The Propane Tank That Starved the Generator in a Cold Snap

The call

A 22 kW air-cooled standby on propane, fed from a 120-gallon tank the previous owner of the house had installed for a gas range and a fireplace. The unit had passed every weekly exercise for two years. During a winter ice storm with the temperature near 5 F, the power went out at 6 p.m. The generator started, transferred and ran the house. Around 9 p.m. the lights began to dim and surge, and a little after that the generator shut down. The controller logged the shutdown as underspeed. It restarted on its own, ran about twenty minutes and shut down again. By the time the customer called the emergency line the house had been without reliable power for two hours and the indoor temperature was falling.

This was a stop-and-contain call before it was a diagnosis. Two people in the house, one of them elderly, an ice storm still going, and a fuel system nobody had looked at yet.

First minutes on site: what came before any test

Gas odor first. The tech walked to the tank and the generator from upwind and smelled the air before touching anything. A propane system that has been cycling an engine on and off in the cold can have a leaking connection, and frost on a fitting can be a leak rather than a draw. No odor. Had there been one, the instruction is fixed: everyone leaves the area, no switches, no engine starts, no phone used near the tank, and the call to the propane supplier and 911 is made from a distance.

CO inside. The customer had tried to warm the kitchen by running the gas range burners and the oven with the door open. The tech checked the house CO alarms (both working, both quiet), read 6 ppm in the kitchen on a handheld meter against 1 ppm in the hallway, had the oven turned off, and explained that an oven is not a space heater because it is not built to vent its combustion products to the room for hours.

Stop the cycling. The tech set the controller to off. A generator that starts, transfers, starves, shuts down and restarts every half hour puts the transfer switch through a cycle every time, hammers the house loads with brownouts, and can damage motors and electronics. A dark house with the unit off was better than a house on a failing source, for the few minutes it took to look.

What the tank showed

The tank was covered in frost from the bottom up to a sharp line about two-fifths of the way up its side. Above the line the steel was dry. The gauge read 45 percent.

That frost line is the signature of heavy vapor draw. Propane leaves a tank as vapor boiled off the liquid surface, and boiling takes heat. The heat comes from the liquid, which comes from the tank walls, which come from the outside air. When the draw is more than the air can supply through the wetted wall, the liquid cools below the air, moisture freezes onto the steel along the wetted area, and the tank's vapor pressure falls with the liquid temperature. A frost line at the liquid level means the tank is being asked for more vapor than it can make.

The supplier's driver, who arrived about forty minutes later, read the tank pressure gauge at 14 psig. A resting tank of propane near 5 F holds roughly 28 psig. The tank was running about 14 psi below its own resting vapor pressure for that temperature because the liquid had cooled well below the outdoor air.

Containing it

Containment came down to three moves, in this order, and one firm no:

• Cut the demand so the tank could keep up. The tech turned off everything the house could do without for one night: the electric water heater, the dryer, the well pump except for short periods, all but two rooms of lighting. With the unit off for about an hour, the tank had partly recovered; the driver's gauge read 20 psig by the time he left.
• Supplier temporary supply. The driver set two 100-pound cylinders on a temporary manifold to supplement the tank, which is the supplier's equipment and the supplier's work. More wetted surface means more vapor.
• Never heat the tank. No torch, no heat gun, no space heater pointed at the steel. The customer's adult son had suggested pouring hot water on it; the tech told him no. A tank with its relief valve sized for a fire is not a pot to warm.
• Generator back on, lighter. With the load trimmed and the cylinders manifolded, the controller went back to auto. The unit restarted, transferred, and ran the rest of the night on about half its earlier load.

The tech left a CO meter reading at the hallway, the oven off, the house warming, and a written note of the loads to keep off until morning.

The trace

The question was whether the tank was the whole story or one of several. Three candidates had to be separated, and the tank pressure gauge plus the unit's test port separate them cleanly.

A frozen or failing regulator. Regulators can ice up, from moisture in the gas or from freezing rain on the vent. A frozen second-stage regulator starves the unit too, and looks the same at the test port. The difference is upstream: with a frozen regulator, tank pressure stays at its resting level while the outlet pressure collapses. Here tank pressure itself had fallen about 14 psi below resting. The regulator was being fed too little, not failing to pass what it was fed. The regulator vents were clear and dry.

The line to the unit. An undersized line drops pressure under load whatever the tank does. Two days later, with the air at 28 F, the tech landed a load bank with the unit locked out under The Generator Auto-Start Lockout Before Service SOP and read the test port static and at full load as in How to Measure Gas Pressure at the Generator, Static and Under Load. Both readings sat inside the maker's spec at the start of the run; the line itself was fine.

The engine. The underspeed log entries fit a fuel starvation: power falls with fuel, speed falls with power. On the same load bank run, with good pressure, the unit held its rated output steadily. The engine was not the cause.

And then the tank failed the same test, warmer. On that load bank run at 28 F, the test port pressure held for about fifty minutes and then began to sag, and a faint frost line formed on the tank at the liquid level. At 28 F the tank lasted longer; at 5 F it had lasted three hours with the house as the load. The tank simply could not vaporize what the unit burns.

The arithmetic the original install skipped

The unit's spec sheet listed, say, 3.4 gallons per hour at full load and 2.2 at half load on propane. At about 91,500 BTU per gallon, half load is 2.2 x 91,500 = 201,300 BTU per hour, about 201,000. The house had been running near half load during the outage.

On gallons alone the tank looked fine. 45 percent of 120 gallons is 54 gallons; 54 / 2.2 = 24.5 hours at half load. A day of runtime, which is the number the previous owner and anyone glancing at the gauge would have trusted.

Vaporization is a rate, not a quantity. The supplier's chart for that tank, read at 5 F and 45 percent fill, gave a continuous vaporization rate of, say, 70,000 BTU per hour, roughly a third of the 201,000 BTU per hour the unit was burning. The tank delivered more than its steady rate for the first hours by drawing down the heat stored in the liquid, which is why the unit ran from 6 p.m. to about 9 p.m. before the pressure fell far enough to starve it. Every weekly exercise, a few minutes at light load, drew so little that the tank never came close to its limit. The exercise could not have revealed this.

The numbers that own this card's fuel figures are in Natural Gas and Propane Heat Content and Generator Consumption, and the tank check procedure is How to Check a Propane Tank for Generator Runtime and Vaporization.

The fix

The supplier set a larger tank, sized from its own vaporization chart so that at the area's design low temperature and at a realistic low fill, the point where the customer's delivery schedule triggers a refill rather than a full tank, its continuous rate covered the unit's full-load demand with room left for the range and the fireplace. Setting the tank, its location and its first-stage regulator are the supplier's work under NFPA 58 as adopted by the jurisdiction; the shop supplied the unit's demand figures and stayed out of the tank work. The 120-gallon tank was removed. The supplier also moved the customer to automatic delivery with a refill trigger well above the low-fill figure used in the sizing.

Confirmation. The shop ran a two-hour load bank test at full load on the next cold morning, 12 F, reading test port pressure every fifteen minutes. It held inside spec for the full two hours with no frost line on the new tank. Remote monitoring showed the next outage, six weeks later, carried for nineteen hours without a fault.

What the exercise could never have shown

A propane standby can pass a hundred weekly exercises on a tank that cannot carry it through one cold night, because an exercise draws a trickle for a few minutes and an outage draws a river for hours at the worst temperature of the year. The check that matters is vaporization at the coldest temperature the site sees and at the fill the tank will realistically be sitting at when the outage comes, not the gallons in the tank on install day. Any propane site with a tank the generator did not come with, or a tank shared with other appliances, earns that check at the site walk and again whenever the tank changes.

Two conditions change the answer. A site that cannot fit a larger tank can sometimes manifold two tanks for more wetted surface, which the supplier designs. And a large unit whose demand outruns any practical vapor tank moves to liquid withdrawal with a vaporizer, a different system with its own rules, covered in Propane Vapor vs Liquid Draw. Neither is a field improvisation.

References

• See related: How to Check a Propane Tank for Generator Runtime and Vaporization
• See related: Natural Gas and Propane Heat Content and Generator Consumption
• See related: How to Measure Gas Pressure at the Generator, Static and Under Load
• NFPA 58, Liquefied Petroleum Gas Code, as adopted by your jurisdiction (tank installation by the supplier)