Five 90 kW compressors out, four two-stage units in: what a smaller installed base buys you
A plant that has grown by adding compressors usually ends up with a fleet nobody designed. Each expansion added a machine of whatever size was available, the lead/lag order was set years ago, and the installed kW figure — the one on the energy report — has stopped bearing much relation to what the plant actually needs.
One industrial robot manufacturer in Shanghai reached that point and replaced five 90 kW single-stage compressors with four two-stage machines: 2 × 90 kW and 2 × 75 kW, all permanent-magnet variable speed. Installed power fell by 120 kW without reducing the capacity the plant needed.
That is worth unpacking, because “we added VSD and saved energy” is the wrong lesson.
The fleet was oversized for its own load
The trigger in this case was not a failed machine. It was reliability and energy pressure: the plant needed better availability and lower specific consumption from the same air demand, and the existing fleet could not deliver either.
A five-machine fleet of identical single-stage units is a classic over-capacity signature. When demand sits between 55% and 75% of one machine’s output, the plant either runs a machine lightly loaded — where single-stage screws are at their least efficient — or cycles machines, which loads the system with start/stop losses and pressure swings. Neither shows up as a fault. Both show up on the electricity bill.
Two changes, not one
The project changed two variables at once, and they are worth separating:
Compression in two stages. At a given discharge pressure, splitting the compression work across two rotors reduces the losses in each stage and lowers the temperature rise per stage. The gain is real but conditional: it is largest at high pressure ratios and at sustained load, and smallest on machines that spend their life at 40% load.
Variable speed across the whole fleet. Permanent-magnet variable-speed drive lets each machine match its output to demand instead of loading and unloading. In a multi-machine station the effect is larger than on a single unit, because the controller can keep the base machine near its best efficiency point and let the variable-speed machine absorb the swing.
The two together are what allowed a smaller installed base to carry the same plant. Two-stage without variable speed would have improved the loaded hours; variable speed without two-stage would have improved the swinging hours.
What installed power does and does not tell you
Dropping 120 kW of installed capacity is a useful headline, but installed power is a design number, not a consumption number. A plant with 450 kW installed and a 25% average load can consume more than a plant with 330 kW installed and a 45% average load.
The question to ask before a fleet replacement is not “how many kW can we remove” but:
- What is the measured load duration curve? Not the peak, not the average — the hours spent in each load band.
- Which machines cover which band? The base should be carried by the most efficient machine at that output; the swing by the machine with the best part-load behaviour.
- What is the standby requirement? A four-machine station has to survive the loss of one unit. Right-sizing must leave that margin intact.
Why the replacement also bought reliability
Fewer machines, each running closer to its design point, is generally a better maintenance position than more machines idling. Running hours per unit fall, load/unload cycles fall, and the number of units to service falls with them.
The reported project also points at a softer benefit that applies widely: the fleet was assessed by an engineering team before the equipment list was fixed, using the plant’s actual duty rather than its nameplate capacity. The replacement was specified against measured demand — which is why the installed power could drop at all.
What to take from it
- Audit the fleet as a system, not machine by machine. A perfectly good compressor in a badly sequenced station is a cost, not an asset.
- Measure the load before choosing between two-stage, VSD and sequencing. Each fixes a different inefficiency, and buying all three where only one is needed is how retrofits fail to pay back.
- Expect the count and the sizes to change. A right-sized fleet rarely has the same number of machines as the fleet it replaces.
- Keep the standby margin explicit through the whole redesign.
The project’s saving was stated as a projection rather than a metered result, so no annual figure is repeated here. The structural change — a smaller installed base carrying the same demand — is the part other plants can copy; the kWh number belongs to the site that measures it.
Planning a compressed-air upgrade? Talk to SIND’s engineers.
Facts as reported by industry press; the project described is not a SIND project.
Planning a compressed air project?
Talk to an engineerPublished November 16, 2026.