Generator Load Bank Testing: Why a Standby Set Needs to Work Hard Once a Year

A standby generator has one job, and it does that job on the worst day of the year: the grid fails, the hospital theatre or the cold store or the telecom shelter goes dark, and the set has seconds to start and carry the load. Most standby sets spend the rest of the year either stationary or running a weekly exercise with little or nothing connected. That is how a machine that "starts every week without a problem" fails in the first hour of a real outage.

A load bank test is the answer to that gap. It puts the set under a controlled, measured load up to its rating, for long enough to prove the engine, the alternator, the cooling system and the fuel system can do what the nameplate says. This article covers why standby sets fail when called on, what a load bank is and why the type matters, what a typical test looks like, where the standards sit, and what the report should contain.

Why Standby Sets Fail When They Are Called On

Light-load running and wet stacking. A diesel engine run at low load does not reach the combustion temperatures it was designed for. NFPA 110 describes the result as wet stacking: unburned fuel or carbon, or both, in the exhaust system, readily indicated by continuous black smoke while the engine runs. Carbon builds up in the exhaust and turbocharger, fuel dilutes the oil, and cylinder bores glaze. Our diesel generator maintenance schedule covers the mechanism in more detail.

Untested cooling. A radiator partly blocked with dust, a slipping fan belt or a thermostat that sticks will pass any no-load run. They show only when the engine has to reject full-load heat for an hour.

Untested fuel supply. Transfer pumps, day tank controls, filters loaded with sediment and fuel that has aged in the tank all limit the set at high fuel flow, not at idle.

Untested alternator and controls. The automatic voltage regulator, excitation system and protection settings are never exercised near their limits by a set running light.

Untested switching. Breakers that have not operated in years and transfer switches that have never transferred real load are part of the standby system too. A load bank does not test them, which is covered below.

What a Load Bank Is

A load bank is a controllable electrical load that absorbs the generator's output and turns it into heat, or into reactive current, in selectable steps. It lets the tester choose exactly how much load the set sees, and for how long, without depending on the building.

Resistive load banks are banks of resistive elements cooled by fans. They draw real power, measured in kW, at unity power factor. They load the engine fully and are the most common type for routine testing.

Reactive load banks use inductors, and sometimes capacitors, to draw reactive power, measured in kVAr. On their own they load the alternator's current capacity without much engine load.

Resistive-reactive load banks combine both, so the test can be run at a chosen power factor, usually the one on the set's nameplate.

Why the Reactive Element Matters

Generator sets are commonly rated at 0.8 power factor lagging, and the arithmetic shows what a resistive-only test leaves out. A 1,500 kVA set at 0.8 power factor is a 1,200 kW set. A resistive load bank at 1,200 kW loads the engine to its full rating, but the alternator carries only 1,200 kVA of current, which is 80 percent of its rated current. The remaining alternator capacity is only reached when 900 kVAr of reactive load is added alongside the 1,200 kW, because 1,200 squared plus 900 squared equals 1,500 squared.

So a resistive test proves the engine and the cooling system. A resistive-reactive test at rated power factor also proves the alternator windings, the voltage regulator and the excitation system at rated current and temperature. For a large set, particularly one that will run motors, pumps and transformers whose loads are inductive, the second test is the one that resembles the real duty.

The three load bank types and what each one proves. Only a resistive-reactive test at rated power factor takes the alternator to its rated current. Source: MIMAH engineering practice; arithmetic shown in the article.

A Typical Test Profile

There is no single profile, and the one in your contract, code or manufacturer's procedure governs. A typical stepped test looks like this.

Before the load. Record the set's identity and ratings, check oil, coolant and fuel levels, inspect for leaks, confirm the load bank connection and cable ratings, and note the ambient and room temperature.

Warm up and step up. Start the set, allow it to reach operating temperature, then increase load in steps, for example a quarter of rated load at a time, up to rated load. Hold each step long enough for coolant, oil and exhaust temperatures to settle.

Hold at rated load. Run at rated load for the duration the specification sets, with readings taken at fixed intervals.

Step down and cool down. Reduce load in steps and allow a cool-down run before shutdown, as the manufacturer recommends.

At each step and interval the log should capture voltage and current on each phase, frequency, kW and kVAr or power factor, coolant temperature, oil pressure and temperature, exhaust temperature where instrumented, battery charging, and any alarms, smoke or leaks. A single set of readings at full load is a snapshot. The trend across the hold period is what shows a cooling system losing ground.

A typical stepped load bank test. The profile in your contract, code or manufacturer's procedure governs. Source: MIMAH engineering practice.

Block Loads, Transient Response and Black Start

A stepped test proves the set can carry its rating. It does not prove the set can pick up load suddenly, and that is what happens when a transfer switch closes during an outage.

A block load test applies load in large steps and records the voltage dip, the frequency dip and the time to recover. ISO 8528-5, part of the ISO 8528 series for reciprocating engine generating sets, sets the limits for these transient deviations and recovery times for each performance class. Many turbocharged diesels cannot accept their full rating in a single step, and manufacturers state the load steps their set can take within its class. The test should confirm those figures, and the site's largest motor start should sit within them. Load rejection, the set's response to losing a large block, belongs in the same test.

Black-start sets carry an extra requirement. A black-start generator must start with no external supply, on its own batteries or air start, close onto a dead bus and pick up its first block of load. MIMAH supplied 1,500 kVA black-start diesel generator sets in Nigeria, and for this duty the test worth running is the one that reproduces the dead-bus start, not a start from a healthy auxiliary supply.

What the Standards Ask For

The most detailed public benchmark is NFPA 110, the standard for emergency and standby power systems. It is a US standard and not binding in most of the markets we work in, but it is well reasoned and widely used as a reference.

In recent editions, NFPA 110 requires diesel sets in service to be exercised at least monthly for a minimum of 30 minutes, either at a loading that maintains the manufacturer's recommended minimum exhaust gas temperature, or at not less than 30 percent of the standby nameplate kW rating. It permits a supplemental load bank to make up the difference where the building load is too small, and its annex points out that 30 percent of a standby rating is a larger figure than 30 percent of the prime rating on the same set, so the basis must be recorded. For diesel sets whose monthly exercise does not meet that minimum, it adds an annual test with supplemental load in stepped percentages of nameplate rating over a continuous period. The exact steps have changed between editions, so work to the edition your authority enforces.

NFPA's own overview of emergency power supply systems summarises the longer test: a Level 1 system, the kind whose failure could cost lives, must be tested at least once every 36 months for the duration of its assigned class, though the test need not exceed four hours.

Load Bank or Building Load?

Running the set on the real building load through the transfer switch proves something a load bank cannot: that the switch transfers, the breakers close and the site stays up. NFPA 110 requires a monthly transfer switch test for exactly this reason.

But the building load is uncontrolled. It is usually well below the set's rating, especially where sets were sized with generous margins, its power factor is whatever the site happens to be drawing, and nobody can step it up to full rating for an hour without risk to the operation. On many sites it is also far too light to prevent wet stacking. Solar hybrid sites make this worse, since the generator runs only when the solar array and batteries cannot cover the load, often briefly and lightly. Our guides to hybrid solar systems and commercial battery storage cover how that duty cycle arises.

The answer is both: building load for the transfer, load bank for the capability. For sites where a failed start has consequences for patients, such as those covered in our guide to solar for hospitals and clinics, both should be on the calendar.

What the Report Should Contain

A load bank test without a report is an anecdote. A useful report contains the following.

Set identification. Make, model, serial numbers, running hours, and the rating used as the test basis, standby or prime.

Load bank and connection. Type, power factor achieved, and the connection point.

As-found condition. Fluid levels, leaks, battery condition and anything corrected before the test.

The readings table. Every step and interval with the electrical and engine parameters listed above.

Transient results. Voltage and frequency dips and recovery times for each block, where a block test was run.

Observations. Smoke, alarms, leaks, abnormal noise or vibration, and any step at which the test was held or stopped.

Verdict and recommendations. Pass or fail against stated criteria, and what needs doing before the next test.

What a load bank test report should contain. Without the readings table and a stated rating basis, the test cannot be compared with the next one. Source: MIMAH engineering practice.

Planning the Test

A load bank dumps the set's full output as heat, so the discharge must be pointed away from people, fuel and the set's own air intake. Cabling must be rated for the full current, the connection made at a proper point with the set isolated from the site load or arranged so a real outage takes priority, and fuel checked for the full test duration.

On frequency, annual is the common pattern for sets whose normal running never reaches a meaningful load, with a longer test on a multi-year cycle. For many commercial and industrial sites in Nigeria, where the generator is part of daily life rather than standby, the loading question is often the opposite one, and our comparison of solar versus diesel generation in Nigeria covers when that duty starts to change.

MIMAH's maintenance team maintains diesel and hybrid generation systems. If you have a standby set that has never been tested at its rating, or a test report you are not sure how to read, get in touch.