Real power (kW) does the work; reactive (kVAR) builds magnetic fields and returns every cycle; apparent (kVA) is what the wiring carries: S² = P² + Q².
PF = P/S = cos φ, between 0 (all reactive) and 1 (no phase shift). It sets the current for the same useful power.
A capacitor leads where an inductor lags. In parallel with the load it supplies the reactive power locally, shrinking the current the cable carries. Q = 2πfCV².
Reactive power goes with V², so a star bank needs three times the capacitance of delta for the same kVAR; delta parts must be rated for full line voltage.
A correction calculation gives reactive power in kVAR, but you order a capacitor in microfarads, and the conversion depends on the voltage each capacitor actually sees: Q = 2πfCV², so the same kVAR needs completely different capacitance at 230 V and 400 V, or in delta versus star. Single phase uses the supply voltage; three-phase delta divides the total by three; star uses line/√3 per capacitor.
Cutting a 133 kVA draw to 105 kVA frees 28 kVA of transformer and cable headroom without changing a conductor. On a site near its supply limit that is often worth more than the reactive charge saved.
At fixed real power current goes as 1/PF, and I²R losses as the square of that. Going from 0.75 to 0.95 cuts the current 21% and the copper losses 38%.
It does not register on a kWh meter and costs nothing directly, but the current it forces through cables, transformers and switchgear causes real I²R losses: what UK DNOs are charging for. Every DNO builds its network around a design PF of 0.95; UK Power Networks’ connection terms require no less, and most apply reactive charges below it. Charges apply to non-domestic sites with half-hourly metering, not domestic bills.
Inductive loads (motors, transformers) make current lag voltage: the common case. Capacitive loads make it lead. Over-correcting flips a lagging PF into a leading one, which some tariffs also penalise; automatic banks switch stages to avoid overshooting past unity as the load changes.
A motor’s reactive draw (for its magnetic field) stays roughly constant regardless of load while its real power scales with the mechanical load, so a lightly loaded motor’s PF drops: the reactive share becomes proportionally larger.
Capacitors can sit right at a large motor, which also cuts losses in the cable feeding it, or as one bank at the main switchboard, correcting the site’s average, which is cheaper to install and maintain.
Every DNO in the UK builds its network around a design power factor of 0.95. A site running well below that forces more current down the same cables to deliver the same useful power, which is exactly why UK Power Networks' own connection terms require no less than 0.95 PF, and why most DNOs apply reactive-power charges below that threshold.
Power at any instant is voltage times current. When the current is out of step with the voltage, there are moments in every cycle when one is positive and the other negative, so the power is negative: the load is handing energy it stored in a magnetic field (or a capacitor's electric field) back to the supply. At a power factor of 0.80 that happens for about a fifth of every cycle. The average, which is what the kWh meter counts, is unchanged; but the energy travels down the cable and back again, and the cable heats both ways.
A capacitor's reactive power is the mathematical opposite of an inductor's — it leads instead of lags. Placed in parallel with an inductive load, it supplies the load's reactive power locally instead of drawing it from the supply, which is what shrinks the current the cable actually has to carry.
Q = 2πfCV², rearranged to C = Q / (2πfV²) — where V is the voltage
across each capacitor, not necessarily the supply voltage. Single phase uses the supply voltage
directly. Three-phase delta divides the total by three (each capacitor carries a third of the
kVAR at full line voltage). Three-phase star uses line/√3 per capacitor, which works out to
three times the capacitance of delta.
Delta, almost always. For the same correction it needs a third of the capacitance, because reactive power scales with the square of the voltage each capacitor sees. The trade is that every capacitor must be rated for full line voltage. Star banks appear where that voltage rating is the harder constraint, or where a neutral connection is wanted for unbalanced switching.
At fixed real power, current scales with 1/PF — so going from 0.75 to 0.95 cuts it by
21%. Because losses go as I²R, the heating in cables and transformers falls by 38%, and
the released kVA is headroom you can use for new load without upgrading the supply.
Yes — too much capacitance flips the phase the other way, giving a leading power factor, which some DNOs also penalise. Automatic capacitor banks switch stages on and off precisely to avoid overshooting past unity as the load changes through the day.
Not directly — UK domestic and small commercial supplies are billed on kWh alone. Reactive-power charges apply specifically to non-domestic sites with half-hourly metering, which is why this mostly matters for industrial and larger commercial premises.
Power factor = real power ÷ apparent power = kW ÷ kVA = cos φ. The reactive power is Q = √(S² − P²) = P tan φ. For example 100 kW at 0.8 power factor is 125 kVA, of which 75 kVAR does no useful work.
The capacitor size is Qc = P (tan φ1 − tan φ2), from the present angle φ1 to the target φ2. To raise 100 kW from 0.8 to 0.95: tan φ1 = 0.75 and tan φ2 = 0.329, so about 42 kVAR.
Average power factor over the period = kWh ÷ √(kWh² + kVArh²). The two meter readings are the real and reactive energy, so this is the same triangle as kW, kVAR and kVA, applied to energy.
Releases in which this page changed, newest last. Derived from the archived copy of every release, not from notes written afterwards — so it reflects what actually shipped. Site-wide passes are left out; they are in the full changelog.