Constraints, congestion and loss factors

The question

Does the cheapest answer satisfy every limit?

Mechanism

Use the cheaper remote plant until the first applicable limit is reached: A = 60 MW. The local plant supplies the residual B = 40 MW. Energy over the hour is 60 MWh plus 40 MWh. Total offer-based cost is 60 × 20 + 40 × 80 = A$4,400. Check balance, non-negative output, each plant's capacity and the export limit separately.

Without the export restriction, A = 100 and B = 0 gives A$2,000 under the same assumptions. The A$2,400 difference is this model's additional production cost caused by the restriction. It is not a generator's lost profit, a consumer bill, a network tariff or a congestion rent. Those require prices, settlement boundaries, ownership and contracts that this model does not contain.

Create a spreadsheet with inputs, decisions, constraint slack and total cost in separate rows. For each ≤ constraint, calculate RHS minus LHS; a negative result fails feasibility. Keep the energy balance residual at zero. If your model selects A = 100 while retaining a 60 MW export limit, the cheap answer is infeasible rather than efficient.

Worked numbers

At 60 MW transfer, A=60, B=40 and cost is A$4,400 for one hour.

Try it in the workspace

Raise demand to 140, then 180 MW.

The question

What is one extra MW of capability worth?

Mechanism

Relax only the export limit from 60 to 65 MW. The new dispatch is A = 65, B = 35 and cost is A$4,100. The saving is A$300 for 5 MW of added transfer over one hour, or A$60 per added MW for this one-hour case. Equivalently, each additional transferred MWh displaces an A$80/MWh offer with an A$20/MWh offer while both remain on the same marginal segments.

At 100 MW export capability, demand is fully served by A. Raising capability to 105 MW then saves nothing because neither demand nor A's capacity has increased. This is why a local sensitivity cannot be multiplied by unlimited capacity or every hour of a year. Repeat the calculation across plausible demand, offer and outage conditions before evaluating an augmentation.

AEMO's marginal value describes an objective-function response to an RHS change; its documented sign depends on the inequality. Our positive saving is not a copied market-data field. Check the equation orientation, coefficients and reported convention before comparing values. In mathematical optimisation, equality at a limit alone need not establish a unique non-zero economic value, particularly with redundant constraints.

Worked numbers

RHS 60→65 saves A$300; 100→105 saves zero.

Try it in the workspace

Compare a +1 RHS perturbation before and after the limit stops binding.

The question

What does a coefficient mean?

Mechanism

Consider the separate teaching inequality 0.5G1 + G2 ≤ 100 MW. With G2 = 70 MW, G1 can be at most 60 MW, assuming no other limits. An extra 20 MW of G1 uses 10 MW of this constraint's headroom. Coefficients are sensitivities in this illustrative equation, not ownership shares or a statement that half the plant's electricity disappears.

Multiplying every term and RHS by two gives G1 + 2G2 ≤ 200. It has the same feasible set but a one-unit change in its RHS is only half as much physical relaxation as before. Compare marginal values only after accounting for scaling. In real analysis, read the constraint description, affected equipment, invocation period and formulation, not just the identifier.

Return to our two-generator model. At demand 140 MW, A = 60 and B = 80 is feasible. At demand 180 MW, the available transfer plus local capacity totals only 160 MW; there is no solution meeting all the stated conditions. Report the 20 MW shortfall. A solver's penalty or relaxed constraint does not create physical supply or prove an acceptable operating outcome.

Worked numbers

0.5G1 + G2 ≤ 100 with G2=70 permits G1≤60.

Try it in the workspace

Use Scaled equation, then check the unchanged feasible set.

The question

Whose value changed?

Mechanism

Reduced output may reflect a binding network limit, resource availability, a plant outage, an offer decision or several factors together. Compare actual and target output, available capability, relevant constraints and timestamps before calling every shortfall curtailment. A technical connection limit and a dispatch constraint are related but different evidence; use Topics 36 and 41 to examine the connection context.

Do not interpret a marginal loss factor of 0.95 as proof that exactly 5% of annual generated electricity was physically lost. Physical energy balance and marginal settlement treatment answer different questions. For an explicitly simplified illustration only, 10 MWh × A$100/MWh × 0.95 = A$950; changing the assumed factor to 0.90 gives A$900. The A$50 difference isolates that one multiplier, not a complete participant settlement calculation. Review Topic 3 and Topic 17 before applying real factors.

For an asset investment, link interval output, capture price, applicable loss treatment and contract allocation. A PPA may allocate curtailed energy, negative-price exposure and deemed generation differently. Before using the model's cost saving as a project benefit, specify whose cash flow changes, who bears construction and outage risk, and which costs remain outside the boundary. Connect the analysis to Topics 22, 23 and 39.

Worked numbers

10 MWh × A$100 × 0.95 = A$950; at 0.90 it is A$900.

Try it in the workspace

Keep MLF revenue sensitivity separate from the lossless constraint desk.

The question

What evidence supports the recommendation?

Mechanism

For a real event, assemble an evidence register: interval and timezone, constraint ID and version, invocation, coefficients, RHS, targets, violation and marginal-value fields, relevant offers, availability and network conditions. Preserve the source file and publication/revision time. Distinguish what was known at decision time from later corrections. An observed price spike and a binding flag are a hypothesis, not sufficient causal proof.

Independent task: use demand 130 MW and export capability 40 MW with the original plant capacities and offers. Solve the dispatch, compare a 10 MW expansion, and calculate the one-hour saving. Then increase demand to 150 MW with the original 40 MW limit. Explain why an apparently attractive price-based answer may still fail feasibility. Show every unit and constraint check before consulting the Worked example section.

Submit a one-page engineering-to-commercial brief: decision, base case, changed assumption, calculated effect, alternative explanation and missing evidence. Include a reproducible spreadsheet and distinguish model output from a market observation. A reviewer should be able to change one input and reproduce your result. Correct arithmetic plus a stated limitation is stronger evidence than an unsupported claim that a network upgrade will pay for itself.

Worked numbers

At demand 130 and transfer 40: cost A$8,000; at transfer 50: A$7,400. Demand 150 with transfer 40 is infeasible.

Try it in the workspace

Submit the dedicated Apply it brief before reading the benchmark.

The question

Does a Victoria–Tasmania spread prove Basslink congestion?

Mechanism

The included AEMO price extract records VIC1 at −A$25.13/MWh and TAS1 at −A$26.78903/MWh at the same published interval. Subtraction establishes a spread. It does not reveal which constraint was invoked, which way power flowed, or which unit and constraint terms contributed to price.

Preserve region, interval, run flag, publication status and file provenance. Then request matching interconnector flow and limit results, constraint invocation/definitions, coefficients, targets and price-setter evidence. Some information is published later; distinguish observation time from decision time.

The two-generator constraint desk is a causal experiment only inside its stated model. A historical price comparison is observational evidence. Write these on separate lines in your briefing so a reviewer can see which claims depend on assumptions.

Worked numbers

VIC1 − TAS1 = −25.13 − (−26.78903) = A$1.65903/MWh, displayed as A$1.66/MWh. A hypothetical 10 MWh price difference is A$16.5903 before losses and contracts.

Try it in the workspace

In Apply it load the five-region snapshot. Record the spread, then name three missing data families before writing a congestion conclusion.

The question

What changed in the constraint evidence register?

Mechanism

AEMO published Constraint Formulation Guidelines v14 on 20 August 2026, effective 1 October 2026. The update includes transmission-loop formulation associated with Project EnergyConnect. Follow the current guidelines and the specific constraint description rather than assuming the network is a single radial corridor.

The document includes IPRR-related text whose operation is conditional on commencement of those provisions. An effective date on a combined document does not make every included future rule operational. Treat that distinction as an explicit evidence field.

FY2026–27 MLF files are listed in the official loss-factor hub with July 2026 publication dates. Use the connection point and applicable period; a previous year's value is not a substitute. The separate MLF Framework Review describes proposed changes, not an automatic implemented smoothing rule.

Worked numbers

In the teaching desk, scale A ≤ 60 to 2A ≤ 120. Dispatch remains A=60, B=40, but a one-unit RHS relaxation now saves A$30 rather than A$60 over one hour.

Try it in the workspace

Compare Base transfer and Scaled equation. Export both input sets and explain why the economic interpretation must include coefficient scaling.