Recycling economics under volatile energy prices

Electricity cost is not a single number per kilowatt hour. It is a function of how much a plant consumes, when it consumes it, and what the contract says about both.

The load profile describes the plant's consumption over time. A plant running a continuous process has a flat, predictable load. One running batch processes, or operating several lines that start and stop independently, has a variable load with peaks. The profile determines both the total energy consumed and the plant's maximum demand.

The process schedule is the operational decision about when that load occurs. Shredding, grinding, washing, extrusion and drying all consume energy at different rates, and some can be scheduled more flexibly than others. Whether the schedule can shift is a property of the equipment, of the storage available between steps, of the labour arrangement and of the throughput the business needs.

Where the schedule can shift, a plant gains a lever it can pull when prices move. Where it cannot — because the process is continuous, or because storage between steps is limited — the plant's exposure is fixed by its design, and the analysis should recognise that rather than assuming flexibility it does not have.

The contracted tariff determines how a change in the market price reaches the plant's cost, and the components respond differently.

A consumption charge per kilowatt hour is the component most people have in mind. It scales directly with the energy used, and it is the part most exposed to market movement where a contract is indexed.

A standing charge does not move with consumption at all. A plant that reduces its consumption does not reduce its standing charge, which matters for any measure justified on a percentage saving.

A demand charge, where one applies, is based on the plant's maximum demand rather than its total consumption. It is the component most often overlooked, and it creates a specific exposure: a plant whose peak demand is driven by a few short periods pays for that peak across the billing period. Reducing total energy without reducing the peak leaves a significant part of the cost untouched.

The contract's structure — fixed, indexed, hedged, or a blend — determines which of these the plant actually carries. A plant on a fixed-price contract has no exposure to market movement during the contract and considerable exposure when it renews; one on a fully indexed contract carries the movement continuously. The two have very different risk profiles and the analysis should reflect which applies.

When energy prices spike, the intuitive response is to stop. The response is frequently wrong, and the reason is restart cost.

Stopping a process carries costs: the materials in process may be lost or degraded, the equipment may take time to return to the required conditions, the labour may still be paid, and the output that would have been produced is not. Starting again may also involve a period of substandard output before conditions stabilise. For a continuous process with high fixed costs, the cost of stopping can exceed the cost of running through a high-price period.

This is the fixed-versus-variable distinction in operational form. A process with a high proportion of fixed cost — capital-intensive equipment, permanent labour, unavoidable standing charges — has a strong incentive to keep running, because the fixed cost is incurred whether or not it operates. A process with mostly variable cost has more freedom to stop.

The analysis should therefore test the decision explicitly: what does it cost to run through the period, and what does it cost to stop and restart? The comparison is specific to the process, but the framework applies generally, and it frequently produces the less intuitive answer.

The following is entirely hypothetical and demonstrates the arithmetic of testing price and yield together. It asserts no actual price, yield or cost.

Suppose a hypothetical plant has an assumed annual energy consumption of 2,000,000 kWh at an assumed price of 0.15 per kWh, giving an assumed energy cost of 300,000. Suppose its assumed revenue is 1,200,000 at an assumed saleable output, so energy is 25% of revenue.

Now suppose the price rises by an assumed 40%. Energy cost becomes 2,000,000 × 0.21 = 420,000, an increase of 120,000. If output and price are unchanged, the margin falls by 120,000.

Suppose also that the higher price leads the plant to run fewer hours, and the reduced throughput causes an assumed 5% reduction in saleable output. At the same assumed output price, revenue falls by 5% of 1,200,000 = 60,000. The combined effect is a margin reduction of 120,000 + 60,000 = 180,000.

The illustration makes the structural point: testing price alone understates the exposure, because in a real plant the response to price affects output. The two must be tested together, and the assumed output reduction is a stated assumption rather than a result.

A further refinement is the offtake price lag. Where the price the plant receives for its output responds to energy costs only after a delay, or does not respond at all, the margin absorbs the whole of an input cost increase for the intervening period. Modelling that lag — and its duration — is part of the sensitivity, and it can be the difference between a manageable period and an unmanageable one.