When a large industrial complex replaced several ageing cogeneration plants with one centralised steam-and-power utility, the steam header conditions had to be reassessed to keep modification costs down across every colocated company.
Industrial complexes that grow over decades end up with a patchwork of utility arrangements. Individual companies build their own cogeneration plants, sized for their own needs at the time, with steam and power systems never designed to talk to each other. That patchwork gets expensive. Smaller cogen units run at lower thermal efficiency than a single larger plant would, and each company carries the capital and maintenance burden of its own generation.
The case here was exactly that: a multi-company site including an alumina refining operation, where several lower-capacity cogeneration plants were retired in favour of one centralised utility supplying steam and power to everyone.
The problem with “just connect the pipes”
On paper, consolidating utilities is straightforward. Build one bigger, more efficient plant, pipe steam and power out to everyone who used to generate their own. Steam distribution networks are less forgiving than that.
Each company’s header had been designed around the pressure, flow and demand characteristics of its own dedicated plant. Merging those companies onto one centralised network changes all of it — pressure drops, flow distribution, header sizing margins, and transient behaviour under variable multi-company demand.
So the question was never “will steam reach each company.” It was whether headers installed long before anyone conceived of a merger could safely handle the new operating envelope, without every company rebuilding its distribution network from scratch. With several independent companies sharing the outcome, holding down the cost of physical modification became a design objective in its own right.
Four questions the study had to answer
- What configuration should the centralised utility have? Unit sizing, redundancy, the steam and power split, how many generation trains and at what capacity. There is no single right answer — configurations trade capital cost, efficiency, reliability and flexibility against each other, and the choice determines what conditions the downstream headers actually see.
- How does it handle each company’s varying demand? Colocated companies do not draw steam and power in sync. Profiles differ by process, by shift pattern, and over time. The network has to be assessed against the realistic range of combined demand, including what happens when one company spikes while others sit low.
- How does annual cost to serve compare across configurations? Not capital cost alone. Fuel, operating and maintenance cost, and the header modifications each configuration would require, compared on a whole-of-life basis rather than lowest upfront spend.
- Which header settings minimise retrofit cost? Which combination of pressure levels and pressure-reduction arrangements keeps the most existing headers inside safe operating limits, so that the fewest physical modifications are needed across all companies.
Toward a systematic framework
Rather than four separate exercises, the work became one framework connecting them. Configuration options were screened against demand variability and annual cost to serve; for each viable configuration, header settings were optimised to minimise retrofit cost. That gives a repeatable path from “what should we build” through to “what will it cost each company to connect,” with cost rather than bare technical feasibility driving the recommendation.
Why it matters beyond this site
Centralising steam and power is, in principle, a clear efficiency win. Fewer, larger, more efficient plants replacing several small ones. That win only materialises if the configuration is chosen with real demand variability and annual cost to serve in mind, and if the existing distribution network — usually the least visible part of the system — can support it without a costly rebuild.
The general lesson: utility centralisation needs configuration selection, demand variability, whole-of-life cost and the legacy network treated as one connected design problem. Four decisions made in sequence will not find the lowest-cost path.
Specific project details, including the site and companies involved, remain confidential. This post describes the problem and approach at a level intended to be useful to others facing similar consolidation work.