Reactive power compensation hardware is constrained by temperature limits in four layered ways: ambient at the cubicle inlet, hotspot inside the dielectric, thyristor junction rise, and busbar joint temperature. Each gate rules out a different topology before a single kVAR is specified [S3][S5].
The 2026 reactive-power planning literature keeps voltage stability and loss reduction as the headline goals, but a 2026 paper on optimised compensation with renewable integration now bounds inverter actions by apparent power S and reactive losses by line thermal limits, with bus voltage kept inside 0.95 to 1.05 p.u. on every bus [S4]. That same thermal envelope governs the static capacitor bank sitting next to the inverter.
Why temperature gates compensation sizing
Power factor is the ratio P/S, where S is the vector sum S = sqrt(P² + Q²) and Q is the reactive power oscillating at 90° to real power; a load at PF 0.7 draws roughly 43% more line current than the same kW at PF 1.0, and that extra I²R heating shows up first at the capacitor terminals, not the transformer [S3].
Reactive power itself is not consumed, but the current that carries it is, and current heats every joint, cable, and dielectric it passes through [S5]. In a typical indoor LV switchroom rated 40°C, a self-healing metallised polypropylene capacitor bank is usually specified for an ambient of –25°C to +45°C with a continuous current limit of 1.3×In (2025-08). When the cubicle sits in a non-air-conditioned feeder pillar above 45°C, kVAR must be derated or the dielectric life halves every 7–10°C above rated, the rule of thumb you will hear from capacitor OEMs and from IEC 60831-1/2 design guidance (2025-08).
Skip any one of these and a bank that passed the spreadsheet fails within a season.
Compensation topologies and their thermal behaviour
Three families dominate 2026 specifications: passive capacitor banks, thyristor-switched capacitor (TSC) + thyristor-controlled reactor (TCR) Static VAR Compensators (SVC), and voltage-source-converter STATCOMs. The GitHub simulation record from Prem Manoj Mule spells out the split: SVC adjusts reactive power via thyristor-controlled reactors and thyristor-switched capacitors and is cheaper but slower, while STATCOM is a VSC-based shunt compensator that generates controllable reactive current independent of system voltage and holds up better in low-voltage conditions [S1].
Comparison of the three on the criteria that decide a 2026 RFQ:
- Passive capacitor bank: lowest cost per kVAR, fastest to install, but no continuous control; thermal limit is purely ambient-driven; switching transients force a 200–300 s discharge reactor; best when load is steady and PF target is 0.95–1.0.
- TSC + TCR SVC: continuous control, fast step (~1/2 cycle), harmonic filtering possible with detuned branches; thermal limit split between capacitor dielectric (T_class D, 55°C hotspot) and thyristor heatsink (junction 125°C max, heatsink sized for 40°C ambient with 70% load); mid-cost; best on fluctuating industrial loads with PF penalty exposure.
- VSC STATCOM: independent of AC voltage for reactive output, sub-cycle response, inherent harmonic filtering; thermal limit is the IGBT module and the DC-link capacitor, both rated for 105–125°C hotspot; higher cost and higher losses (typically 0.5–0.8% of rated MVA); best on weak grids, renewables interconnection, or where voltage must stay inside the 0.95–1.05 p.u. band cited in the 2026 compensation literature [S4].
A reactive-power planning paper published in 2026 in Scientific Reports formalises the weak-bus identification step that often drives this choice: it ranks candidate buses by a composite voltage stability index, sizes the compensator against the minimum stability margin, and confirms that reactive resources should be located at the weak buses first rather than distributed evenly [S2].
Limit switches, monitors, and the control loop around a compensation panel
Reactive compensation hardware does not sit in isolation. The cubicle door carries a limit switch for the isolator, the position feed is mirrored to a limit switch box wired to the SCADA, and a temperature controller on the heatsink or cubicle interior trips the bank on overtemperature before the dielectric or the IGBT module is damaged. [S3]
The temperature monitor chain is what ties compensation back to the reactive power compensation master: a PT100 or thermocouple on the hottest phase, setpoints at the dielectric class temperature minus 10 K (alarm) and minus 5 K (trip), and a temperature measurement loop that the controller uses to block the next switch-in until the bank has cooled. Skipping this chain is the single most common reason a capacitor bank fails its second summer.
Spec gates that decide fit on a 2026 RFQ
Decision checklist, in the order an evaluator should walk it:
1. PF target and penalty clause: many contracts cap reactive consumption at 50% of active consumption before penalties apply, so the kVAR target is set by the bill, not by engineering preference [S3].
2. Voltage band: 0.95–1.05 p.u. at the point of common coupling is the 2026 planning norm for transmission-connected sites [S4]; below 0.95 p.u. the STATCOM topology earns its premium.
3. Thermal envelope: ambient, altitude, harmonic spectrum, duty cycle. Each topology above has a different failure mode if any of these is wrong.
4. Switching duty: more than 20–30 operations per hour pushes passive banks toward TSC, and TSC toward STATCOM, because each hard-switch event re-stresses the contactor or thyristor and the inrush reactor.
5. Standards: IEC 60831-1/2 for LV capacitors, IEC 61921 for power factor correction banks, IEEE 519 or IEC 61000-2-2/2-4 for harmonics, and the project-specific ATEX or IECEx zone if the panel is indoors in a classified area (2025-08).
Limits, failure modes, and what the datasheet will not tell you
Capacitor banks are robust until they are not. The three failure modes seen most often on operating sites are (a) dielectric dry-out from sustained over-temperature, (b) contact welding from inrush on a partly discharged bank, and (c) resonance with the supply inductance when a detuned reactor is omitted on a site with significant harmonic voltage (THDu above 5%). The datasheet will not flag (c) because the resonance point depends on the actual short-circuit MVA at the busbar, which the OEM does not know. [S3]
Reactive power itself, as the 2026 Switchcraft guide underlines, is not a free or useless quantity: it is the energy stored and released each cycle in magnetic and electric fields, and it is the mechanism by which voltage is sustained across a transmission line [S5]. Compensation is the engineering of that storage; the temperature limit is the engineering of where and how fast the storage device can sit and still survive the next 100 000 hours.
Trackable signal: watch for the 2026 revision of IEC 60831 family publications and for grid-code updates that tighten the 0.95–1.05 p.u. band on transmission-connected renewable plants, since both will move the cost optimum between passive, SVC, and STATCOM topologies on the next planning cycle. The Industrial Procurement Signal write-up on [US grid build-out and Senegal gas-to-power]((/news/industrial-procurement-signal-power-transmission-20260811-617c66eb.html)) and the [AI data-centre grid-demand forecast]((/news/ai-data-centers-push-2026-2030-grid-demand-forecast-into-a-new-gear.html)) are the two upstream signals that will set the reactive-power headroom inside which every 2027 RFQ will be written.
This topic is covered further in Warehouse Safety Glove Selection: EN 388, EN 511 and Task-Specific Spec Gates.