2017 for Finland and Ireland. Data is from entso-e.

First block of columns is generation, second block is emissions used by EM with "total emissions" as they are calculated by EM. In the third block I had changed emission factor for peat (to 1140 instead of 230). Hope this may help you to present another example :)

Peat in Finland and Ireland.xlsx

[HansHyde]

@corradio (@brunolajoie)

The common unit for all calculations with the EM/LCA method is the kilowatt-second. Not one kilowatt per second, nor one kilowatt per 3600 seconds (1 hour). Do you see the difference between these two?

For all calculations with the EM/LCA method, the common denominator (which is fixed & exact) is used to determine the variables; i.e., 700 g per kilowatt-hour for "Unknown", 820 g per kilowatt-hour for "Coal", 12 g per kilowatt-hour for "Nuclear" or 24 g per kilowatt-hour for "Hydro".

These Variables have an implicit bias against the common denominator.

The bias for coal, oil, natural gas, etc., is greater than 90% of that value IS direct carbon emission from combustion.

The bias for nuclear, hydro, wind or solar, is that less than 95% of the value IS direct carbon emissions from the "fuel". For nuclear, the percentage is somewhere less than 0.0001%.

Increasing or decreasing the variable's value; i.e., Estonia Oil at ~1,100 g per kilowatt-hour is not un-doing the inherent bias, it is only hiding it deeper.

Now, say there are 3,000 MW of hydro and 3,000 MW of nuclear. This would equal over 95% of hydro's operational (produced electricity) carbon emissions in real-time being over-reported, and over 99.9999% of nuclear's operational (produced electricity) carbon emissions in real-time being over-reported.

On the other side of the spectrum, say there are 3,000 MW of gas & 3,000 MW of coal. This would equal 90% of coal's operational carbon emissions in real-time being reported, while 10% of its non-operational emissions are not being reported. This percentage changes based on how far coal's EM/LCA value shifts from LCA median value, and more specifically from the Specific Carbon Emissions of Combustion of Fuel that is likewise an exact value. The same would be true for gas.

And in hindsight, this is what the California data was showing, with a 7.6% variance between EM/LCA and SCEF calculated values, and a ~26% variance for coal. What are those percentage showing? They are showing (approximately) the difference the LCA median value (used by EM) and the SCEF values for the fuels. By increasing the EM/LCA values for a country, you are not getting "cleaner", more accurate data, you are only hiding the inherent bias within the LCA unit of measure - the g CO2 per kilowatt-hour.

This means for a Country rate, were we have a 50/50 high carbon/low carbon generation mix such as NY; 6,000 MW units are heavily biased to over-report emissions (> 90%) and 6,000 MW are heavily biased to under-report emissions (~10%) where 90% represents real direct emissions now, you have a common denominator of 12,000 units to average them and arrive at a "Carbon Intensity" Rate.

Look at either Chart 2 or Chart 3 again. Both show emissions (a volume) calculated in both models increasing in the evening, yet the EM/LCA calculated Carbon Intensity Rate barely reacts. The SCEF Rate increases proportionately, as it should.

Here is your problem, like I have tried to explain prior, the EM/LCA metric is "flawed". The "flaw" is that it has an inherent bias within its common denominator that is being "washed" when one has a roughly 50/50 low carbon/high carbon mix. The bias is multiplied in country's like France or Norway, where there actual CO2 emissions should be lower than reported. And in countries like Poland, EM is under-reporting them (where one can simply increase the coal LCA median value and discount the bias, while making the data look reasonable).

There was one "anomaly" in the resulting data calculations for the two models that was driving me crazy, and I rewrote the calculations 100 times to verify Rates were correct - that my units & conversions were balancing, etc.

That calculation was expected CO2 emissions from the EM/LCA model within the 5 minute increments of time. The EM/LCA model shows even more emissions than that of the SCEF (CO2 Emissions Total) model within the same time period (EM - CO2 MT/5min), yet for the two Rates (Instantaneous), SCEF is higher (404.3) versus EM/LCA (333).

How can Rates be opposite the actual quantities calculated within the time period each Rate is reporting? The inherent bias within the common denominator have contradictory effects of disproportionate bias.

[HansHyde]

Thanks @alixunderplatz - I will see what I can do. Much smaller data set than I am used to working with! :D

[HansHyde]

Hi @alixunderplatz - I will run your data. Can you give me further insights as to what "other" consists of? For example, solar or district heat & power?

Also, regarding Fossil Peat, could you offer me insights as to its 1) Extraction, 2) Processing & 3) Transport (to the power plant) efficiencies might be. Possibly, it is similar for coal mining, as per this "Efficiency of Energy Conversion" model. https://www.ems.psu.edu/~radovic/Chapter4.pdf

Then with each efficiency, also give me (an estimate) of the 2 or 3 forms of energy used to perform each Step.

For example, Coal has an Extraction efficiency of 66%, Processing 98% and Transport 95%. Coal's primary energy (fuel) for Extraction is oil (gasoline or diesel) and Electricity (from coal/nat gas/nuclear).

Thanks

[alixunderplatz]

@HansHyde hey Hans, unfortunately, I have no idea what "other" consists of in that case - exclude it if you like and if it is insignificant.

I didn't think of all of this when supplying the data to you :D Basically, yes, it is very similar to coal (lignite) because it is "mined" from surface mines. But I cannot give you any further infos on that.

[alixunderplatz]

@HansHyde You need a source where it's quite well known how the fuels are extracted and where ideally no "unknown/other" category is given, right? In that case, I'd suggest data for Poland - mainly domestic hard coal and lignite, some pumped storage in peaks and wind penetration up to 20%. If you like, I can support you by downloading some data for you and leave it here in the same format as for the Finland/Ireland data?

[HansHyde]

@alixunderplatz

Frist point, the above is not critical on accuracy, I will model Peat as Coal from an emissions perspective of the fuel. As you can see here, "it comes out in the wash", as each efficiency as multiplicative, while the carbon emissions are additive per unit fuel delivered to the electric generator.

Second, if you were to get me some data, if you could put the times across the top (column header) & each fuel as a row, that would be better. Cells are still MW reported in the rtd.

EM's calculated data is not necessary, as I calculate it with the same method as how it is calculated in the excel file you posted.

[alixunderplatz]

@HansHyde here's something for you ;) Entire hourly data of 2018 so far for Poland, aligned horizontally, pre-sorted ;)

PL_01-10Jan2018.xlsx Hope this will help :)

[HansHyde]

@alixunderplatz Hi Alex,

I am still working through the modeling, improving the full lifecycle calculations of the SCEF, which then allows me to repeatedly confirm the EM/LCA formulas are correctly calculated.

By doing so, I am confirming the formulas used are correct in both models. I then carry forward all the formulas across the timestamps, and look at the displayed results in the Charts (not at the individual timestamp calculated values). Hence, I'm looking at the forest, not the trees, where I can confirm the tree types & the numbers per type are weighted accurately & proportionately represented.

In the SCEF modeling, all of the following LifeCycle Analysis "modes" I can automate, to produce different Rates & Emissions...

1. Emissions from the 4. Power Plant Fuel only
2. Emissions from the 4. Power Plant Fuel, and Emissions from 1. Fuel Extraction, 2. Fuel Processing & 3. Fuel Transport to the Power plant & 5. Emissions from Power Plant Operations (excluding fuel).
3. Emissions from 1. Extraction all the way through to 6. Electric Delivery at a meter.

What the SCEF model does not include (yet), "expected" Emissions from the power plant resulting from A. Construction, B. Physical Plant existence during its operational lifetime, & C. Decommissioning. Although, all would agree this is a very small quantity regardless the fuel or plant type.

The EM/LCA model claims to capture complete LCA Rate & Emission's "volume", hence SCEF modeling 1. --> 6. PLUS A. --> C.

The results I continue to see are;

1. Carbon Intensity Rates - SCEF models, regardless of "mode", are significantly higher than EM/LCA
2. Carbon Emissions ("volume" or mass) - SCEF models, regardless of "mode" are significantly lower than the EM/LCA.
3. Carbon Intensity Rates vs Carbon Emissions - SCEF models; SCEF Rate corresponds/tracks SCEF Emissions "volume" accurately, EM/LCA models; EM/LCA Rate is "buffered" and is NOT corresponding to the EM/LCA Emissions "volume" it calculates.
4. SCEF Rate tracks with the EM/LCA Emissions "volume", EM/LCA Rate does not track with the SCEF Emissions "volume".

SCEF models calculate Emission "volumes" first from Electric Output, then determine the SCEF Rate.

EM/LCA models calculate Rate first from Electric Output, then determine the EM/LCA Emissions "volume". This "volume" would be what you would expect by multiplying the Rate with the Time Interval (of the Rate) in seconds.

The IPCC LCA's values are aggregations of monthly or yearly reported data which is used to support a "theory" that broad trends in emissions from electric generation can be accurately represented at the instantaneous time interval of Electric Generator Output (which is 1 second).

The theory is, data compiled/calculated from monthly or yearly intervals, can be "reverse" applied at an interval of 1 second where 1 second is the unit of the realtime data sourced (Electric Output).

1. Monthly; 1 divided by ( 30 days x 24 hr/d x 3600s/hr ) = 1 divided by 2,592,000 = 0.00000039
2. Yearly; 1 divided by ( 365 days x 24 hr/d x 3600s/hr ) = 1 divided by 1 divided by 31,536,600 = 0.00000003

EM is attempting to apply this "theory", a practical application of the IPCC LCA "theory" through the EM/LCA model.

Analysis of the data, THAT the EM/LCA model IS calculating from instantaneous Electric Output (reported across intervals of varying times (5 min, 1 hour, etc)) is revealing EM's practical application of the IPCC LCA theory contains flaws. The "flaw" is that the theory contains biases within its root constant variables - grams CO2 (by fuel type) per Electric Output Unit (1 kW per second).

There are two biases at play;

1. Variation (difference) from Specific Carbon Emissions from Combustion of Fuels & the EM/LCA gram CO2 eq /kWh, and
2. Variation (difference) from, either Monthly or Yearly modeled Time Intervals, & the Instantaneous Interval of Electric Output reported (1 kW per second).

These biases are carried forward. Hence, the model continues to return data using the biased root constant variables, and the returned data confirms the model, not the theory.

Thus, for both Carbon Intensity Rate and Carbon Emissions, the returned data from the EM/LCA model is confirmation of ANY or ALL biases contained within.

Repeated confirmation of a bias or of multiple biases is NOT confirming theory or factual evidence, it is confirming its own biases. Confirmation Bias shows what one would like to see (or would expect to see), but not the data (or the facts) as they really are.

The EM/LCA model is confirming its inherent biases, not accurately representing modeling without bias or conducted with scientifically verified evidence/calculations.

The numbers are not "lying", but the practical application of the IPCC LCA theory is.

I don't have the time or the technical resources to turn around the visual evidence (Charts) to confirm the existence of the multiple biases, that demonstrate themselves in various forms dependent upon the reported data in different Electric Systems (Parser, Countries, States).

[HansHyde]

@alixunderplatz

The results from the data you provided for Finland over 3 days. Some fuels/plants were not included (didn't sort them yet), but the results were consistent with my NYISO analysis independent of inclusion of all data.

Green Line is the Carbon Intensity Rate that would be calculated from the EM model. Purple Line is the Carbon Intensity Rate calculated from the SCEF model Yellow Line is the difference between the two Rates Black Line is the SCEF calculated Emissions on a fuels-basis (tracking

Output used in model

SCEF Emissions Modeling

EM Emissions Modeling

[HansHyde]

@alixunderplatz I can only surmise, that from the 3rd image I posted above... "Chart 3 - CO2 Emissions Expected - EM/LCA Modeling" that shows imports/exports for NYISO in Output only, is what makes the EM Carbon Intensity Rate so "reactive" (fluctuates high & low) in the EM webpage interface display. Carbon Intensity's for imports/exports are not "balanced" within the Country Parser, they are balanced against the Carbon Intensity Rate of the Country Parser & neighboring Country Parser's Carbon Intensity Rate. These are much different calculations.

[alixunderplatz]

@HansHyde one question: is coal missing in the second and third image (SCEF Emissions Modeling, EM emissions modelling) of your post or did you not include it in the CO2 calculation? EDIT: got it, you wrote you didn't include some :)

[alixunderplatz]

@HansHyde okay, I read through all of this issue here once again and I try to give a brief summary of the content in my words:

1.) carbon-intensive generation's actual emissions are higher than LCA based ones, because the LCA median value is somewhat based on a yearly capacity factor / plant utilization factor (x % of yearly possible gen if a unit ran on 100% of its power through the entire year) (totally not sure if I got this one correctly.)

2.) Using SCEF (Specific Emissions for Combustion of Fuels) assumptions will result in significantly higher instantly/hourly released emissions for carbon-intensive generation (from coal and gas)

3.) Using SCEF assumptions will result in barely any (~<2% of gas emissions) emissions released by non-combusting renewables (wind, PV, hydro) because they are only released during their "assembly".

4.) My interpretation of this image:

4.1) The green line (EM carbon intensity rate) is calculated by electricitymap's approach: (fuel1generation X emissionfactor1 + fuel2gen X ef2 + fuel3gen X ef3) divided by (fuel1gen + fuel2gen + fuel3gen)

For some reason, the green line is not reacting to the increasing/decreasing generation from gas. I assume, the reason is that hydro generation is also increasing/decreasing at the same time, which is keeping a rather "constant" generation ratio between hydro and gas.

4.2) The purple line (SCEF rate including extraction, transport, processing) is calculated using your approach with instant emissions and SCEF which is raising and lowering the g/kWh way stronger. For the considered mix, it is also results in significantly higher total emissions per generated kWh (factor 4-5 higher).

Interim conclusion from what I understand (I think we've had similar words somewhere already):

• using SCEF will represent emissions from fossil electricity generation better, because emissions are actually higher than the used LCA numbers
• using SCEF will nearly blow off emissions from renewable generation, because their emissions mostly comes from "once in a lifetime" emissions during creation, assembly and construction

Did I get it right, Hans? :D feel free to briefly comment and correct my bullet points.

I did not yet consider any Export/Import related stuff, nor did I recalculate anything. If you like and if it's okay for you, drag and drop that Excel file here, so we/I can take a closer look and make sure what numbers the results are based on.

[HansHyde]

@alixunderplatz

Du bist richtig, coal was not included - my bad. I had expanded my modeling to allow "mixes" of fuels like NYISO has (oil & nat gas), so had not yet made a "coal only" category.

But here it is with coal. The results are the same... EM Rate is "flatline", albeit slightly higher with the inclusion of coal.

Finland Generation - Control

SCEF model "Coal : Oil Split" is 100% coal

EM model "EM - Coal" equals 820g CO2eq/kWh

Call me crazy, but the data you sent me showed EM Carbon Intensity Rate for Finland peaked at 255.4 g CO2eq/kWh in the 01:00 - 02:00 hour of the 24th, with all fuels accounted for. The lowest SCEF Carbon Intensity Rates I can find are in the 1,300s, and this does not include "other" or "fossil peat".

[HansHyde]

@alixunderplatz

to answer one of your questions, Hydro at 24gCO2/kW would require 34.2x the Output to "balance" 1 Coal at 820gCO2/kw. In other words, 34.2 GW of Hydro would be required to offset 1 GW of Coal. The above chart with coal shows an 8x difference in Rates. The EM Carbon Intensity Rate calculation is not only wrong, but it cannot function with any level of consistency at any interval of time other than 1 year.

Take an entire Country's fleet knowing its Installed Capacity, its Capacity Factors & the GWhr each fuel/plant type generated in that year, and yes EM might be able to use the IPCC LCA median gCO2eq/kWr values to arrive at a fair approximation of CO2 emissions (both direct & indirect).

But in "realtime" the practical application is failing.

[alixunderplatz]

@HansHyde First, I want to make sure which axis belongs to which data in 2nd and 3rd image. The surfaces to the left vertical axis and line graphs to the right vertical axis? Am I right?

Something around 1,300 g/kWh shoudn't even be possible for the entire mix. Some of the worst emission stuff I could find for burning low quality lignite with an efficiency under 30% is in that range of specific emissions. Not for good quality hard coal (used in Finland). Are you sure you did divide the CO2 emissions properly? Even with mining, processing and transport included it seems unbeliavable high.

Here are the specific CO2-only (not CO2eq) emissions for the worst electricity source of Germany, lignite power plants of 2015 (took the now inactive units out, source: energy-charts.de):

Estimating, that the other emitted greenhouse gases burning lignite will not raise the emissions from operation beyond 1,300, I think something is not converted properly in your data. Even though this might be data averaged for the year (yearly emissions divided by yearly generation), a source like coal or lignite should not be able to pull the entire generation emissions above a well known "maximum value" (like for German lignite 1,000 to 1,300 g/kWh on average).

If I assigned both axis (axisses? whats the correct plural? :D axes?) properly, something is weird with your "CO2 emissions in tonnes per 5 minutes". Right now Germany's coal (generation 33 GW) is emitting 400 t CO2eq/minute, so about 2,000 t/5minutes according to EM model. The same is almost released by the red surface for natural gas in the center image for Finland, generating ca. 1 GW only.

[HansHyde]

@alixunderplatz if you can find me a better number for Heat Rates of German lignite burners than the ones I have from the US EIA average Heat Rates by plant type, which is consist across all steam or gas turbines for 15+ years running, I would be happy to try them out.

As to German lignite, specific carbon emissions from combustion of any coal type vary only a small amount from each other. I have also weighted them based on their consumption proportions in the US for 2010.

As coal & nuclear plants in the US average ~40 years of age, and Germany rebuilt its coal fleet in the early 2000s, I suspect the German's might have SuperCritical technology giving them a better Heat Rate than the US fleet. I would be happy to try,

[HansHyde]

@alixunderplatz - for the SCEF model Chart, numbers reported are MT per hour. I did not change the Chart primary axis title to reflect that the model was running calculations based on an interval of time = 1 hour. Not 5 minute intervals for NY's data.

[HansHyde]

@alixunderplatz

And with peat, side by side.

And PS, don't forget Rates on the right are not the same as the values inputed into the Parser. The Rates are in kilowatt-second.

[alixunderplatz]

@HansHyde How do I convert the "kW-second" rates to g/kWh rates (for comparison)? I still have issues understanding this unit properly. Is kW-second equal to "kWs", so "kWh/3600"?

[HansHyde]

@alixunderplatz .... http://www.energylens.com/articles/kw-and-kwh

[HansHyde]

I'll back track my numbers and calcs, delete a bunch of "copied" workbooks & worksheets and verify all is in order.

But looking at IPCC 2014 Climate Synthesis Report, I can't for the life of me understand how EM arrived at these "default" values, when lifecycle of coal is north of 1000, lifecycle of gas is around 620 and lifecycle of oil is not even listed (yet it is much closer to coal than gas in emissions at the power plant).

And IPCC has no direct emissions from biomass, yet landfill gas gets assigned as methane???

[alixunderplatz]

@HansHyde :DDDD that link absolutely made my day ... hahahahah .:D ... Just to clarify: I am totally aware of the difference between "power" and "energy" and power being a rate of "energy per time" :D

It is just that I can't seem to understand whether the "second" in your kilowatt-second ...

1. is an index (usually written a bit lower next to the actual unit) or
2. belongs to "kW" as "factor" so it is an energy unit "kWs" or "1000 Ws" (like "kWh") or
3. is the power in "kW which is present during a second (as a result producing a certain amount of energy during that intervall of one second)

The "second" is my problem, nothing else. i'd like to put your calculations into a relation to what I am used to work with (kW, kWh, ...). The kW-second is an "unusual" unit you came up with and - for me - it is causing trouble in comparing things to it.

[HansHyde]

@alixunderplatz

1 Joule = 1 Watt per second 1 Watt per second = 1 Joule 1 kW per second = 1,000 Joules 1 MW per second = 1,000,000 Joules

These are "vertical" calculations so to speak, with a width of 1 second.

Rates I am familiar with are 60 Miles per Hour 100 kiloMeter per Hour Heat Rates - the inverse of the efficiency of a power plant and TBH, I would be very careful when you pull an "efficiency" claim out of literature. Is it the Capacity Factor of the plant (MWh divided by (nameplate rated output x 8760 hours of the year)) or is it the Heat Rate of the plant itself? Two different things.

"grams per kiloWatt per hour" - is that some IPCC invention? I kind of think it is...

If I turned on a 1000 watt heater for 1 minute, I would be furious if my utility company sent me a bill for 1 kilowatt hour because they charged me for a Rate.

This is some sort of "horizontal" calculation... across time like Albert Einstein. 🔢

Any Outputs we are getting from a grid operator are effectively Megawatts per second (a "vertical" calculation).

Hence, 1 "kilowatthour" = 3,600,000 Watt-seconds or 3.6million Joules

[HansHyde]

@alixunderplatz

I "cleaned" my model making sure all was in order, added additional capacities, and re-ran numbers.

Again, with NY (a roughly 50/50 mix of low carbon & higher carbon), the EM model Rate under reported emissions (yellow line shows the differences in Rates) and was "unresponsive" to highest carbon generation.

Again, with the Finland data, the EM model's Rate is all over the place. Going up when it should be going down, and down when it should be going up. To be honest, it can't calculate an emissions Rate to save its life with any level of consistency.

I outlined this in great detail to Bruno & Oli 3 days ago and I have spent way too many hours building & testing models in a concerted effort to root out real "fixes" to improve the EM model. They have been unresponsive.

Please excuse that I did not reformat the charts for Finland's lesser quantity of timestamps... all is in order. Rates on the right are equally based on both charts (regardless what "unit" we give them.

The "Stacked Bars" are incremental volumes of emissions, the lines are Rates. I don't know why, but when I try to change to 1 hour increments (from the 5 min it was built for), it "breaks" the model. My guess... it is the same reason taking rates based on yearly (monthly at best) values and constructing "averages", then attempting to apply that in realtime, "breaks" the model.

PS, don't worry, no data was "double" reported - I have run out of colors. All data you gave to me was run. The "Other" was ran through as Oil in my model, and 700 "other" in EM's model. At the power generator, oil clears my model at about the same rate in proportion to coal, as "other" would be in proportion to coal in the EM model.

[brunolajoie]

Again, all this boils down to 2 potential improvements we could make to the EMap. Improvements does not mean the initial model was wrong or falted. You've got to start with something: The least wrong something.

1. Move away from IPCC and use instead "country specific" carbon-intensity factor per fuel. This is WIP in #738 and i'm planning to try to use ecoinvent database. IPCC was used at first because if you have to choose one and only one number for all countries, the IPCC median value was the least wrong.

2. Break down Lifecycle analysis carbon-intensity figure (gCO2eq/kWh) between Operational + Non-Operational factors. 2.1 Operational factors (gCO2eq/kWh) could be multiplied by the real time MW output to give operational emissions. This is the variable part of the country carbon intensity. 2.2 Non-operational factors (if we have them in gCO2eq/kW) could be multiplied by the capacity installed in the given country, to output a "fixed", non-variable part of the country grid carbon intensity.

We're well aware of that, and only missing a standardized database of LCA/Operaional/Non-Oerational emissions factor per country. We need a standardized database in order to compare country with fair underlying assumptions.

[HansHyde]

@brunolajoie

The IPCC number is a longitudinal factor, across time. The "realtime" data output is a horizontal factor. You are mixing them up with the EM model and this is exposing serious biases within your calculations and the information that is being displayed.

Why you would go with "country specific" carbon-intensity factor per fuel is beyond me. There are Specific Carbon Emissions of Combustion of Fuel that are universal, regardless of country. It is the power plant's Heat Rate (the inverse of its efficiency) that determines its Output (MW) per unit fuel it uses. And it is the Heat Rate that varies broadly depending upon the type of Prime Mover in the power plant; i.e, steam boiler, gas combustion turbine, gas combustion turbine with HRSG & steam turbine. This is how a NG steam boiler might get only half the Output per unit fuel as a Combined Cycle plant.

If you want to go "country specific", do so based on the composition of the Prime Mowers; i.e., 60% NG Combined Cycle, 20% NG Steam Boilers, 10% NG Gas Turbines, not some barely distinguishable variation in the primary fuel type. Do you really think you are going to obtain scientifically derived values for energy content of two different NG streams in two different countries? Even the Specific Carbon emissions difference of the 3 primary types of coal combusted in power generation is less than 2% between them, where the Heat Rate (technology used) of the plant determines how much CO2 emissions per unit Output it produces. If Germany is using Super or Ultra Critical boilers with "dirty" lignite, while Bulgaria or Poland is using traditional boilers with subbit coal, who do you think's Output is "dirtier"? Which impacts more, the <2% specific carbon emissions of fuels or the up to 15% (30-->45%) efficiency differences of the Prime Mover?

It is honestly not rocket science to "reverse" engineer using average heat rates & specific carbon emissions and peg this against realtime output. But it is absolutely taking a "walk-of-faith" calculating as EM is currently, especially as unfortunately, no one seems to be comprehending how this calculation method is failing EM --> plus, it is turning out results that seem plausible, but are horribly inconsistent.

Unfortunately, cleaning up data reporting irregularities (timestamps other than on 5 minute intervals) took me over 10 hours to do, across over 40,000 timestamps of data from NY and I'm giving up on Week 2 after 4 hours, but here is longer longitudinal analysis of the data, where you can see EM's Carbon Emissions rate is being "buffered". The biases in the calculations are working differently at different times giving a mostly flat, "unresponsive" plotting of the Carbon Intensity Rate against that which would be expected.

Look at the Pink Lines... this is the Carbon Intensity Rate EM is calculating if Imports/Exports are not included in the secondary balancing between Country Parsers calculations. How does that represent either models' emissions.

Look at the Yellow Lines... this is the difference in Carbon Intensity Rates. The electric output data used is the same for both models. On the 5th day's peak, EM is off by over 75%! And this is with the SCEF model only calculating for direct operational emissions against EM's full LCA emissions. I add Extraction, Processing & Fuel Delivery to SCEF, that peak is over 700 grams while EM is around 370/80 grams.

There is something significantly wrong with the EM calculation methodology of Carbon Intensity Rate.

[brunolajoie]

When I said "coutry specific carbon intensity factor per fuel", I meant "country specific carbon intensity factor per type of power plant that we have on the EMap, namely, coal, gas, oil, wind, solar, goethermal, nuclear, biomass and other". There are obvious differences between countries that depends on the specific type of power plant (age, sub-categories such as OCGT vs CCGT etc..., quality of fuel used, rampups/down patterns, typical heat rates/efficiencies and even annual load factors as for LCA values) installed in each countries, and these differences are important and analysed by various studies such as described in #738 . These differences are based on historical analysis. As already said numerous time, we won't do it ourself, as we don't have the manpower nor the scientific legitimacy to compute these figures for each country. We need an external, standardized database, peer reviewed or highly reputable, to refine our model.

[HansHyde]

@brunolajoie

If you guys want to test this, then the easiest way would be for @jarek to turn off Import/Exports to NYISO from Ontario and @systemcatch to turn off Import/Exports from NEISO and look at NYISO's Carbon Intensity Rate as calculated by EM and how it responds during 24 hrs as the mix varies. Let's not forget, the area under "Carbon Intensity in the Last 24 Hours" is NOT a quantity of emissions, it is a colored scale of the Rate. These are not the same thing. From the EM Carbon Intensity Rate, one cannot calculate the actual (expected) emissions.

Now, take the 2 charts above and pick a time, say the first 20:00 for NYISO's 1st Week of 2018.

Calculate the emissions actual in both graphs, using numbers on the left scale & the stacked column values.

SCEF shows ~1,700 metric tonnes CO2 per 5 minutes. EM shows ~7,850,000 grams CO2 per second.

Do the math on either to get a common denominator of time, either 5 minutes or 1 second, and this is your quantity of emissions per equal unit of time at 20:00.

Now compare that to both the EM and the SCEF Rates which are about 330 & 500 respectively at 20:00.

What doesn't match?

The stacked bar values on the EM Chart represents emissions that would be expected if EM applied the IPCC "grams per kWhr" to the Realtime Generation Output, but EM is not doing this. It is only calculating a rate using the IPPC values & output of each type.

If the EM Rate is not even matching (responding to) the actual (expected) emissions the IPCC rates would anticipate if they were calculated, what is it measuring? And more importantly, where is the check on it? There is none. And even worst, the values it is turning out somehow seem "plausible" even though the source data tells us something completely different.

[HansHyde]

I suspect this is going to sting a bit, but here goes...

Equal unit of time basis at 20:00 on Jan 1st NYISO with ~20,460 MW of reported generation within the state - no imports/exports

Carbon "Intensity" Rates

1. SCEF - 500
2. EMap - 330

Calculated Carbon Emissions

1. SCEF model - 1,700 MT CO2 per 5 minutes --> 5,667,000 grams CO2 per second
2. EMap (modeled) - 7,850,000 grams CO2 per second --> 2,355 MT CO2 per 5 minutes

How can EMap report a Rate LOWER THAN SCEF, when the Carbon Emissions calculated using either the ...

1. the Specific Carbon Emissions of Combustion of Fuel (direct operational only) model, or
2. the IPCC median "grams CO2eq/kWh" for fuel/plant type LCA calculated emissions ARE HIGHER???!!

The SCEF carbon emissions Rate (purple line) matches the SCEF calculated emissions, as one would expect. The SCEF carbon emissions Rate (purple line) tracks with the carbon emissions not calculated on EMap, but that would be expected from using the IPCC values it is using. The EMap Rate does NOT track with the calculated emissions from either IPCC values or SCEF models.

It is not just that EMap's Rate is some percentage (higher or lower) different than the SCEF Rate consistently, it is that EMap's Rate is independent of all other modeled results. It is operating of its own accord and it is consistently inconsistent with all other data inputs or calculated outputs.

Before, I said the EMap model of IPCC theory was "simplistic". It is not only that it is simplistic, but the evidence continues to show the calculations & the fundamental Maths are flat out incorrect.

When multiple people here cannot consistently understand the fundamental metric being used, what a "grams per kWh" actually is or is not, and cannot easily reverse calculate fuel consumed or carbon emissions expected from an Power Output rate in the data EMap is using to make its calculations, there is something seriously suspect in the modeling.

The modeled Maths are wrong. More specifically, EMap's fundamental Carbon Intensity Rate's algebra is incorrect.

Whether EMap uses IPCC median values, country specific values or Specific Carbon Emissions values, does not matter when the fundamental algebra is incorrect.

And if this is the fundamental algebra equation/methodology EMap has been using from Day 1, with its first Carbon Intensity Rate displayed, then every time/country stamp that was calculated & recorded/displayed is equally wrong.

Which makes this... https://pro.electricitymap.org generated data that can NOT be independently verified or substantiated for accuracy or consistency.

And going further, makes pages 25 thru 53 of this report, using EMap's "exclusive" data, unable to be independently verified or substantiated. http://energyforhumanity.org/wp-content/uploads/2017/11/European_climate_leadership_report_2017_WEB.pdf

Here the report claims to be able to calculate & rank countries by "total carbon emissions from electricity generation", yet as demonstrated above, EMap does not calculate carbon emissions... and if it did, its Carbon Intensity Rate cannot be used to calculate the emissions that would be expected from the EMap Rate.

If I am not mistaken, unless domestic versus foreign generation is weighted accordingly, they should not be stacked, but side by side.

My perspective of that report does not matter. What does matter is that is used data from EMap that cannot be independently & consistently verified for accuracy, and then used that data to strengthen/substantiate the position it was making, then produced a report for public consumption.

At this point, it appears I will need to step up my data-analysis game with different chart types. Any guesses on what the above charts will look like when standard deviation is plotted?

The (EXTREMELY) unfortunate reality is, this is (UNBELIEVABLY) easy to fix moving forward...

[HansHyde]

@alixunderplatz

From the data you sent from Finland including timestamps for 72 hours (3 days), the following was determined, run equally across all timestamps with Outputs per Fuel as provided. I did not use any fancy models or calculations, only the numbers within the data set.

The Error of Actual Emissions Rate divided by EM's "Carbon Intensity Rate" across 72 timestamps...

1. Lowest value = 738%
2. Highest value = 992.6%

When 1,140 g/kWh replaced 230 g/kWh for "Fossil Peat" as you were suggesting, The Error of Actual Emissions Rate divided by EM's "Carbon Intensity Rate" across 72 timestamps...

1. Lowest value = 934.6%
2. Highest value = 1331.2%

The percentages are as reported, decimal placements are accurate.

If the 1,140 g/kWh for Fossil Peat was implemented for Finland, the data would suggest this made the margin of error increase if all other values are were not modified.

[brunolajoie]

Hello Hans,

• I'd really like to understand what message you are trying to tell us about algebra being wrong, but it's not concise enough for me to make sense of it in a Github discussion. Could you pinpoint in one or two sentences what "specific algebra" is wrong in our computation, using the same vocabulary and units we use in the EMap project ?

• You can perfectly stack domestic + imported components in the carbon intensity of a given country grid, to my point of view. Why not? Take Austria: domestic production had a carbon intensity of 100gCO2eq/kWh (mostly hydro), but they were importing a lot of german coal, so that from a consumer perspective, overall carbon intensity was 200gCO2/kWh. 100g/kWh local + 100g/kWh imported. Why wouldn't it sum up to you? Remember that we take the point of view of consumers. As for 'carbon rate (in tCO2/min)': It consists of local emissions, plus imported ones, minus exported one. The export component here is important to avoid double-counting. What's the big algebra problem with all that?

• Also, don't worry, our model has been set up to allow back in time computations.