Energy trading internships

A $20 price spread can produce only $20 of total margin

Fictional delivery obligation: sell 10 MWh at $60/MWh and buy 10 MWh at $40/MWh. With complete delivery and no other costs, margin is 10 × ($60 − $40) = $200.

Now suppose the supplier delivers only 7 MWh and charges you only for that amount. You still owe the customer 10 MWh. Buying the missing 3 MWh at $100/MWh gives revenue of $600 and costs of 7 × $40 + 3 × $100 = $580. Margin is $20, not $200. The replacement purchase costs $180 more than buying those same 3 MWh at $40.

At a $120/MWh replacement price, total purchase cost becomes $640 and the result is a $40 loss. This example excludes transport, penalties and other contract terms. The lesson is to connect the price view to the delivery obligation; it is not a prediction about an actual market.

Compare forecasts without borrowing information from the future

A fictional decision is made at 09:00. The forecast available then predicts prices of $40, $50 and $60/MWh for three later intervals. Actual prices are $45, $48 and $72. Absolute errors are $5, $2 and $12, so mean absolute error is $19/3, about $6.33/MWh.

A baseline of $42 for every interval has errors $3, $6 and $30, averaging $13/MWh. The first forecast is better on this small sample. But if its third prediction was revised to $70 after 09:00, using that revision would reduce the reported error without improving the forecast available for the original decision.

Store the issue time as well as the time being predicted. Evaluate the version that existed when a decision was possible. Then keep forecast error separate from trading profit: a profit calculation additionally needs a position, execution price, settlement convention and costs.

• Forecast available at 09:00: 40, 50, 60. Correct mean absolute error: 6.33.
• Later revised forecast: 40, 50, 70. Mean error would be 3.00, but it is not the original 09:00 forecast.
• A chart of the later forecast cannot substantiate the earlier trading decision.

Make the application match the team's function

For analytics, show the dated forecasts and their errors. For scheduling, use the shortfall example to explain the unresolved quantity and time. For risk, explain how the replacement-price assumption changes the outcome. These are more useful application examples than a general claim that you follow energy markets.

Ask what the intern can prepare independently and what needs approval before affecting a commercial process. A strong sample can be entirely synthetic, with its assumptions visible; it does not need proprietary prices, positions or customer contracts.

Before you apply

Eligibility

Use the country-specific commercial or trading posting. A technical internship and a commercial internship at the same company can require different academic preparation.

Timing

Shell describes assessed internships as generally 10–12 weeks. bp separates opportunities by country. Neither overview confirms an available trading-desk seat.

Location

Confirm the desk or commercial office and expected market hours, not the company's headquarters address.

Pay

Ask for guaranteed pay, placement length and relocation terms. An unrelated quantitative-trading firm's package is not a useful estimate for this offer.

Sources: Shell assessed internships · bp early careers

Job titles to search

Trading or commercial intern

Ask which commodity and decision your analysis would support.

Scheduling or operations intern

Ask how quantities, delivery times and exceptions are recorded and resolved.

Risk or market analytics intern

Ask which exposure or outcome you would calculate and independently check.

Where to look

Search routes, not confirmed vacancies.

Compare the work and try an example

Use a role comparison or worked exercise to choose your preparation before tailoring an application.

• Energy trading, scheduling and risk internships: distinguish the work
• Midstream commercial versus engineering internships
• Utility internships beyond engineering: finance, contracts and project work