Substation Voltage Regulation, LDC & CVR

Acceptable Customer Voltage

In the United States, ANSI C84.1 establishes standard voltage ranges for electric power systems.

For a typical residential service with a nominal voltage of 120 volts, its normal service range, called Range A, extends from 114 volts minimum to 126 volts maximum. For simplicity, we'll ignore the fact that people actually have split-phase service voltages of double these values.

This gives us a total acceptable voltage range of 12 volts.

These limits apply at the customer's service point, not directly to the primary distribution feeder.

For now, we'll treat the feeder's 120-volt-base readings as if they were service voltages. We'll account for the transformer and secondary drops when we get to the voltage budget.

Bus-Only Regulation

So how do we keep all the customers on a substation within these bounds?

The easiest way is bus-only regulation. The substation controller monitors the bus voltage and commands the transformer load tap changer to raise or lower it as needed. Its job is to keep the bus within a regulation band around a fixed setpoint.

Simple regulation like this only requires three parameters:

• Bandcenter: the target voltage, such as 122.5 V.
• Bandwidth: the total width of the band. A 2 V bandwidth centered on 122.5 V spans 121.5 to 123.5 V. One tap step moves the bus about 0.75 V (5/8% of 120 V), so the bandwidth must be wider than one step, or the LTC will hunt back and forth across the band. A rise of 0.9 volts from the bandcenter is still inside the band; a rise of 1.9 volts is outside it and starts the controller's time delay toward a tap correction.
• Time delay: how long the voltage must stay outside the band before the controller acts, typically tens of seconds.

If load increases while the tap position stays fixed, the greater current generally produces more voltage drop through the transformer's internal impedance, tending to lower the low-side bus voltage. It is also possible for the highline supply voltage to change.

The controller doesn't need to know which of these caused the change: if the measured bus voltage moves outside its regulation band, it can command a tap change, after its time delay, to bring the voltage back toward the setpoint.

The controller only responds to the voltage at the bus.

In this demonstration, you play the controller by moving the tap yourself. The green band on the meter is the 114–126 V customer range, not the narrower regulation band we just described.

In practice, it usually monitors just one phase of that bus, even though the LTC changes taps on all three phases together. Loading is rarely perfectly balanced, so a lightly loaded monitored phase can read high and stay inside the band while the voltage on the other phases sags. Apart from the following demonstration, the examples on this page assume balanced three-phase loading.

Voltage Drop Along the Feeder

From the substation bus, feeder circuits head out into the surrounding neighborhood.

Along the way, pole mounted transformers supply houses, businesses, and other electrical loads. Some of these transformers are tapped from the main feeder, while others are supplied through laterals branching off in different directions.

Every load draws current, and that current has to travel through some portion of the distribution circuit to reach its destination.

This is important because every section of conductor has impedance.

We know from Ohm's law that the voltage difference across an impedance is related to the current flowing through it:

Since we're dealing with AC circuits, both current and impedance are complex quantities. The voltage difference depends on the resistance and reactance of the conductor, along with the magnitude and phase angle of the current. In this simplified discussion, we'll mostly talk about the resulting change in voltage magnitude.

For a single section of line, a good working approximation for that change in magnitude is:

Here, θ is the power-factor angle between the voltage and the current.

One note on units: with primary amps and ohms, this gives primary volts. To put the result on our 120-volt base, divide by the VT ratio, which is 60 on a 12.47 kV system.

At unity power factor, only the resistance matters. With lagging current, both terms add, and at ordinary power factors, a poorer power factor means more drop for the same amps. With leading current, such as from a large capacitor bank at light load, the reactance term subtracts. The drop shrinks, and the voltage can even rise along the line.

You can play with this demonstration to get a feel for how each term contributes to the overall voltage drop.

Hang on to this equation. We'll come back to it later.

For now, we can simplify things by saying that greater current flowing through a given section of conductor generally produces a greater voltage drop under ordinary lagging-power-factor loading.

Another important thing to keep in mind is that the current isn't the same everywhere along the feeder.

Consider a section of line relatively close to the substation. That section carries current supplying all the loads connected downstream of it.

Farther down the feeder, some of those loads have already branched off. Consequently, the remaining sections of conductor may carry considerably less current.

To determine the voltage at a particular point, we need to calculate the voltage difference across every section of conductor between that point and the substation, then subtract that 'drop' from the source voltage on the bus.

The following demonstration is a simplified version of this idea. An imaginary feeder has three tapped loads, which divide it into sections. You can change each load, or move where it is tapped, and visualize the corresponding voltage drop across the line. A real feeder works the same way, with many more loads and one section between each pair of them.

On a real feeder, there could be many laterals branching from the main line, each with its own load and length. As the following demo shows, this can complicate the range of voltages spread across the entire line.

Under ordinary loading conditions, voltage generally decreases as we move farther from the substation.

For instance, we might start with 122.5 volts at the substation bus and end up with 118 volts at the end of a relatively long feeder.

That reduction isn't necessarily a problem. In fact, distribution systems are designed to accommodate a certain amount of voltage drop.

The question is how much voltage drop we can tolerate.

The Limits of Substation Voltage Regulation

Suppose our substation bus is regulated to 122.5 volts.

At the end of a long lateral, we measure a customer service voltage of 113 volts.

In our simplified model, that's a total voltage drop of 9.5 volts between the substation bus and the customer's service point.

Since 113 volts is below the minimum acceptable service voltage, we need to do something about it.

The simplest solution would be to raise the bus voltage by about one volt.

Assuming the voltage drop stays about the same, that would bring our distant customer up to approximately 114 volts while raising the bus to 123.5 volts.

(If you are wondering whether raising the bus voltage would also change feeder current, know that it can, depending on the loads. To keep this discussion manageable, our demonstrations hold each load's current fixed, as if it didn't depend on voltage. We'll come back to that when we look at how loads respond to lower voltage.)

Problem solved.

But what happens if the customer is considerably farther away, or the circuit has more load?

Let's imagine that the same customer instead measures 109 volts.

Now we have a total voltage drop of 13.5 volts.

To bring this customer up to the minimum acceptable voltage of 114 volts, we'd need to increase the substation bus voltage by approximately five volts, again assuming the feeder drop stays the same.

That would put our bus at 127.5 volts.

Unfortunately, we might have customers connected relatively close to the substation who would now experience excessive voltage.

This illustrates a fundamental limitation of regulating an entire feeder from one location.

We can raise or lower the voltage of the entire circuit, but doing so doesn't eliminate the voltage drop between the substation and the most distant customers.

If we assume that our first customer sees essentially the same voltage as the substation bus, and that customer must stay at or below 126 volts, we have a theoretical voltage-drop budget of 12 volts before the last customer falls below 114 volts.

In a real distribution system, some of that voltage allowance must also accommodate drops through pole mounted transformers and secondary tri-plex conductors, as well as operating margin.

Therefore, the amount available for primary feeder voltage drop is generally less than the full 12 volts.

The regulation band takes its share, too. The bus isn't held exactly at its bandcenter; it can sit anywhere inside the band. A 2-volt bandwidth means 1 volt below and 1 volt above the setpoint.

To keep customers within 114–126 volts throughout that bus-voltage swing, we need to reserve 1 volt at each edge of the acceptable range. That reduces the theoretical feeder voltage-drop budget from 12 volts to 10 volts, before allowing for transformer and secondary drops.

Given this, it might make sense to decrease the bandwidth setting, but that can cause the tap changer to operate much more often, increasing wear and tear.

Let's ignore complicated branching for now and think of the line as a single, even voltage drop from the sub to the farthest part of the line. Using only the bus voltage slider, see if you can fit the whole shaded strip inside the acceptable voltage range. Then click Beyond Regulation and try again. The strip shows the regulation band: the bus can sit anywhere within 1 volt of its setpoint, and the whole profile moves with it.

As the demo shows, long feeders with substantial impedance and high loading can exceed these limits, making it impossible to regulate the line voltage from the bus alone.

One common solution is to install additional voltage regulators farther down the circuit. These regulators can raise the voltage locally without raising the voltage at every customer closer to the substation.

Shunt capacitor banks help, too. By supplying reactive power locally, they reduce the reactive current, and with it the X sin θ part of the drop, on every section between the bank and the substation. Reconductoring with larger wire reduces R and X directly.

But those are topics of their own. Here, we'll concentrate on regulating the substation bus.

What Happens When the Load Changes?

So far, we've looked at a feeder under a single set of loading conditions.

But in the real world, electrical loads are constantly changing, sometimes rapidly.

People turn on heating and cooling equipment, commercial buildings begin their workdays, industrial loads cycle on and off, and residential demand changes throughout the day.

As these loads change, the current flowing through different sections of the distribution system changes with them.

And because voltage drop depends on current, the voltage drop along the feeder changes throughout the day.

Let's consider a feeder with two different loading conditions.

During light loading, maybe the voltage drop between the substation bus and our distant customer is only 1.5 volts.

During heavy loading, maybe it increases to 6.5 volts.

With our substation bus regulated at a constant 122.5 volts, we get the following results:

Notice that both conditions are perfectly acceptable.

Our distant customer never experiences an undervoltage condition.

So what's the problem?

Strictly speaking, there isn't one. Our fixed bus setpoint is doing its job.

But while we've maintained a constant voltage at the substation, the voltage at our distant customer has changed by five volts.

A fixed bus can keep this customer in range all day, as long as the peak drop fits the budget, and varying the bus won't change that. What varying the bus can change is the rest of the day: less voltage swing at the far end, and lower voltage everywhere at light load. That second effect becomes important in Part 3.

Suppose instead that we want to maintain approximately 118 volts at that distant customer as loading changes.

Maybe we have an important customer, like a factory, that requires more tightly controlled voltage.

We could accomplish this by varying the substation bus voltage as the loading changes.

During light loading, when we're losing only 1.5 volts along the circuit, we would need a substation bus voltage of just 119.5 volts.

During heavy loading, when we're losing 6.5 volts, we would need to raise the bus to 124.5 volts.

Our results would look like this:

In this idealized example, we've accomplished two things.

First, we've reduced the voltage variation experienced by our distant customer. Instead of seeing a five-volt swing between light and heavy loading, the customer's voltage remains approximately constant.

Second, we've allowed the substation voltage to fall during light loading, when the additional voltage isn't needed to support this customer. We would still need to check the highest-voltage customers closer to the substation.

But for now, the important point is that we've changed the way the substation voltage responds to changes in feeder loading.

Rather than maintaining a constant voltage at the substation regardless of conditions, we're adjusting the bus voltage to compensate for changes in the electrical system downstream.

Introducing Line-Drop Compensation

This is the core idea behind line-drop compensation, commonly called LDC.

We use the load current, measured by a current transformer (CT) on the power transformer, to estimate how much additional voltage is needed to compensate for voltage drop in the distribution system.

Under ordinary forward power flow and lagging-power-factor loading, light load generally calls for little compensation.

As loading increases, the estimated feeder drop increases, so the controller calls for a higher bus voltage to compensate.

In effect, LDC makes the desired bus voltage depend on load current instead of holding it to one fixed target.

Inside the controller, the same idea usually runs the other way around. The controller subtracts its estimate of the line drop from the measured bus voltage to get a compensated voltage, its estimate of the voltage out at the regulation point, and regulates that to the bandcenter. Either way, the result is the same:

• With LDC enabled, the bandcenter is the voltage you want at the regulation point, not at the bus.
• The bus runs at roughly the bandcenter plus the compensation.
• At no load, the compensation is zero and the bus sits at the bandcenter.

That's why LDC alone doesn't lower the bus at light load. It only adds voltage as current rises. To get the lower light-load bus voltage from the last example, the bandcenter itself has to be set lower. LDC is a control strategy, not something that happens automatically just by enabling it.

There's an important distinction here.

LDC doesn't actually reduce the impedance of the feeder, and it doesn't eliminate voltage drop. It simply changes the regulated bus voltage to compensate for the expected drop.

It doesn't create any extra headroom, either. Whether a feeder can be held within range depends only on its peak drop, so LDC can't rescue a feeder that's beyond regulation.

It also doesn't necessarily measure the voltage at the distant location we're trying to regulate.

Instead, it estimates the required compensation using measurements available at the substation.

For this to work, the controller needs some way of relating the measured current to the expected voltage drop in the distribution system.

There are two common approaches to doing this.

Traditional R/X Compensation

The traditional method uses the resistance and reactance of the line to estimate voltage drop.

The controller measures voltage and current, including their phase relationship, and uses configured resistance and reactance values to estimate the voltage at a selected point downstream. (Remember that voltage drop formula from earlier?)

This is often described as regulating a theoretical load center, although the chosen point is better understood as a regulation point rather than necessarily the mathematical center of all connected loads.

This approach is useful when we're regulating a specific feeder with reasonably well-understood electrical characteristics.

However, things become more complicated when a single substation bus supplies several feeders with different lengths, impedances, and loading patterns.

In that situation, there may not be one set of resistance and reactance values that accurately represents the voltage behavior of the entire distribution system.

Z Compensation

Another approach, used in Beckwith tap changer controllers, is known as Z compensation.

Despite the name, this method does not require us to enter the actual complex impedance of a particular feeder.

Instead, we specify a desired bus-voltage increase at a reference level of load current. On a Beckwith M-2001D, that's the voltage raise at the CT's rated output of 200 mA.

For forward-load operation, the controller then increases the compensation in proportion to the magnitude of the measured current. With the raise set to 7 V, for example, the controller adds 1.75 V when the CT is delivering a quarter of its rating (50 mA).

In other words, we're establishing a relationship between load current and the amount of voltage boost applied at the substation.

This makes Z compensation particularly useful for substation bus regulation, where one transformer may supply several different feeders. Beckwith's application note describes exactly that case: no single line impedance applies, the transformer CT measures the total current from all the feeders, and the way that current splits between feeders isn't constant.

Unlike traditional R/X compensation, Z compensation doesn't attempt to calculate the precise voltage at a particular downstream location using line impedance and current phase angle.

It's a more general method of increasing bus voltage as system loading increases.

But that simplicity comes with tradeoffs.

Power factor is the clearest example. The real drop on the feeder still follows I(R cos θ + X sin θ), but Z ignores θ. Tuned at one power factor, Z under-compensates when the power factor gets worse, and over-compensates when it improves or goes leading. R/X tracks those changes, but only for the one impedance it was set for.

Because the controller is responding to measured current magnitude rather than the actual voltage at every customer, its compensation may not always match the voltage drop occurring on a particular feeder.

Different loading patterns, changes in power factor, and unequal loading among feeders can all affect how closely that compensation matches the actual voltage drop on any one feeder.

Multiple-Feeder Substations

We've touched on this twice already, but it's worth looking at directly: a single LTC controlling a bus that supplies several feeders.

Each feeder has different impedance, loading, and voltage requirements.

Whichever compensation method we use at the bus, the controller responds to the current measured through the transformer. That current represents the combined loading of the connected feeders, not necessarily the voltage drop on any particular one. Z compensation doesn't solve this problem; it simply stops pretending to know one feeder's impedance.

An increase in current on a relatively short feeder could cause the controller to raise the bus voltage even though the additional voltage isn't needed elsewhere.

Conversely, a long feeder with substantial voltage drop might require more compensation than the controller provides.

This makes choosing appropriate LDC settings more complicated, particularly when we're also trying to lower overall operating voltage, as we'll see with conservation voltage reduction (CVR) in Part 3. The controller lab at the end lets you try it: its Short Lines / High Load preset draws about three times the current of Long Lines / Low Load, yet has less voltage drop.

First-Customer Protection

Raising the substation voltage to support distant customers can also raise the voltage of customers close to the substation. If the LDC settings are too aggressive, or several feeders peak at once, the bus can climb past what the first customer should see.

Tap changer controllers guard against this with an upper voltage limit, commonly called first-house protection. Beckwith's manual calls it the First Customer Protector. It has two parts:

• Block raise: above an overvoltage limit, set somewhere above the top of the band, the controller stops raising the bus, no matter how much compensation it calculates.
• Runback: if the voltage climbs higher still, past the limit plus a deadband of about two tap steps, the controller lowers the tap immediately, with no time delay.

First-customer protection protects the first customer. It can't do anything for the last one: whenever it's holding the bus down, it's withholding compensation the distant customers were counting on.

Consequently, LDC settings must still be selected carefully to avoid creating overvoltage conditions in the first place.

How Loads Respond to Lower Voltage

Distribution loads are commonly modeled as a mix of three idealized types: constant impedance, constant current, and constant power. Together, these are known as ZIP loads (the Z here refers to impedance and has no relation to Z compensation). Holding everything else constant, each type responds differently when the voltage at its terminals drops by 5%:

Constant impedance loads draw less current as voltage decreases. A 5% voltage reduction results in approximately a 9.75% reduction in power consumption, because power varies with the square of the voltage when resistance stays fixed.

Constant current loads maintain approximately the same current as voltage changes, so power consumption changes proportionally with voltage. True constant-current behavior is relatively uncommon at the AC input of individual appliances, but it's a useful component of aggregate distribution load models over a limited operating range.

Constant power loads maintain approximately the same power consumption regardless of voltage. Many modern electronic power supplies behave this way. When voltage decreases, they draw additional current to maintain power: a 5% voltage reduction raises their current by about 5.3%, to 1 / 0.95 of its original value. This can actually increase conductor losses on the distribution system.

Real feeders contain a mixture of all three.

This is the simplification we made earlier. The feeder demonstrations on this page hold each load's current fixed, so every load behaves like a constant-current load. On a real feeder, lowering the bus also changes the load currents, and with them the feeder drop.

So where does the energy reduction come from?

Some loads simply operate at reduced output without compensating. Incandescent lighting was the classic example: less voltage, less power, less light. It was largely phased out of U.S. general-service lighting in 2023, and the LED lighting that replaced it behaves much more like a constant-power load. Certain continuously operating fans and motors may consume less power, although their mechanical output may also decrease. Resistive heaters will produce less heat, but their thermostats will run them longer to reach the same setpoint, so they draw less power but use about the same energy over time.

There can also be efficiency improvements. Distribution transformers generally experience lower excitation losses at reduced voltage, and some lightly loaded motors may operate more efficiently. Other motors may experience increased losses.

In other words, reducing instantaneous power doesn't necessarily reduce energy consumption, and reducing energy consumption doesn't necessarily mean equipment is operating more efficiently.

The actual results depend on the mixture of connected loads and how they respond over time.

And that mixture keeps changing. LED lighting, electronic power supplies, and inverter-driven heat pumps all tend to behave more like constant-power loads than the equipment they replaced.

What the Studies Show

CVR has been studied for decades, both through field measurements and computer simulations.

One relevant study was published in 2010 by Pacific Northwest National Laboratory (PNNL). Researchers simulated 24 representative distribution feeders across the United States, incorporating different feeder configurations, load characteristics, and operating conditions.

The simulations accounted for thermostatically controlled equipment, including the increased operating time associated with reduced heating output.

Depending on the feeder, simulated peak demand and annual energy fell by roughly 0.5–4%.

The simulated scheme lowered the substation regulator's voltage while holding the end-of-line primary voltage at about 118 volts.

Extrapolating the results nationally, the researchers estimated that implementing CVR across all distribution feeders could reduce electrical energy consumption by approximately 3.04%.

Keep the date in mind, though. That load mix predates most of the shift to LED lighting and electronic loads we just described, so treat this as a ballpark for older load mixes rather than a forecast.

The study also found that reductions in distribution losses accounted for only about 1–2% of the savings. The other 98–99% came from changes in customer consumption.

PNNL-19596 — Evaluation of Conservation Voltage Reduction on a National Level (2010)

CVR and Line-Drop Compensation

Let's go back to the feeder from earlier, with 1.5 volts of drop at light load and 6.5 volts at heavy load, originally regulated at a fixed 122.5 volts.

Suppose we implement CVR by lowering the bandcenter to 119.5 volts. At light load, our distant customer drops from 121.0 to 118.0 volts. That's the reduction we wanted. But at heavy load, the distant customer falls to 113.0 volts, below the 114-volt minimum.

One approach is to combine the lower CVR bandcenter with line-drop compensation.

This allows the controller to maintain a relatively low bus voltage during light loading while raising it as feeder current increases.

However, LDC can partially or completely offset the voltage reduction we're trying to achieve with CVR.

Suppose we keep the 119.5-volt bandcenter and enable Z compensation.

At light loading, suppose the controller calculates one volt of compensation:

119.5 + 1.0 = 120.5 volts

We're operating two volts below our original bus voltage.

Now suppose loading increases and the controller calculates five volts of compensation:

119.5 + 5.0 = 124.5 volts

We're now operating two volts above our original setpoint.

Here's the whole comparison, showing bus and distant-customer voltage:

Z compensation fixed the heavy-load undervoltage. But at heavy load, every customer on the feeder is now two volts above where they started.

The controller is behaving exactly as configured, but under these conditions, the LDC compensation has more than offset the CVR bandcenter reduction.

That result comes from where we set the bandcenter, not from LDC itself. The 119.5 V setting lifts our distant customer from 116 to 118 volts at heavy load, and those two extra volts show up on every customer. Let's add two more settings: the lowest fixed bus that stays in range, and the same Z compensation with the bandcenter lowered as far as it will go.

A fixed 120.5-volt bus matches our 119.5 + Z setting at light load and beats it by four volts at heavy load. If that were the best LDC could do, there'd be no reason to use it.

Now look at the last row. It uses the same Z compensation with the bandcenter at 115.5 volts. It matches the best fixed bus at heavy load and runs four volts lower at light load. A fixed bus can only put the distant customer at the floor at peak. LDC can keep that customer near the floor all day.

That's the real case for combining LDC with CVR. It's also why LDC can erase the reduction: if the bandcenter isn't lowered enough to pay for the boost, the bus ends up above where it started. And on a bus with several feeders, current doesn't track any one feeder's drop, so the boost can't be tuned this tightly.

Does It Average Out?

A feeder that spends most of its time lightly loaded may operate near the lower CVR bandcenter for much of the day, with higher voltages occurring only during periods of heavy loading.

But we can't determine the energy savings simply by averaging bus voltages.

Heavy-load periods account for a greater share of energy consumption. If LDC raises voltage above the original operating level during those periods, the additional consumption could offset savings achieved during lighter loading.

The result depends on the loading profile, the actual customer-voltage changes, and the voltage sensitivity of the connected loads.

To know what effect CVR has on any given circuit, real world data measured from actual customers needs to be collected and analyzed.