中文版 · Blockchain learning guide · Previous article
Blockchain learning · Article BC097
A blockchain energy claim becomes useful when you can reconstruct what was counted. Start with the equipment and observation period, reconcile the electricity records, and only then examine emissions factors and activity denominators. A smaller displayed number can result from excluding infrastructure rather than improving it.
This lesson follows one fictional catalogue service through three reports that initially appear inconsistent. You will reconcile consensus machines, RPC access, indexing, networking and facility overhead; calculate a separate emissions estimate; and test whether a lower per-record ratio establishes an absolute reduction. Every numerical value is invented for learning. None describes a real network, Chinese electricity supply, environmental certification or investment.
The practical outcome is a short, defensible report whose boundary travels with its number. You do not need a universal ranking of blockchain technologies to do this work.
Choose your reading route
In this guide
- 1. Name the system before accepting its total
- 2. Read the five electricity lines
- 3. Explain why the reports say 48, 80 and 110
- 4. Test a boundary change honestly
- 5. Calculate emissions as a separate result
- 6. Keep emissions and electricity changes distinguishable
- 7. Choose the activity denominator explicitly
- 8. A falling ratio can accompany a growing total
- 9. Handle uncertainty without manufacturing certainty
- 10. Write the report a reader can check
- 11. Exercise: approve, qualify or reject
- Frequently asked questions
- Sources and further reading
Read sections 1–4 to resolve the electricity reports. Sections 5–8 cover emissions and denominators. Use sections 9–11 for uncertainty, the reporting exercise and model answers, then the FAQ and sources for further study. Readers wanting the wider accounting framework can first read How Carbon Accounting Works.
Back to contents · Next section
1. Name the system before accepting its total
Our fictional service records catalogue updates on a small permissioned blockchain. Its reporting objective is the operational electricity used by a specified service equipment set during one complete 24-hour observation window. It is not an inventory of a company, a whole industry or every device capable of querying the chain.
The included set comprises consensus machines, dedicated RPC machines, a dedicated indexer, dedicated network equipment and an explicitly assigned share of facility overhead. These support different jobs. Consensus equipment participates in maintaining the ledger; RPC machines answer application requests; the indexer creates searchable views. Removing an item from the spreadsheet does not remove its role from the service.
For this exercise, all quantities are supplied as reliable, aligned measurements or explicitly approved allocations. There are no overlapping submeter readings within the five included categories. The service performs no other task on those dedicated machines. These assumptions make the arithmetic decidable; a real reviewer would need evidence for them.
End-user devices, equipment manufacture, transport, replacement and disposal are excluded. Any wider internet infrastructure beyond the listed dedicated network equipment is also excluded. This report therefore cannot establish a whole-life environmental footprint. An excluded category is outside the stated calculation, not demonstrated to be zero.
Back to contents · Next section
2. Read the five electricity lines
The common 24-hour window contains these quantities:
- Consensus machines: 48 kWh.
- Dedicated RPC machines: 12 kWh.
- Dedicated indexer: 6 kWh.
- Dedicated network equipment: 4 kWh.
- Assigned facility overhead: 10 kWh.
Adding them gives 80 kWh. The equipment categories sum to 70 kWh before the separately supplied overhead. Do not apply another overhead multiplier to the completed 80 kWh: this dossier has already included the assigned overhead once.
The U.S. Energy Information Administration’s units explanation distinguishes power from electricity used over time. Here kWh is an energy quantity over the observation window. Dividing 80 kWh by 24 hours gives an average power of 10/3 kW, about 3.33 kW. It does not reveal the peak load or show that demand was constant.
A label reading “80 kW used yesterday” would need correction. It confuses a power unit with the supplied daily energy total. Conversely, a single instantaneous power reading cannot reconstruct yesterday’s consumption without information about how power varied through the day.
Back to contents · Next section
3. Explain why the reports say 48, 80 and 110
Report A says “consensus electricity: 48 kWh.” Its narrow number agrees with the consensus line. If A is clearly labelled, it need not be erroneous. However, relabelling it “complete service electricity” would omit 32 kWh of included supporting operations and overhead.
Report B says “specified service operational electricity: 80 kWh.” It includes all five categories and publishes the exclusions. That is the answer to our reporting objective.
Report C shows a shared rack and associated facility allocation totalling 110 kWh. The dossier independently assigns 30 kWh of that aligned total to unrelated services and 80 kWh to our catalogue service. The allocation is assumed reliable and exhaustive for this example. Subtracting 30 from 110 reconciles C with B.
Do not add C to B. C already contains the energy attributed to the catalogue service. Doing so would count it twice. Nor should an analyst subtract 30 merely because that makes the numbers agree: the independent allocation evidence is a necessary premise, not an arithmetic convenience.
This reconciliation produces a useful correction request. Ask A to keep its consensus-only label, B to retain its category list, and C to show the shared-service allocation. There is no need to accuse a report of fabrication simply because its scope differs.
Back to contents · Next section
4. Test a boundary change honestly
Suppose a presentation compares A’s 48 kWh with another service’s complete 80 kWh and announces a 40% saving. The subtraction is arithmetically correct: the difference of 32 is 40% of 80. The comparison is not supported as a complete-service saving because the boundaries differ.
This matters particularly when a blockchain application outsources access and indexing. A project might operate few machines itself while relying on substantial external infrastructure. “Our office meter” and “the equipment supporting our service” answer different questions. The fictional report follows its specified service boundary regardless of which team pays an invoice.
A repaired comparison needs aligned categories, time windows and service requirements. Even then, equal numbers of records do not guarantee equivalent functionality, availability or verification responsibilities. The present dossier contains only one service, so it cannot establish that a competing architecture would use more or less electricity.
A consensus label alone also cannot supply the missing measurement. This lesson does not estimate any real proof-of-work or proof-of-stake network. Protocol design is relevant background, but the actual reporting claim still needs an explicit equipment model or measured boundary.
Back to contents · Next section
5. Calculate emissions as a separate result
The dossier divides the complete 80 kWh into two non-overlapping electricity groups. Group X supplies 60 kWh with an invented applicable factor of 0.4 kg CO2e per kWh. Group Y supplies 20 kWh with an invented applicable factor of 0.7 kg CO2e per kWh. Both factors are stipulated to cover the same limited operational-electricity emissions definition required by this exercise.
The result is 60 × 0.4 + 20 × 0.7 = 24 + 14 = 38 kg CO2e. Electricity remains 80 kWh. The emissions calculation translates those energy quantities under supplied factors; it does not change how much electricity the service used.
The weighted factor is 38/80 = 0.475 kg CO2e per kWh. Taking the simple average of 0.4 and 0.7 would give 0.55, incorrectly weighting unequal energy groups equally. Applied to 80 kWh, that would produce 44 kg CO2e rather than 38.
These factors are not official national values, market-based claims or a complete lifecycle model. GHG Protocol’s Scope 2 Guidance addresses accounting for purchased energy and requirements concerning contractual instruments. Our small calculation does not demonstrate compliance with that guidance. A production report would need the applicable methodology and evidence for its factor selection.
Back to contents · Next section
6. Keep emissions and electricity changes distinguishable
Consider an independent factor-only variation. Hold electricity at 80 kWh and both group quantities unchanged, but supply a revised valid factor of 0.5 for Group Y under the same exercise definition. The revised result is 24 + 10 = 34 kg CO2e.
That is a reduction of 4 kg CO2e, or 4/38 of the original estimate, approximately 10.53%. It is not an electricity saving. The machines’ supplied consumption did not change. The example establishes the effect of the changed factor within this model; it does not explain why a real factor changed or demonstrate a causal effect of buying a certificate.
Conversely, a lower electricity total need not produce a proportionally lower emissions estimate if the applicable factors change. Keeping both result columns visible prevents one quantity from silently standing in for the other.
No offset, removal or avoided-emissions credit is supplied. None may be invented to reduce the 38 kg CO2e baseline. Likewise, the absence of embodied-equipment estimates does not license a “zero lifecycle emissions” conclusion.
Back to contents · Next section
7. Choose the activity denominator explicitly
During the same baseline window, the service receives 1,000 submitted catalogue requests. Exactly 800 distinct requests achieve successful application execution and the supplied finality criterion before the window ends. The other 200 do not meet that combined endpoint within the window; their further outcomes are not supplied.
Using all submissions gives 80/1,000 = 0.08 kWh per submitted request. Using successful-finalized requests gives 80/800 = 0.10 kWh per such request. Both quotients can be reported if their labels remain visible. Neither number proves the incremental electricity caused by adding one request.
The corresponding emissions allocation using the successful-finalized denominator is 38/800 = 0.0475 kg CO2e per counted request. It is an allocation of the entire supplied period total across that cohort, not a physical measurement at each transaction. The service may also expend energy on unsuccessful requests, maintaining readiness and serving reads.
A claim that the remaining 200 were permanently rejected goes beyond the dossier. Some could still be pending; we have not been told. A count of submitted requests also cannot replace a count of accepted records when a report promises successful completion.
Back to contents · Next section
8. A falling ratio can accompany a growing total
In a separate next-day variation, the same equipment-category boundary uses 100 kWh over another complete 24 hours and records 2,000 successful-finalized requests. Assume the same endpoint definition. The average allocation is 100/2,000 = 0.05 kWh per counted request.
Relative to the baseline 0.10, that ratio falls by 50%. Yet absolute electricity increases from 80 to 100 kWh, a 25% rise. Both statements are true. Neither should be hidden when assessing the report.
We do not have next-day electricity group quantities or factors, so we cannot compute next-day emissions. We also lack a controlled experiment that isolates the cause of the changed intensity. More activity sharing ongoing operational loads is one possible mechanism, but the numbers alone do not identify it.
This is why reporting a service’s aggregate electricity, output count and defined ratio together is more informative than displaying only the most flattering percentage. It preserves the evidence needed for readers with different questions.
Back to contents · Next section
9. Handle uncertainty without manufacturing certainty
Replace the baseline overhead point estimate of 10 kWh with a supplied possible interval of 8–14 kWh, leaving the other 70 kWh fixed. The complete energy total is then 78–84 kWh. Dividing by the fixed 800 successful-finalized requests gives 0.0975–0.105 kWh per counted request.
This interval is a sensitivity range from stipulated possible overhead values. It is not a statistical confidence interval and carries no probability statement. Do not attach “95% confidence” because the endpoints look formal.
We have not assigned the uncertain overhead change between Groups X and Y, so this variation does not establish a corresponding emissions range. Multiplying both endpoints by the baseline weighted factor would introduce an extra allocation assumption. State it if you choose to explore it; do not treat it as supplied evidence.
The Cambridge Bitcoin Electricity Consumption Index methodology illustrates why estimates need visible assumptions: its model provides estimates based on equipment and other assumptions and distinguishes power demand from annualised electricity. An annualised value is not automatically the electricity actually observed over the previous year. We borrow that methodological caution, not a current network figure or its model parameters.
Back to contents · Next section
10. Write the report a reader can check
A defensible baseline statement is: “During the specified 24-hour window, the included consensus, RPC, indexing, dedicated networking and assigned facility overhead totalled 80 kWh. Applying the two supplied operational-electricity factors produced an estimate of 38 kg CO2e. There were 800 successful-finalized catalogue requests, giving an average allocation of 0.10 kWh per counted request.”
Follow it with the exclusions and shared-meter reconciliation. Identify the source records, allocation version, factor definitions and endpoint rule. Explain that the figures are synthetic if presenting this classroom exercise. Do not relabel them as measured results from a live eduKateSG network.
A short evidence pack should allow a reviewer to reproduce the five-line sum, the 110 minus 30 allocation, the two-factor emissions calculation and the cohort count. If source records are digitally signed or anchored to a chain, that can help establish specified record integrity under the relevant verification assumptions. It does not independently establish meter calibration, complete equipment coverage or the appropriateness of an emissions factor.
Back to contents · Next section
11. Exercise: approve, qualify or reject
Evaluate these claims using only the baseline dossier unless a variation is named.
- “The complete specified service used 48 kWh.”
- “The 110 kWh shared total includes 30 kWh attributed to unrelated services.”
- “The baseline emissions estimate is 38 kg CO2e under the supplied factors.”
- “Every extra catalogue request necessarily consumes 0.10 kWh.”
- “The next-day variation halves average electricity per successful-finalized request and reduces total electricity.”
- “The uncertain-overhead variation gives an energy range of 78–84 kWh.”
- “A valid chain anchor proves the report includes every relevant machine.”
Model answers: reject claim 1, because 48 covers consensus only. Approve claim 2 under the explicit reliable allocation assumption. Approve claim 3 with its factor and scope qualification. Reject claim 4: a period-total allocation is not a marginal-consumption experiment. Qualify claim 5: its first half is correct, but total electricity increases 25%. Approve claim 6 as a stipulated sensitivity range, without a confidence level. Reject claim 7: record integrity does not establish boundary completeness.
For an extension, use the factor-only variation. The correct answer is 34 kg CO2e with electricity still 80 kWh. Explain why an electricity-efficiency claim would be unsupported even though the emissions estimate is lower.
Back to contents · Next section
Frequently asked questions
Is electricity per transaction always meaningless?
No. A clearly labelled period-average allocation can answer a defined reporting question. Problems arise when its denominator is hidden or it is presented as the causal electricity cost of one additional transaction.
Can we compare two protocols using this fictional service?
No real-network ranking follows. A useful comparison would need aligned service requirements, equipment boundaries, observation periods and credible measurements or models for both systems.
Does outsourcing remove energy from the service?
It can change an organisation’s reporting relationship to the activity, but it does not make the supporting equipment disappear. Follow the chosen boundary and applicable accounting method rather than the location of an invoice.
What should we learn next?
Use the carbon-accounting prerequisite for scopes and lifecycle methods. Then apply the same evidence discipline when designing a pilot: decide in advance what resource boundary and completed-service endpoint its acceptance test will use.
Back to contents · Next section
Sources and further reading
- EIA, Measuring electricity: power and energy units; checked 12 October 2026.
- GHG Protocol, Scope 2 Guidance: the 2015 guidance page, which also records the consultation period ending 31 January 2026; checked 12 October 2026. Consultation is not treated here as an adopted replacement standard.
- Cambridge CBECI methodology: model assumptions, power demand and annualised electricity; checked 12 October 2026. No live network totals are used.
- How Carbon Accounting Works: the broader accounting prerequisite.
The essential habit is simple: preserve the equipment boundary, time window, factor definition and denominator whenever an energy number travels.