The heat is small; the duty is not
A modern string inverter is 98.5-99 % efficient at its best point. That reads like a thermal non-event until you convert the missing percent into watts.
At 250 kW output, one to one and a half percent of loss is 2.5-3.8 kW of continuous heat. It is not a spike: on a clear day the machine sits near rated output from mid-morning to late afternoon, session after session, for decades. Central inverters in the megawatt class scale the same arithmetic upward, which is why the largest units move to liquid loops or very large forced-air paths rather than a single fan tray.
Three facts set every fan decision that follows. The heat is modest but continuous. The air available to carry it away has already been heated by the sun, so the sink is at 45-50 C, not 25 C. And the enclosure that has to reject the heat is rated IP65, which means the short path - air straight through the power chamber and out again - is deliberately closed.
The rest of this article turns those three facts into selection criteria: a loss budget, a derating target, an architecture choice, an airflow calculation, a filter and altitude margin, a night-time condensation strategy, and a replacement plan for a component that will not outlive the asset.
The derating curve is the real specification
The rated power on the front of the datasheet is conditional. The condition is an ambient temperature, and above it the machine quietly stops delivering what its nameplate promises.
IEC 62109-1 treats this as a design feature that has to be tested, not an emergency behaviour that happens in the field. Its thermal clause sets total temperature limits rather than rise limits; equipment rated above 50 C ambient has to be tested at the maximum rated ambient within 5 C of the declared value; and equipment with automatic derating has to be tested under as many conditions as needed to catch the worst case, including at least the maximum ambient before derating and the maximum ambient with derating. The reference conditions for those tests are themselves modest - 15-40 C, no more than 75 % relative humidity, 75-106 kPa - which is exactly why the rating above 40 C has to be proven separately.
What manufacturers publish as a result is a curve, and across the industry the shape is similar even though the knee moves by model. Full output holds to 40 C, and often to 45 C on outdoor models rated -25 to +60 C. Above the knee the output falls roughly linearly, commonly on the order of one to two percent of rated power per kelvin, until the machine reaches its shutdown limit somewhere in the 50-60 C band.
The commercial consequence is sharper than the thermal one. The derating window lands on solar noon - the hours with the most irradiance and, on many markets, the most valuable kilowatt-hours. A site that spends three months a year above 45 C can lose several percent of annual yield to a curve that was fixed by the thermal design. Moving the knee up by a few kelvin is worth more than any other change a fan can buy.
So the fan is not specified to prevent a shutdown. It is specified to move the derating knee, and the way to hold the specification honest is to state the target as an ambient temperature at which full output must still be delivered.
| Ambient temperature | Common behaviour | What it costs the plant |
|---|---|---|
| Up to 40 C | Full rated output on most outdoor models | Nothing - this is the design point |
| 40-45 C | Full output on many models; some begin a linear derate | The first kilowatt-hours of the hot season |
| 45-50 C | Linear derate, commonly 1-2 % of output per 1 K | A few percent of annual yield at hot sites |
| 50-60 C | Deeper derate; some models reach their shutdown limit | Output lost exactly at solar noon |
| Above 60 C | Beyond the rating of most outdoor models | Not a cooling problem - a selection problem |
IP65 closes the short path
An outdoor inverter is sealed against dust and water. Sealing is the right call for a 25-year asset, and it moves the hard part of the heat path into the wall.
IP65 under IEC 60529 means dust-tight and protected against water jets. It is the de facto rating for outdoor PV inverters, and for good reason: a ventilated cabinet in a desert site ingests silica, and in a coastal or agrivoltaic site it ingests salt. Both are conductive, both destroy creepage distances, and neither can be cleaned off a populated board.
Closing the cabinet means the heat has to cross the wall. Inside, a stirrer fan moves air across a finned baseplate so the power stage sees a moving, mixed air volume instead of a stagnant one. Outside, either natural convection or a fan bank moves air across an external fin stack. The first fan lowers the module-to-wall resistance; the second lowers the fin-to-sky resistance. Neither touches the junction-to-case resistance, which belongs to the power module and its mounting.
The temperature ladder makes the split visible. With 45 C ambient, a typical sealed design lands near 65 C at the fin surface, around 70 C in the internal air, roughly 90 C at the module case, and holds the junction under about 125 C. Every step is a thermal resistance in series, and the fan only controls the last two. A cooling project that spends its budget on the fan and ignores the fin area, the mounting interface or the internal stirrer will move the numbers far less than it expects.
Two architectures, two fan duties
Sealed and ventilated cabinets ask the fan for different things, and they fail in different ways. Pick the architecture before comparing datasheets.
A sealed machine treats the fan as a heat-transfer enhancer: it improves the convection coefficient on surfaces that would otherwise rely on natural circulation. It is tolerant of fan failure - a large finned aluminium body still rejects heat by convection, so a stopped external fan usually produces a slow temperature climb and a derate, not an immediate trip.
A ventilated machine treats the fan as the heat path itself. Air enters through a filtered louver, passes over the power stage, and leaves. It is efficient per watt of fan power and cheap to build, but the fan is now a single point of thermal failure, and the filter in front of it is a maintenance item with a pressure cost.
The choice is environmental, not ideological. Hot, dusty, salty or high-pollution sites favour the sealed machine despite its larger and heavier enclosure. Temperate, indoor or prefabricated-skid installations with a maintenance schedule favour the ventilated one.
| Aspect | Sealed (IP65) | Ventilated (IP54 / louvered) |
|---|---|---|
| Heat path | Internal stir, wall, external fins, ambient | Intake, across the power stage, exhaust |
| What the fan controls | Fin-to-sky and module-to-wall resistance | Airflow, and therefore the whole temperature rise |
| Contamination | External surfaces only - washable | Internal - creepage, corrosion, blocked filters |
| Filtration | Usually none required | Required, with a cleaning or change interval |
| Acoustics | External fan noise radiates directly | The enclosure provides some attenuation |
| Failure mode | Slow climb into derate; convection residual remains | Fast climb; minutes to tens of minutes to trip |
| Best fit | Desert, coastal, high-pollution, unattended | Temperate, indoor or skid, maintainable |
From watts to a working point
One equation turns the loss budget into airflow, and one correction turns airflow into a fan specification.
For air, the required volume flow is q = P / (rho x cp x dT), where P is the heat to reject, rho is air density, cp is about 1005 J/(kg K), and dT is the allowed air temperature rise across the cabinet. Take the 3 kW of a 250 kW string inverter, dry air at 45 C with a density near 1.11 kg/m3, and a 15 K rise: q is about 0.18 m3/s, roughly 650 m3/h. Allow 20 K instead and the requirement falls to about 485 m3/h.
The trap is what the two numbers imply. A 15 K rise on a 45 C day puts the exhaust near 60 C and the internal air near 65-70 C, and every component in the chamber ages at that temperature. A 20 K rise is cheaper in airflow and more expensive in capacitor life. The right value is the one the derating target and the lifetime model agree on, not the one that makes the fan small.
The correction is the system curve. The fan has to deliver that volume against the installed path: louver or grille, filter if ventilated, fin stack, turns, and the system effect of a fan mounted close to an obstruction. A fan selected on its free-air figure routinely loses a third of its delivery once installed. Specify the duty as a volume flow at a stated static pressure, measured or modelled on the installed geometry, at the worst-case ambient air density.
For sealed designs the same equation applies to the external air path, but the useful dT is across the fin stack surface rather than through the chamber, and it is usually smaller. There, fin area and fin efficiency set the ceiling; a bigger fan buys a lower convection resistance and diminishing returns after that.
Dust, filters and the pressure budget
Where air is admitted, dirt follows. Filter pressure drop is not a maintenance footnote - it is a term in the airflow equation that changes every week.
A ventilated cabinet starts with a clean filter pressure drop on the order of 25-40 Pa and is normally alarmed at something like 150-250 Pa. Because the fan rides its own curve, that rise in resistance is not absorbed: it moves the operating point left, and flow falls. A design with no margin ends up derating in week six of a dust season for reasons that look electrical.
Two rules follow. Size the duty at the dirty-filter condition, not the clean one, and alarm on differential pressure rather than on a calendar interval. In desert and agricultural sites a filter can load in days after a single event, and a fixed quarterly schedule guarantees either unnecessary service or a late one.
Sealed machines trade filter maintenance for surface maintenance. Dust on the external fin stack insulates it; dust on fan blades unbalances them and raises bearing load and noise. Both are washable, and both belong in the site's preventive work list - which is why external fins should be reachable without opening the sealed chamber.
Corrosion deserves the same treatment as dust. Coastal and agrivoltaic sites need fan materials, coatings and bearing sealing selected for salt-laden air, and the sealed architecture is usually the cheaper answer because it keeps that air out of the electronics entirely.
Altitude: less air, and wider clearances
Plateau sites are attractive for irradiance and punishing for cooling. Two separate deratings apply, and they are not the same one.
The thermal one is air density. Standard atmosphere gives about 1.225 kg/m3 at sea level, 1.007 at 2000 m, 0.909 at 3000 m, 0.819 at 4000 m and 0.736 at 5000 m. A fan is a constant-volume machine: it moves roughly the same cubic metres per minute, but each cubic metre carries proportionally less mass, so the heat it removes falls with the density ratio - to about two thirds of sea-level capacity at 4000 m. Available static pressure falls with density too, which makes the system curve steeper at exactly the moment the duty is hardest.
The practical correction is to size the volume flow by the inverse of the density ratio. A cabinet that needs 650 m3/h at sea level needs on the order of 950-1000 m3/h at 4000 m to reject the same kilowatts at the same temperature rise, and the fan has to produce it against the thinner air's lower pressure capability.
The electrical one is clearance. IEC 60664-1 gives altitude correction factors for clearances: 1.00 at 2000 m, 1.14 at 3000 m, 1.29 at 4000 m and 1.48 at 5000 m. Equipment intended for 4000 m needs its clearances multiplied by 1.29, which is why inverter datasheets commonly state a full-power altitude limit - historically 4000 m, and now increasingly 5000 m as plateau plants are built higher.
Both deratings land on the same machine at the same site, and only the first one is the fan's problem. Confirm which altitude the nameplate rating refers to before promising a cooling fix for a unit that is electrically derated.
Night: the quietest duty is the one that protects the asset
At sunset the power goes to zero and the thermal problem does not. The hours after shutdown are when condensation is made.
PV inverters are routinely rated for 0-100 % relative humidity including condensation, because that is the reality of the duty: a warm, humid chamber cooling through the night will cross the dew point on its coldest metal. What follows is not dramatic - no trip, no alarm - but it is cumulative: tracking on insulation surfaces, corrosion of terminals and fins, and, on populated boards, electrochemical migration.
Fan strategy is part of the answer. A low night-time speed, or intermittent runs, keeps internal air mixed and the coldest surfaces slightly warmer than still air would leave them; combined with a heater it is the standard anti-condensation approach. Sealed designs add a second option: a breathable membrane vent that equalises pressure across the enclosure - important when a sealed chamber cycles through 40 K a day - while keeping liquid water and dust out.
Two cautions. Low-speed operation must still be monitored: a fan creeping at a few percent of rated speed can fail to start or stall without the tachometer output ever looking wrong, so specify a feedback signal and test the low-duty corner. And any night-time run has to be within the site's acoustic limit, which for plants near housing is often the binding constraint rather than the thermal one.
Fan life against a 25-year asset
The plant is financed for two decades or more. The fan is not. Design for that difference instead of discovering it in year seven.
Twenty to twenty-five years is 175,000 to 219,000 hours. A typical DC fan carries an L10 bearing life quoted in the 40,000-70,000 hour range at 40 C, and bearing life falls steeply with temperature; the widely used rule of thumb is that every 10 K of additional bearing temperature halves it, though the actual curve comes from the fan's own datasheet and test method, not from the rule. Either way, the fan is a scheduled replacement item in a machine that is not.
That makes serviceability a design requirement, not a nice-to-have. A fan tray that can be swapped from the front, without opening the sealed power chamber, without draining anything and without special tools, turns a truck roll into a short stop. Connectors, cable length, and the mechanical interface should be specified with that swap in mind, and the spare should be defined as part of the BOM.
Instrumentation closes the loop. Tachometer or rotation-detection feedback on every fan, reported to the plant SCADA, converts a hidden failure into an alarm; trending current, speed and acoustic changes turns it into a prediction. On multi-fan banks, N+1 means one failure is a work order rather than an outage - provided the surviving fans are actually sized to hold the derating target in the worst case, which is a duty statement, not a redundancy claim.
What MEGA Tech can do here - and what we need from you
We build cooling fans for exactly this duty: specified around the site conditions, and proven on the customer's own cabinet before the BOM is frozen.
What we do is selection and customization around a real duty cycle. Frame size, thickness, voltage, speed curve, PWM control, FG/RD feedback, connector, cable length, bearing and lubrication, and protective measures for dust or salt-laden air are all adjustable. Where an existing unit is being replaced - including machines built by Sanyo Denki (SAN ACE), SUNON, Delta, NMB or ebm-papst - we work from the installed duty and measurements rather than from the frame size alone, because equal outlines do not guarantee equal delivery on a given system curve.
What we need from you is the duty, not just the part number: rated power and the efficiency or loss curve, the derating curve or the ambient temperature at which full output must hold, site ambient range and solar exposure, altitude, pollution degree and cleaning interval, the enclosure geometry (fins, ducting, filter, louvers), the acoustic limit, the life target and replacement strategy, and any measured system curve or thermal data from the existing build.
Where we stop is also part of the answer. We do not supply the coil or liquid loop, the control electronics, or the cabinet safety design: IP rating, clearances, creepage and insulation coordination under IEC 62109 and IEC 60664 belong to the equipment manufacturer. Our scope is the air side - the last two resistances in the chain - and the evidence that it works in your enclosure.
- Fan and blower selection against the installed system curve at worst-case ambient and dirty-filter conditions.
- Customization: frame, thickness, voltage, speed curve, PWM, FG/RD, connector, cable, bearing and lubrication, coating for dust or salt.
- Replacement and second-source programs validated by measurement on the real cabinet, not by outline comparison.
- Sample build, instrumented test and a documented working point before production release.
Recommended Fan Starting Points
These catalogue series may be suitable starting points for this application. Final selection depends on the installed airflow, static pressure, voltage, noise target and operating environment.
Request a sample for validation on your equipment.
Engineer FAQ: Solar Inverter Cabinet Cooling
Short answers to the questions that come up when a plant is already built and the numbers are not what the datasheet suggested.
Q: Is a higher-RPM fan the fix for a cabinet that derates every afternoon? A: Only if the installed path can use the extra flow. Check the system curve first: with a loaded filter, a tight louver or a badly placed fin stack, more speed mostly buys noise and a worse operating point.
Q: Should we ventilate a sealed IP65 inverter to cool it? A: Generally no. Ventilation buys thermal headroom and pays for it with contamination, filter maintenance and a faster failure mode. Improve the fin stack, the external airflow and the internal stirrer before opening the enclosure.
Q: How much airflow does a 250 kW inverter need? A: From q = P / (rho x cp x dT). At 3 kW of loss, 45 C air and a 15 K rise, about 650 m3/h of delivered flow - then correct for installed resistance and for altitude density before selecting the fan.
Q: Why is the cabinet still hot an hour after sunset? A: Thermal mass and a lagging air path. Cabinet air peaks after ambient, which is also why the derating band sits in the early afternoon rather than at noon.
Q: Can we run the fans slowly overnight to stop condensation? A: Yes, and it is common practice, usually with a heater. Specify a feedback signal and test the low-duty corner, because a stalled fan can look healthy to a supervisory input.
Q: Do we need N+1 fans on a sealed cabinet? A: It depends on the residual natural-convection capacity. Ventilated machines need redundancy because the fan is the path; sealed machines often ride through one failure into a derate, but only if the surviving fans are sized for it.
Q: What can MEGA Tech customize for this duty? A: Frame, thickness, voltage, speed curve, PWM, FG/RD, connector, cable, bearing and lubrication, and coatings for dust or salt - selected against your measured duty and validated on your cabinet.
Questions for the Inverter Cooling Design Review
A short checklist that keeps a derating complaint from turning into an endless fan swap.
- At what ambient temperature must full rated output still be delivered, and is that the number the fan is specified against?
- Where does the loss actually go - power stage, magnetics, capacitors, cabling - and is the budget agreed in watts?
- Is the architecture sealed or ventilated, and is that choice driven by site contamination or by cost?
- What is the delivered airflow at the installed system curve, at dirty-filter condition and worst-case ambient density?
- What is the allowed air temperature rise, and does the resulting internal temperature satisfy the capacitor and module life model?
- What altitude is the machine rated for, and have both the density derating and the clearance correction been applied?
- What happens on the night cycle: dew-point crossing, heater or fan strategy, and the pressure-equalisation path?
- How is fan health reported to SCADA, and what is the failure mode of a single fan in this architecture?
- Can the fan be replaced on site without opening the sealed chamber, and is the spare defined in the BOM?
- Has the candidate fan been measured on the actual cabinet across the real duty cycle, or only on a bench curve?
