Skip to content
LumenaryAcademy
Power 70 min

AC power and three-phase systems

Real, reactive and complex power, power factor and how to correct it, and why almost all the world's electricity is made and moved on three wires.

Code tested · 4AI-draftedAwaiting review by an electrical engineer

Your guides: Chloe · The Physicist, Margaret · The Bench Engineer, Gideon · The Design Reviewer

MotivationH

Why does the grid run on three wires?

Look up at a high-voltage transmission line and count the conductors on each side of the tower: three, or a multiple of three. Walk past the substation at the edge of a town, or the switch room of a factory, a hospital or a data centre: the busbars come in threes. Inside almost every power station on Earth, the generator has three sets of windings, and the electricity it makes is three AC voltages, each a third of a cycle behind the one before.

Only at the very end of the chain, in the socket in your kitchen, does the power arrive as a single phase: one live conductor and a neutral, taken from one of the three. That is why most people never meet three-phase power, and why it is worth a lesson.

AC circuits, phasors and impedance taught you to turn sinusoids into complex numbers, and introduced complex power and the power factor for a single load. Here we take those tools to the scale of power systems: where complex power comes from in the time domain and why it is conserved, how to correct the power factor of a whole factory, and why the answer to "how should we move power?" favours three phases so strongly that the whole world adopted them within a few decades of the first demonstrations in the 1880s and 1890s.

Note

Safety. Everything in this lesson is calculation and simulation. Mains electricity, and three-phase supplies above all (400 V between lines in much of the world), can kill. Never open a distribution board, a socket, a plug or anything connected to the mains: that work is for qualified electricians with the right training, tools and isolation procedures. Safe experiments with electricity use small batteries or a current-limited low-voltage bench supply.

DefinitionsH

Power in an AC circuit

Start again from the time domain, because the three-phase argument later needs it. Take one load fed from an AC source. Its voltage and current are sinusoids at the same angular frequency ω=2πf\omega = 2\pi f, with RMS values VV and II, and the current lags the voltage by an angle φ\varphi (for an inductive load φ>0\varphi > 0; for a capacitive load φ<0\varphi < 0):

v(t)=2 Vcos⁡ωt,i(t)=2 Icos⁡(ωt−φ).v(t) = \sqrt2\,V\cos\omega t, \qquad i(t) = \sqrt2\,I\cos(\omega t - \varphi).

The instantaneous power is their product, and the identity 2cos⁡Acos⁡B=cos⁡(A−B)+cos⁡(A+B)2\cos A\cos B = \cos(A-B) + \cos(A+B) splits it in two:

p(t)=v(t) i(t)=VIcos⁡φ⏟constant  +  VIcos⁡(2ωt−φ)⏟oscillates at twice the supply frequency.p(t) = v(t)\,i(t) = \underbrace{VI\cos\varphi}_{\text{constant}} \;+\; \underbrace{VI\cos(2\omega t - \varphi)}_{\text{oscillates at twice the supply frequency}}.

The oscillating term averages to zero over a cycle, so the average power, the rate at which energy actually leaves the source for good, is VIcos⁡φVI\cos\varphi. When φ≠0\varphi \neq 0, p(t)p(t) even dips below zero for part of each cycle: energy stored in the load's magnetic field (or electric field) flows back to the source. That energy shuttles to and fro twice per cycle, doing no net work, yet the current that carries it is real current that heats every wire on the way.

Here, as a reminder from AC circuits, phasors and impedance, are the quantities engineers use.

Definition— Real, reactive, apparent and complex power

With phasors V=V∠θv\mathbf V = V\angle\theta_v and I=I∠θi\mathbf I = I\angle\theta_i (RMS magnitudes) and φ=θv−θi\varphi = \theta_v - \theta_i:

  • Complex power S=V I∗=P+jQ\mathbf S = \mathbf V\,\mathbf I^{*} = P + jQ, where ∗^{*} is the complex conjugate.
  • Real power P=VIcos⁡φP = VI\cos\varphi, in watts (W): the average power, the part that does work.
  • Reactive power Q=VIsin⁡φQ = VI\sin\varphi, in volt-amperes reactive (var): the amplitude of the energy that sloshes back and forth. By convention Q>0Q > 0 for an inductive load, which is said to absorb reactive power, and Q<0Q < 0 for a capacitive load, which supplies it.
  • Apparent power ∣S∣=VI=P2+Q2|\mathbf S| = VI = \sqrt{P^2 + Q^2}, in volt-amperes (VA): what the wires and transformers must be sized for, because their heating depends on current, whatever kind of power it carries.
  • Power factor pf=P/∣S∣=cos⁡φ\mathrm{pf} = P/|\mathbf S| = \cos\varphi, called lagging when the current lags the voltage (inductive) and leading when it leads (capacitive).

Three facts make complex power the workhorse of power engineering. It is a single complex number that carries both PP and QQ. For an impedance Z=R+jXZ = R + jX carrying current II, S=VI∗=ZII∗=I2R+jI2X\mathbf S = \mathbf V\mathbf I^{*} = Z\mathbf I\mathbf I^{*} = I^2 R + jI^2 X: resistors absorb only real power, inductors (X>0X > 0) absorb reactive power, capacitors (X<0X < 0) supply it. And it is conserved:

Theorem— Conservation of complex power

In any circuit of sources and impedances at one frequency, the sum of the complex powers absorbed by all elements is zero. Equivalently, the complex power supplied by the sources equals the sum of the complex powers absorbed by the loads, so real and reactive power each balance separately.

Proof

Number the nodes, pick one as the reference, and write Vk\mathbf V_k for the voltage across element kk and Ik\mathbf I_k for its current, both in the same (passive) direction. Each element joins two nodes mm and nn, so Vk=Um−Un\mathbf V_k = \mathbf U_m - \mathbf U_n where U\mathbf U are node voltages. Then ∑kVkIk∗=∑nodes mUm(∑k leaving mIk−∑k entering mIk)∗.\sum_k \mathbf V_k \mathbf I_k^{*} = \sum_{\text{nodes } m} \mathbf U_m \left(\sum_{k \text{ leaving } m} \mathbf I_k - \sum_{k \text{ entering } m} \mathbf I_k\right)^{*}. By Kirchhoff's current law each bracket is zero, so the sum is zero. (This is Tellegen's theorem; it uses only KCL and the fact that voltages are differences of node potentials, so it holds whatever the elements are.)

∎

The power triangle pictures it: PP along the real axis, QQ up the imaginary axis, S\mathbf S the hypotenuse at angle φ\varphi.

VIφP = |S| cos φ (W)Q|S| (VA)φ
Left: phasors for an inductive load. The current I lags the voltage V by the angle phi. Right: the power triangle of the same load. Real power P is horizontal, reactive power Q vertical, and the apparent power |S| is the hypotenuse at angle phi; the power factor is cos phi.
Worked examplesH

Power factor and its correction

Most industrial loads are inductive: induction motors need magnetising current, transformers too, and fluorescent ballasts and welding sets add more. A factory full of motors may run at a power factor of 0.7 to 0.8 lagging. That costs money upstream. For the same real power, the current is 1/pf1/\mathrm{pf} times larger than it needs to be, so cables, switchgear and transformers must be larger, and I2RI^2R losses in them are 1/pf21/\mathrm{pf}^2 times larger. Many utilities therefore charge large customers for reactive energy or for a low power factor.

The fix is to supply the reactive power locally. A capacitor connected in parallel with the load supplies QC=V2/XC=ωCV2Q_C = V^2/X_C = \omega C V^2 of reactive power, cancelling part of the load's QQ. The load itself is unchanged; it draws exactly the same current as before. What changes is the current in everything upstream of the capacitor, because the reactive current now circulates only between the capacitor and the load.

The calculation is the power triangle twice. To raise the power factor from pf1\mathrm{pf}_1 to pf2\mathrm{pf}_2 at constant PP:

QC=P(tan⁡φ1−tan⁡φ2),φi=arccos⁡pfi.Q_C = P\left(\tan\varphi_1 - \tan\varphi_2\right), \qquad \varphi_i = \arccos \mathrm{pf}_i .

Designers usually aim for 0.95 or so, not 1.0. The last few per cent of correction need the most kvar for the least benefit (the slope of tan⁡\tan is steepest near 0.95 to 1.0), and a bank sized for unity at full load would make the power factor leading when the load is light, which can push voltages up.

We will size a real bank in a moment, once we know how three-phase loads are connected.

IntuitionH

Generating three phases

Put a magnet on a shaft inside a ring of iron and wind a coil on the ring. As the shaft turns at a steady speed, the flux through the coil rises and falls, and the coil produces a sinusoidal voltage, one cycle per revolution for a two-pole magnet. Now wind two more identical coils on the same ring, rotated 120∘120^\circ and 240∘240^\circ round from the first. The magnet passes each coil a third of a revolution after the previous one, so the three voltages are identical sinusoids shifted by a third of a cycle:

va=2Vcos⁡ωt,vb=2Vcos⁡(ωt−120∘),vc=2Vcos⁡(ωt−240∘).v_a = \sqrt2 V\cos\omega t,\qquad v_b = \sqrt2 V\cos(\omega t - 120^\circ),\qquad v_c = \sqrt2 V\cos(\omega t - 240^\circ).

The order in which they peak, aa then bb then cc, is the phase sequence. In phasors, with the operator a=1∠120∘=ej2π/3a = 1\angle 120^\circ = e^{j2\pi/3},

Va=V∠0∘,Vb=a2 Va=V∠−120∘,Vc=a Va=V∠120∘.\mathbf V_a = V\angle 0^\circ,\qquad \mathbf V_b = a^2\,\mathbf V_a = V\angle{-120^\circ},\qquad \mathbf V_c = a\,\mathbf V_a = V\angle 120^\circ.

Since aa is a cube root of unity other than 1, 1+a+a2=01 + a + a^2 = 0: the three phasors of a balanced set add to zero. At every instant the three voltages sum to zero too. Almost everything special about three-phase systems follows from that one identity.

abcωt0one cycle (20 ms at 50 Hz)VaVbVc120°
A balanced three-phase set. Left: the three phase voltages over one cycle, each a third of a cycle (120 degrees) behind the one before; at every instant they add to zero. Right: their phasors, equal in length and 120 degrees apart.
DefinitionsH

Wye and delta: line and phase quantities

The three sources (or three loads) can be connected in two ways.

Definition— Wye (star) and delta connections

In a wye (Y, or star) connection, one end of each of the three elements is joined at a common neutral point, and the other ends go to the three lines. In a delta (Δ\Delta) connection, the three elements form a triangle, each connected between two lines, and there is no neutral. Phase quantities are the voltage across and current through one element; line quantities are the voltage between two lines and the current in one line.

In a wye, each line current is a phase current, but each line voltage is the difference of two phase voltages. In a delta, each line voltage is a phase voltage, but each line current is the difference of two phase currents. With 1+a+a2=01 + a + a^2 = 0 the differences are easy:

Vab=Va−Vb=Va(1−a2)=3 Va∠30∘(wye),Ia=Iab−Ica=Iab(1−a)=3 Iab∠−30∘(delta).\mathbf V_{ab} = \mathbf V_a - \mathbf V_b = \mathbf V_a(1 - a^2) = \sqrt3\,\mathbf V_a\angle 30^\circ \quad\text{(wye)}, \qquad \mathbf I_a = \mathbf I_{ab} - \mathbf I_{ca} = \mathbf I_{ab}(1 - a) = \sqrt3\,\mathbf I_{ab}\angle{-30^\circ} \quad\text{(delta)}.

So line voltage is 3\sqrt3 times phase voltage in a wye, line current is 3\sqrt3 times phase current in a delta, and the other quantity is equal. That 3\sqrt3 is where the familiar pairs come from: 230/400 V in Europe, India, Africa and much of Asia (2303≈400230\sqrt3 \approx 400), 120/208 V and 277/480 V in North American commercial buildings. A European house usually takes one phase and the neutral (230 V); a workshop takes all three lines (400 V between them).

Whatever the connection, the total complex power of a balanced load is three times one phase's:

S3ϕ=3 VphIph∗,∣S3ϕ∣=3VphIph=3 VLIL,P3ϕ=3 VLILcos⁡φ,\mathbf S_{3\phi} = 3\,\mathbf V_{\text{ph}}\mathbf I_{\text{ph}}^{*}, \qquad |\mathbf S_{3\phi}| = 3V_{\text{ph}}I_{\text{ph}} = \sqrt3\,V_{\text{L}}I_{\text{L}}, \qquad P_{3\phi} = \sqrt3\,V_{\text{L}}I_{\text{L}}\cos\varphi,

where φ\varphi is the angle of the phase impedance, not the angle between a line voltage and a line current (those differ by 30∘30^\circ).

abcNneutralwye (star)abcdelta
Left: a wye-connected load, three impedances meeting at a neutral point N; each line current flows through one impedance, which sees the line-to-neutral voltage. Right: a delta-connected load, three impedances forming a triangle between the lines; each impedance sees a full line-to-line voltage, and each line current is the difference of two branch currents.
IntuitionH

Why three phases?

Three phases won for three separate reasons, any one of which would have been worth having.

1. Constant power. One phase delivers power that pulses at twice the supply frequency, as we saw. A balanced three-phase system does not.

Theorem— Balanced three-phase power is constant

If three phases carry voltages 2Vcos⁡(ωt−θk)\sqrt2 V\cos(\omega t - \theta_k) and currents 2Icos⁡(ωt−θk−φ)\sqrt2 I\cos(\omega t - \theta_k - \varphi) with θk=0∘,120∘,240∘\theta_k = 0^\circ, 120^\circ, 240^\circ, the total instantaneous power is constant: p(t)=3VIcos⁡φp(t) = 3VI\cos\varphi at every instant.

Proof

Each phase contributes VIcos⁡φ+VIcos⁡(2ωt−2θk−φ)VI\cos\varphi + VI\cos(2\omega t - 2\theta_k - \varphi), as in the single-phase calculation. The angles 2θk2\theta_k are 0∘0^\circ, 240∘240^\circ and 480∘≡120∘480^\circ \equiv 120^\circ, so the three oscillating terms are cosines of equal amplitude 120∘120^\circ apart, whose sum is the real part of VI ej(2ωt−φ)(1+a+a2)=0VI\,e^{j(2\omega t - \varphi)}(1 + a + a^2) = 0. Only the three constant terms remain.

∎

Constant electrical power means constant torque on the generator's shaft and on a motor's: no vibration at twice the supply frequency, smoother running, longer bearings. A large single-phase machine would shake itself and its foundations at 100 or 120 Hz.

2. Less conductor. Suppose you must deliver power PP over a distance ℓ\ell at unity power factor, with the voltage VV between any two conductors fixed (it sets the insulation) and a fixed total loss PlossP_{\text{loss}}. With two wires, the current is P/VP/V, the loss is 2(P/V)2ρℓ/A12(P/V)^2\rho\ell/A_1, and the copper volume needed comes to 4ρℓ2P2/(V2Ploss)4\rho\ell^2P^2/(V^2P_{\text{loss}}). With three wires, the current is P/(3V)P/(\sqrt3 V) in each, the loss is 3 (P/3V)2ρℓ/A33\,(P/\sqrt3 V)^2\rho\ell/A_3, and the volume comes to 3ρℓ2P2/(V2Ploss)3\rho\ell^2P^2/(V^2P_{\text{loss}}). Three-phase needs three quarters of the copper, and in a balanced system no neutral return is needed at all, because the three line currents add to zero.

3. A rotating magnetic field. Feed three coils spaced 120∘120^\circ apart round a stator with three currents 120∘120^\circ apart in time, and the magnetic field they make together has constant strength and turns steadily at the supply frequency, without any moving part. Put a conductor or a magnet in that field and it is dragged round: that is the induction motor and the synchronous motor of Transformers and electric machines. A single coil only makes a field that pulses back and forth along one axis.

The first program checks reasons 1 and 3 numerically.

python
import numpy as np

f = 50.0                     # hertz
w = 2 * np.pi * f
V, I = 230.0, 10.0           # RMS voltage and current of one phase
phi = np.arccos(0.8)         # the current lags the voltage by phi: power factor 0.8 lagging

# Complex power from the phasors: S = V I*
V_ph = V * np.exp(1j * 0.0)
I_ph = I * np.exp(-1j * phi)
S = V_ph * np.conj(I_ph)
print(f"S = {S.real:.0f} + j{S.imag:.0f} VA,  |S| = {abs(S):.0f} VA,  pf = {S.real / abs(S):.2f}")

t = np.linspace(0, 1 / f, 2000, endpoint=False)     # one cycle of the supply
def p_phase(shift):
    """Instantaneous power v(t) i(t) of one phase whose voltage is delayed by `shift` radians."""
    v = np.sqrt(2) * V * np.cos(w * t - shift)
    i = np.sqrt(2) * I * np.cos(w * t - shift - phi)
    return v * i

p1 = p_phase(0)
p3 = sum(p_phase(k * 2 * np.pi / 3) for k in range(3))
print(f"one phase:    p(t) from {p1.min():6.0f} to {p1.max():6.0f} W, average {p1.mean():5.0f} W")
print(f"three phases: p(t) from {p3.min():6.0f} to {p3.max():6.0f} W, average {p3.mean():5.0f} W")

print("\nOne phase over one cycle (each # is 250 W; | is zero):")
for k in range(0, 2000, 125):
    n = int(round(p1[k] / 250))
    bar = " " * (2 + min(n, 0)) + "#" * max(-n, 0) + "|" + "#" * max(n, 0)
    print(f"t = {t[k] * 1000:5.2f} ms {int(round(p1[k])):6d} W  {bar}")

# The field of three coils 120 degrees apart round a stator, each along its own axis.
coil_angles = np.radians([0, 120, 240])
print("\n t (ms) | three coils, balanced currents: size  angle | one coil alone: field")
for t_ms in np.arange(0, 20.1, 2.5):
    i3 = np.cos(w * t_ms / 1000 - coil_angles)               # currents 120 degrees apart in time
    bx, by = np.sum(i3 * np.cos(coil_angles)), np.sum(i3 * np.sin(coil_angles))
    single = round(np.cos(w * t_ms / 1000), 3) + 0.0          # one coil: a field along its own axis
    print(f"{t_ms:7.1f} | {np.hypot(bx, by):38.3f} {np.degrees(np.arctan2(by, bx)) % 360:6.0f} | {single:21.3f}")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The single phase delivers 1840 W on average but swings between −460-460 W and 41404140 W; the three phases deliver exactly 5520 W at every instant. The field of the three coils keeps a constant size, 1.5 times the peak field of one coil, and its direction advances 45∘45^\circ every 2.5 ms, one full turn per cycle of the supply. The lone coil's field just grows, shrinks and reverses.

ComputationC

Balanced loads and the per-phase method

A balanced three-phase system has three equal source voltages 120∘120^\circ apart and three identical impedances in every set of three. In a balanced wye–wye system the neutral points are at the same potential (the neutral current is Ia+Ib+Ic=Ia(1+a2+a)=0\mathbf I_a + \mathbf I_b + \mathbf I_c = \mathbf I_a(1 + a^2 + a) = 0), so you can imagine them joined by a perfect wire. Then each phase is an independent single-phase circuit, and the other two are copies rotated by 120∘120^\circ and 240∘240^\circ. That is the per-phase method:

  1. Replace every delta by its equivalent wye. For identical impedances, ZY=ZΔ/3Z_Y = Z_\Delta/3.
  2. Draw phase aa alone, from its line-to-neutral source voltage through the line impedance to the load and back by the ideal neutral.
  3. Solve it as an ordinary AC circuit.
  4. Get phases bb and cc by multiplying by a2a^2 and aa; multiply powers by 3; convert back to line or delta quantities with the 3∠30∘\sqrt3\angle 30^\circ factors.

A balanced three-phase network of any size, a whole transmission system for a power-flow study, becomes a single-line problem. Engineers draw it that way too, as the one-line diagram you will meet in The power grid.

Example— A factory's capacitor bank

A factory is supplied at 400 V line to line, 50 Hz. Its load is 250 kW at a power factor of 0.78 lagging, mostly induction motors. The utility charges for a power factor below 0.95, so the engineer specifies a three-phase capacitor bank to reach 0.95.

Reactive power. Q1=Ptan⁡(arccos⁡0.78)Q_1 = P\tan(\arccos 0.78) and Q2=Ptan⁡(arccos⁡0.95)Q_2 = P\tan(\arccos 0.95); the bank supplies the difference, about 118 kvar in total, a third per phase.

Capacitance. Each capacitor of a delta-connected bank sits between two lines and sees 400 V, so CΔ=(QC/3)/(ω⋅4002)C_\Delta = (Q_C/3)/(\omega \cdot 400^2). Connected in wye, each capacitor sees only 400/3≈231400/\sqrt3 \approx 231 V and must be three times larger to supply the same kvar. Delta is the usual choice at low voltage because capacitance costs more than voltage rating; at medium and high voltage, banks are usually wye-connected, partly because each unit then needs only phase-voltage insulation.

Benefit. The line current falls in proportion to pf1/pf2\mathrm{pf}_1/\mathrm{pf}_2, so losses in the supply cable fall in proportion to (pf1/pf2)2(\mathrm{pf}_1/\mathrm{pf}_2)^2. The motors draw exactly what they drew before.

The second program does this example, after a per-phase solution of a delta load fed through a cable.

python
import numpy as np

a = np.exp(2j * np.pi / 3)               # the operator a = 1 at 120 degrees
f, V_LL = 50.0, 400.0
w = 2 * np.pi * f
V_an = V_LL / np.sqrt(3)                 # phase a's voltage to neutral, our reference phasor (angle 0)

# --- 1. The per-phase method: a delta load fed through a cable -------------------------------
Z_delta = 12 + 9j                        # ohms, each of the three load branches
Z_line = 0.15 + 0.10j                    # ohms, each conductor of the feeding cable
Z_Y = Z_delta / 3                        # the equivalent wye load
I_a = V_an / (Z_line + Z_Y)              # solve phase a alone
I_lines = I_a * np.array([1, a**2, a])   # phases b and c lag by 120 and 240 degrees
V_load_an = I_a * Z_Y
S_load = 3 * V_load_an * np.conj(I_a)
print(f"line currents: {abs(I_lines[0]):.2f} A at angles "
      f"{', '.join(f'{np.degrees(np.angle(x)):.1f}' for x in I_lines)} degrees")
print("the three line currents add to zero:", abs(I_lines.sum()) < 1e-9)
print(f"load line voltage {abs(V_load_an) * np.sqrt(3):.1f} V;  load takes {S_load.real / 1e3:.2f} kW and "
      f"{S_load.imag / 1e3:.2f} kvar;  cable loses {3 * abs(I_a)**2 * Z_line.real / 1e3:.2f} kW")

# --- 2. Correcting a factory's power factor with a capacitor bank --------------------------
P = 250e3                                # watts
pf_old, pf_new = 0.78, 0.95
Q_old = P * np.tan(np.arccos(pf_old))
Q_new = P * np.tan(np.arccos(pf_new))
Q_C = Q_old - Q_new                      # what the capacitors must supply
print(f"\nQ before {Q_old / 1e3:.1f} kvar, after {Q_new / 1e3:.1f} kvar: the bank supplies {Q_C / 1e3:.1f} kvar")

C_delta = Q_C / 3 / (w * V_LL**2)        # each capacitor sees the line voltage
C_wye = Q_C / 3 / (w * V_an**2)          # each capacitor sees the phase voltage
print(f"per-phase capacitance: delta {C_delta * 1e6:.0f} uF (rated {V_LL:.0f} V), "
      f"wye {C_wye * 1e6:.0f} uF (rated {V_an:.0f} V)")

I_old = P / (np.sqrt(3) * V_LL * pf_old)
I_new = P / (np.sqrt(3) * V_LL * pf_new)
print(f"line current {I_old:.0f} A -> {I_new:.0f} A; cable loss falls to {(I_new / I_old)**2:.0%} of before")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The delta load of (12+j9) Ω(12 + j9)\ \Omega per branch is (4+j3) Ω(4 + j3)\ \Omega per phase in wye; through the cable it draws 44.58 A per line, the three currents 120∘120^\circ apart, and the cable's resistance costs 0.89 kW while the load voltage sags to 386.1 V. For the factory, the bank must supply 118.4 kvar: 785 µF per phase rated for 400 V in delta, or 2355 µF rated for 231 V in wye.

DefinitionsH

The per-unit system

A power system mixes voltages from 400 V to 400 kV, and impedances that look tiny on one side of a transformer and huge on the other. Engineers tame this by expressing every quantity as a fraction of a base value.

Definition— Per-unit values

Choose a three-phase base power SbaseS_{\text{base}} (for a whole system, often 100 MVA) and, in each voltage zone, a base line-to-line voltage VbaseV_{\text{base}} (usually the nominal voltage). Then Ibase=Sbase3 Vbase,Zbase=Vbase2Sbase,I_{\text{base}} = \frac{S_{\text{base}}}{\sqrt3\,V_{\text{base}}}, \qquad Z_{\text{base}} = \frac{V_{\text{base}}^2}{S_{\text{base}}}, and any quantity in per-unit (pu) is its actual value divided by its base. The bases in neighbouring zones are linked by the transformers' turns ratios.

Three things make this worth the trouble. The 3\sqrt3s vanish: in per-unit, line and phase voltages of a balanced system have the same magnitude, and S=VI∗S = VI^{*} holds as written. Ideal transformers vanish: a 132/33 kV transformer with bases 132 kV and 33 kV maps 1 pu to 1 pu, so it becomes a wire (plus its own small impedance). Values fall in familiar ranges: a transformer's impedance is typically 0.05 to 0.15 pu on its own rating whatever its size, so an engineer can spot a data-entry error at a glance.

Example— Moving a transformer's impedance to the system base

A 50 MVA, 132/33 kV transformer has an impedance of 0.10 pu on its own rating (manufacturers quote it as 10%). The network study uses a 100 MVA base. Impedance in pu scales with SbaseS_{\text{base}} and inversely with Vbase2V_{\text{base}}^2, so znew=zold SnewSold(VoldVnew)2=0.10×10050×1=0.20 pu.z_{\text{new}} = z_{\text{old}}\,\frac{S_{\text{new}}}{S_{\text{old}}}\left(\frac{V_{\text{old}}}{V_{\text{new}}}\right)^2 = 0.10 \times \frac{100}{50} \times 1 = 0.20\ \text{pu}. In ohms that is 0.20×1322/100=34.848 Ω0.20 \times 132^2/100 = 34.848\ \Omega seen from the 132 kV side, or 0.20×332/100=2.178 Ω0.20 \times 33^2/100 = 2.178\ \Omega seen from the 33 kV side. The two differ by the turns ratio squared, (132/33)2=16(132/33)^2 = 16, as they must; the per-unit value is the same from either side.

Every power-system tool, from commercial packages to open-source ones such as pandapower and PyPSA, works in per-unit internally.

ExplanationH

50 Hz and 60 Hz around the world

Every AC grid runs at a nominal frequency of either 50 Hz or 60 Hz. The choice was made in the 1890s, by companies, before anyone thought of a world standard. Lower frequencies suit long lines and the big, slow machines of the day; higher frequencies give smaller transformers and stop electric lamps from flickering visibly. In Germany AEG settled on 50 Hz, and most of Europe followed; in the United States Westinghouse settled on 60 Hz. Each region's choice then spread with its equipment and engineering influence.

As of 2026, the pattern is roughly this (the IEC's World Plugs list gives each country):

  • 50 Hz: Europe, nearly all of Africa, most of the Middle East, most of Asia including China, India and Indonesia, Australia and New Zealand, and in South America Argentina, Chile, Uruguay and Paraguay among others.
  • 60 Hz: North and Central America, much of northern South America including Brazil, Colombia and Peru, and in Asia South Korea, Taiwan, the Philippines and Saudi Arabia.
  • Both: Japan. Eastern Japan, including Tokyo, runs at 50 Hz with generators first bought from Germany; western Japan, including Osaka, at 60 Hz with American ones. The two halves can exchange power only through a few frequency-converter stations, and that bottleneck limited how much help western Japan could send east after the earthquake and tsunami of 2011.

Some special systems use other frequencies: several European railways (in Germany, Austria, Switzerland, Sweden and Norway) run on single-phase 16.7 Hz, a legacy of early traction motors, and aircraft use 400 Hz because transformers and motors shrink as frequency rises, and weight is what matters in the air.

Neither frequency is better in any important way, and converting a country would cost more than any gain. Equipment rated for one usually tolerates the other if its voltage matches, but induction motors and old clocks run 20% faster or slower, and transformers designed for 60 Hz may overheat at 50 Hz, as their core flux rises by 20% for the same voltage.

ExplanationH

The professional view

In practice, three-phase power is the domain of power-system engineers at utilities, transmission and distribution operators, consultancies and large industrial sites. A few things they deal with every day:

  • Reactive power is managed, not just corrected. Generators supply or absorb reactive power to hold voltages up or down, and grid codes require them to do so over a specified power-factor range. Transmission operators add shunt capacitors and reactors, synchronous condensers (synchronous machines that spin without a load, purely to make or absorb vars), and STATCOMs, power-electronic converters that do the same in milliseconds (see Power electronics).
  • Capacitor banks are engineered, not just sized. Banks in factories are split into steps switched by an automatic controller as the load changes; they include discharge resistors because a disconnected capacitor stays charged to a dangerous voltage; and where there are many variable-speed drives, they include detuning reactors, because a capacitor bank and the supply's inductance form a resonant circuit that harmonic currents can excite.
  • Harmonics break the simple picture. Rectifiers in computers, LED drivers and drives draw non-sinusoidal currents. Then P2+Q2<∣S∣2P^2 + Q^2 < |S|^2, the power factor is no longer cos⁡φ\cos\varphi, and definitions must be extended; the IEEE's standard 1459 defines power quantities under non-sinusoidal and unbalanced conditions.
  • Studies are done in software. Power-flow, short-circuit and stability studies of whole networks run in per-unit, on one-line models of balanced systems, with symmetrical components for unbalanced faults. MIT's free Introduction to Electric Power Systems covers the classical methods in more depth.
History & global context

A short history

The rotating magnetic field was discovered twice in the 1880s: by Galileo Ferraris in Turin, who built a working model in 1885 and published in 1888, and by Nikola Tesla, whose polyphase motor patents were granted in 1888 and licensed by Westinghouse. In Germany, Mikhail Dolivo-Dobrovolsky at AEG developed the three-phase system with its induction motor, wye and delta connections and three-phase transformer, and in 1891 a three-phase line of about 175 km carried power from a hydroelectric plant at Lauffen am Neckar to the electrical exhibition in Frankfurt, the demonstration that convinced the industry. Two years later, Charles Proteus Steinmetz showed engineers how to use complex numbers for AC circuits, the method this lesson has used throughout. In 1918 Charles Fortescue published the method of symmetrical components, still the basis of fault calculations in every utility.

Common misconceptionsH

Common misconceptions

Misconception

"Three-phase needs three wires where single-phase needs two, so it uses 50% more copper." It uses fewer, thinner wires for the same job. For the same power, voltage between conductors and losses, three wires need only three quarters of the copper of two, as the calculation in this lesson shows, because each carries less current and, when the load is balanced, the three currents return through one another, with no separate return conductor.

Misconception

"Power-factor correction cuts my electricity use." It cuts the current and losses upstream of the capacitors, and reactive-power charges where a tariff has them. The load's real power and the energy it uses are unchanged. Gadgets sold to households with the promise of cutting the bill this way do essentially nothing: domestic meters bill real energy.

Misconception

"400 V three-phase means 400 V per phase." In a 230/400 V wye system each phase is 230 V to neutral; 400 V is the voltage between any two lines, 3\sqrt3 times larger.

Misconception

"The neutral of a three-phase system carries the return current of all three phases, so it must be three times as thick." In a balanced system the neutral current is zero. The real danger is the opposite surprise: loads such as rectifiers draw third-harmonic currents (and other multiples of three), and those are in phase in all three lines, so they add in the neutral instead of cancelling. In buildings full of electronic loads the neutral current can exceed the phase current, which is why such neutrals are sized up.

Exercises

Exercises

1. AC power and three-phase systems

Warm-upHuman

A single-phase load absorbs complex power S=6+j8\mathbf S = 6 + j8 kVA. Which statements are true? Select all that apply.

Choose all that apply

2. AC power and three-phase systems

Warm-upHuman

A balanced three-phase motor takes 30 kW of real power from a 400 V (line-to-line) supply at a power factor of 0.85 lagging. What is the line current, to the nearest ampere?

3. AC power and three-phase systems

CoreHuman

A bakery's three-phase load is 120 kW at a power factor of 0.70 lagging. How many kvar of capacitors (total, all three phases) raise its power factor to 0.95 lagging? Give the nearest whole number.

4. AC power and three-phase systems

CoreHuman

Two three-phase capacitor banks supply the same 60 kvar on the same 400 V, 50 Hz lines, one connected in delta and one in wye. How does the capacitance of each capacitor in the wye bank compare with each one in the delta bank?

Choose one

5. AC power and three-phase systems

StretchHuman

A 230/400 V wye-connected supply feeds three identical banks of rectifier loads, one on each phase. Each phase current contains 10 A RMS at the fundamental frequency and 6 A RMS at the third harmonic, and nothing else. What is the RMS current in the neutral, in amperes?

6. AC power and three-phase systems

CoreComputational

Write capacitor_bank(P, pf_old, pf_new, V_LL, f, connection), which returns (Q_C, C): the total reactive power in var that a three-phase capacitor bank must supply to raise a balanced load of real power P (watts) from lagging power factor pf_old to pf_new, and the capacitance in farads of each of its three capacitors, for connection equal to "delta" or "wye". The tests use the factory of the lesson (250 kW, 0.78 to 0.95, 400 V, 50 Hz) and a 60 Hz, 480 V case.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

7. AC power and three-phase systems

CoreComputational

Write two helpers for per-unit calculations on three-phase systems:

  • base_values(s_base, v_base) returns (I_base, Z_base) for a three-phase base power in VA and a line-to-line base voltage in V.
  • change_base(z_pu, s_old, v_old, s_new, v_new) converts an impedance from one base to another.

The tests include the transformer of the lesson (0.10 pu on 50 MVA, 132 kV, moved to 100 MVA) and check that a change of base never changes the impedance in ohms.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Further reading and resources

  • Introduction to Electric Power Systems (6.061) · James L. Kirtley, MIT OpenCourseWare · 2011Free lecture notes on three-phase circuits, per-unit systems, transformers and machines.
  • Complex quantities and their use in electrical engineering · Charles Proteus Steinmetz · 1893The paper, read at the International Electrical Congress in Chicago, that made complex numbers the everyday tool of AC engineering.
  • World Plugs · International Electrotechnical Commission (IEC)The IEC's list of the voltage, frequency and plug types used in each country.
  • pandapowerAn open-source Python library for power-flow and short-circuit studies of three-phase networks, all in per-unit underneath.
  • PyPSAPython for Power System Analysis, an open-source toolbox for simulating and optimising power systems.