Electronics Guide

The Back-to-Back Structure

Both converters are pulse-width-modulated voltage-source inverters built from insulated-gate bipolar transistors or, at higher ratings, from integrated gate-commutated thyristors. They share a common DC-link capacitor, which serves three purposes at once: it holds the DC voltage that both converters modulate against, it buffers the switching-frequency and low-order ripple each converter produces, and it provides the path through which real power moves from one converter to the other.

The series converter couples to the feeder through injection transformers, one per phase, whose secondaries sit in the line between the supply and the load. The shunt converter couples to the load bus through interfacing inductors, either directly or through a transformer where the bus voltage exceeds the converter rating. Each converter needs a passive ripple filter to keep switching harmonics out of the network, and the series stage needs a bypass path so that full load current can flow around the injection transformers when the conditioner is out of service or faulted.

The DC-link voltage must be high enough for both converters to synthesize their required outputs with modulation margin to spare. In a three-phase, three-wire system the shunt converter generally sets the requirement, since it must impress current against the full load-bus voltage. The floor follows from the modulation limit of a two-level bridge rather than from any rule of thumb: sinusoidal pulse-width modulation synthesizes a peak phase voltage of half the DC-link voltage, so the link must exceed twice the peak phase voltage, while space-vector modulation reaches a peak of the DC-link voltage divided by the square root of three and relaxes that floor by about fifteen percent. Practical designs add headroom above whichever floor applies, because a converter operating at its modulation limit has no authority left to force current during a transient. Raising the DC-link voltage improves current tracking but increases switching loss and device stress, so the choice is a genuine compromise rather than a free parameter.

Right-Shunt and Left-Shunt Configurations

The order of the two converters along the feeder matters. In the right-shunt arrangement, which dominates practice, the series converter sits nearest the supply and the shunt converter connects at the load bus downstream of it. The shunt converter therefore works against the regulated load voltage, which is clean and constant by construction, and its DC-link regulator draws its active current from that stable bus. This is the configuration assumed throughout the remainder of this article unless stated otherwise.

In the left-shunt arrangement, the shunt converter connects on the supply side and the series converter sits between it and the load. The shunt converter then faces the unregulated supply voltage, which complicates its reference generation, but it can supply the real power the series converter needs without that power passing through the load bus, and it can help precharge the DC link before the series stage is enabled. The arrangement also changes fault behavior, because a load-side fault presents differently to each converter.

Phase and Wiring Variants

Single-phase UPQCs protect individual sensitive machines and appear frequently in laboratory work and in low-power commercial products. Three-phase, three-wire conditioners serve balanced industrial loads and delta-connected distribution. Three-phase, four-wire conditioners are required wherever single-phase nonlinear loads share a neutral, because the neutral carries the triplen harmonic currents that add rather than cancel; these designs use a four-leg converter, a split DC-link capacitor with the neutral tied to the midpoint, or three independent H-bridges with isolated outputs.

At medium voltage the two-level converter gives way to multilevel structures. Cascaded H-bridge and modular multilevel converters reduce device voltage stress, lower the harmonic content of the synthesized waveform, and in some designs remove the need for injection transformers entirely. Transformerless UPQC topologies are attractive because the injection transformers are bulky, contribute loss, and introduce saturation and inrush problems, but removing them shifts the isolation and voltage-withstand burden onto the semiconductors and the DC-link insulation.

Multi-Feeder and Distributed Variants

Several extensions reuse the shared DC link across more than one point of connection. The interline UPQC, described by Amit Kumar Jindal, Arindam Ghosh, and Avinash Joshi in IEEE Transactions on Power Delivery in 2007, connects the two converters to two independent feeders instead of to one: the shunt converter regulates the bus voltage of the first feeder while the series converter regulates the voltage across a sensitive load on the second, and the shared DC link carries power between them. A healthy feeder can therefore support voltage restoration on a disturbed one. Multi-converter arrangements extend the idea to additional feeders and loads. A distributed variant, explored most prominently in the Italian "Open UPQC" work, separates the two stages geographically: the series unit sits at the distribution substation and serves many customers, while shunt units sit at individual customer premises and communicate with it. These variants trade added communication and coordination requirements for better utilization of expensive converter capacity.

Placing distributed generation or storage on the DC link converts the conditioner into a hybrid machine. A battery, supercapacitor bank, or photovoltaic array connected through a DC-DC stage allows the series converter to sustain deep sag compensation for far longer than a capacitor alone permits, and it allows the conditioner to carry the load through a complete interruption. The addition erases much of the efficiency advantage the UPQC holds over a double-conversion uninterruptible power supply, so it is justified only where interruptions as well as sags threaten the process.

Series Converter Functions

The series converter regulates the voltage presented to the load. Its primary task is compensating sags and swells, which it does by injecting the vector difference between the desired load voltage and the measured supply voltage. Because it sits in series, it also isolates the load from supply-side voltage harmonics: by synthesizing the negative of the distorting component, it presents the load with a clean sinusoid even when the upstream bus is distorted. The same mechanism corrects voltage unbalance, since negative-sequence and zero-sequence components of the supply voltage are simply further terms in the injection reference.

A second and less obvious series function is harmonic isolation in the current domain. Controlled to emulate a large resistance at harmonic frequencies while remaining transparent at the fundamental, the series converter damps the resonance that would otherwise couple load harmonics into the supply impedance. This harmonic-isolation mode was the stated purpose of the series stage in the original Fujita and Akagi formulation, where the objective was to block harmonic propagation between a subtransmission system and the distribution system beneath it, and it remains useful where a facility sits on a feeder carrying capacitor banks.

Shunt Converter Functions

The shunt converter regulates the current drawn from the supply. It injects a current equal and opposite to the distorting portion of the load current, so that the source sees only a balanced, sinusoidal, in-phase fundamental. In doing so it performs four distinct jobs: it cancels load harmonic current, it supplies the load reactive power so that the source operates near unity displacement power factor, it balances the fundamental current across phases when the load is unbalanced, and, in four-wire systems, it cancels the neutral current.

The shunt converter also carries a fifth responsibility that has no counterpart in a standalone active filter: it regulates the DC-link voltage for the entire conditioner. An outer voltage loop compares the measured DC-link voltage against its setpoint and produces the amplitude of a fundamental active current component that the shunt converter adds to its compensation reference. In steady state this current merely covers converter losses. During a disturbance it becomes the mechanism by which real power reaches the series converter, and it is what allows the pair to function without external storage.

The Circulating Real-Power Loop

The most important quantitative property of a UPQC is where the real power comes from during a voltage sag. Consider a right-shunt conditioner holding load voltage at nominal magnitude while the supply sags to a fraction of nominal. The load continues to draw its rated power, so the current the source must deliver rises in inverse proportion to the depressed supply voltage. That same current flows through the series injection transformers. The real power the series converter delivers is therefore the product of the injected voltage and the elevated source current, and it flows out of the DC link.

Because the DC-link capacitor stores very little energy, that power must be replenished immediately, and the shunt converter supplies it by drawing an extra fundamental active current from the load bus. The result is a closed loop: the source delivers additional power at the depressed voltage, the series converter injects it back into the line, the shunt converter takes an equal amount out of the load bus, and the DC link returns it to the series converter. The load itself sees nothing but nominal voltage. During a swell the loop reverses, with the series converter absorbing power and the shunt converter returning it to the supply.

The arithmetic of that loop is unforgiving, and it is exact rather than approximate. With in-phase injection, the real power handled by each converter equals the load power multiplied by the depth of the sag and divided by the fraction of nominal voltage the supply still holds. A sag to seventy percent of nominal therefore loads each converter with roughly forty-three percent of load power. A sag to fifty percent loads each converter with the full load power. This is why a UPQC rated to hold nominal voltage through very deep sags offers little economic advantage over a full-power uninterruptible supply, and why published ratings are almost always quoted for moderate sag depths.

Injection Strategies and Their Power Consequences

The choice of injection angle sets how much real power the loop must carry. Three families are in common use, and the literature labels them by the quantity the series converter exchanges.

In-phase injection, sometimes designated UPQC-P, aligns the injected voltage with the supply voltage. It minimizes the injected magnitude for a given sag depth and it restores the load voltage magnitude exactly, but every volt injected is in phase with the current, so the exchange is entirely real power. This strategy imposes the circulating loop described above in full.

Quadrature injection, designated UPQC-Q, orients the injected voltage perpendicular to the source current. The series converter then exchanges only reactive power, the DC link carries no net fundamental real power, and no energy storage is needed at all. The cost is geometric: restoring the load voltage magnitude with a perpendicular injection requires a rapidly growing injected voltage as the sag deepens, and it displaces the load voltage phase by an angle that grows just as quickly. Compensating a sag to fifty percent of nominal this way demands an injection of roughly eighty-seven percent of nominal voltage and shifts the load voltage phase by about sixty degrees. Both numbers are right-triangle geometry: the depressed supply voltage and the perpendicular injection are the legs, the restored load voltage at nominal magnitude is the hypotenuse, and the phase shift is the angle between them. Quadrature injection is consequently practical only for shallow sags and for loads indifferent to phase.

Pre-sag injection restores both the magnitude and the phase angle the load saw before the disturbance, which makes the event entirely transparent to phase-sensitive equipment such as thyristor converters and synchronization circuits. It demands the largest injection and the largest real-power exchange of the three. Minimum-energy strategies search for the injection angle that meets the voltage-magnitude target while minimizing real power, accepting whatever phase shift results, and they extend the depth a given converter and storage combination can cover.

Sharing Reactive Compensation Between Converters

During normal operation the series converter sits nearly idle, injecting only enough voltage to correct residual distortion and regulation error, while the shunt converter carries the entire load reactive and harmonic burden. That asymmetry wastes installed capacity. The UPQC-S concept, introduced by Vinod Khadkikar and Ambrish Chandra in IEEE Transactions on Power Electronics in 2011, puts the idle series capacity to work. Its mechanism is power-angle control: the series converter injects its voltage at a deliberately chosen angle so that it delivers active and reactive power at the same time, which is why the letter S, standing for complex power, labels the scheme. Part of the load reactive demand is then met by the series stage, the shunt converter has less reactive current to produce, and its rating falls accordingly. During a deep sag the control shifts the series converter back toward voltage restoration and returns the reactive duty to the shunt stage.

A related family of optimization approaches, often labeled minimum-VA design, treats the injection angle and the split of reactive duty as free variables and solves for the combination that minimizes the total volt-ampere rating of the two converters for a specified compensation envelope. The practical value of these methods is that they let a designer state the requirement honestly, as a sag depth and duration together with a harmonic and reactive target, and then size hardware against it rather than against a worst case that never occurs.

Converter Sizing in Practice

The two converters are rated on different quantities. The series converter carries full load current at all times, regardless of whether it is injecting, so its current rating is fixed by the load; its voltage rating is set by the deepest sag the installation must ride through, and the circulating-power arithmetic above bounds where that decision usefully stops, since in-phase injection through a sag to half of nominal already loads the converter with full load power. The shunt converter never carries load current, so its current rating covers only the harmonic, reactive, and unbalance components it must supply plus the active current the DC-link loop demands during sags; its voltage rating is fixed by the load-bus voltage.

Energy storage sizing follows from the injection strategy. A pure quadrature scheme needs only enough DC-link capacitance to buffer ripple. Any scheme involving real-power injection needs capacitance, or supplementary storage, sufficient to hold the DC-link voltage within its regulation band while the shunt loop responds, and enough energy to cover the event if the supply cannot deliver the extra current. Voltage sags are predominantly short events: IEEE 1159 places its instantaneous and momentary classes below three seconds, and the great majority of sags recorded in utility power quality surveys fall into those classes. Storage sized for a fraction of a second therefore addresses most of the exposure, and extending it further quickly becomes a question about interruptions rather than about sags.

Reference Generation

Both converters need a reference waveform, and generating it means separating the fundamental positive-sequence component from everything else. Instantaneous reactive power theory, the p-q formulation introduced by Akagi and colleagues, transforms three-phase voltages and currents into the stationary alpha-beta frame, computes instantaneous real and imaginary power, and separates each into a constant term corresponding to the fundamental and an oscillating term corresponding to distortion and unbalance. Compensating the oscillating terms yields the reference directly.

Synchronous reference frame control takes the alternative route of rotating the quantities into a d-q frame locked to the fundamental. In that frame the fundamental positive-sequence component appears as a constant, so a low-pass filter separates it from the harmonics with no phase error at DC. The method is intuitive and numerically robust, and its accuracy depends almost entirely on the quality of the angle supplied by the synchronization loop.

Symmetrical component extraction is a third approach, particularly suited to unbalanced supplies, in which the positive, negative, and zero-sequence components are computed explicitly and the series reference is built from the difference between the desired positive-sequence voltage and the measured sequence content.

Grid Synchronization

Every reference-generation method depends on knowing the supply phase angle, and a UPQC must track that angle through exactly the disturbed conditions that make tracking hard. A conventional synchronous reference frame phase-locked loop degrades badly under unbalance, because the negative-sequence component appears as a double-frequency oscillation on the estimated angle. Decoupled double synchronous reference frame designs and second-order generalized integrator structures separate the sequences before the loop filter and hold accurate lock through unbalanced sags. Loop bandwidth is a deliberate compromise: a fast loop follows genuine phase jumps quickly, while a slow loop rejects distortion and noise. Many designs freeze or slow the loop during a detected event so that the conditioner references the pre-disturbance angle rather than chasing the fault.

Converter Current and Voltage Control

The inner loops force each converter to produce its reference. Hysteresis current control is simple, inherently fast, and robust, but its switching frequency varies with operating point, which complicates filter design and thermal management; adaptive and space-vector hysteresis variants constrain the frequency band. Proportional-integral control in the rotating frame gives zero steady-state error at the fundamental but tracks harmonics poorly unless a separate rotating frame is dedicated to each order. Proportional-resonant control in the stationary frame provides infinite gain at selected frequencies and handles several harmonic orders with one controller structure, which has made it the common choice for active compensation.

Repetitive control exploits the fact that the disturbance repeats every fundamental period, storing one period of error and correcting it on the next; it achieves excellent steady-state harmonic rejection at the cost of a one-period delay in transient response. Deadbeat and finite-control-set model predictive controllers compute the switching state that minimizes a predicted error, giving very fast response at the cost of sensitivity to parameter error and heavy computation. Practical designs frequently combine approaches, using a fast proportional or predictive term for transient response alongside a resonant or repetitive term for steady-state accuracy.

Detection and Transition Behavior

Compensation is useful only if the disturbance is detected quickly. Peak and root-mean-square detection over a full cycle is simple but too slow for sag correction. The d-q transformation gives a magnitude estimate within a fraction of a cycle, and space-vector or instantaneous-magnitude methods react faster still. Total detection and response inside a quarter cycle is the usual design target, which is fast enough to act well before susceptible equipment reaches the limits described by the ITI (CBEMA) voltage tolerance envelope.

Transitions deserve as much design attention as steady-state performance. Entering compensation abruptly excites transformer inrush and DC-link transients; leaving it abruptly returns the load voltage to the supply in one step. Ramped entry and exit, coordinated with the phase-locked loop state, keeps both events benign. The control must also handle the case where the disturbance exceeds the conditioner capability, degrading gracefully to partial compensation rather than losing regulation entirely.

Injection Transformers

The series injection transformers determine much of the physical size, cost, and loss of a UPQC. Their turns ratio trades converter current against converter voltage: a high step-up ratio lets a low-voltage converter inject a large line voltage but forces a correspondingly larger converter current. They must tolerate full load current continuously and full fault current momentarily, and they must not saturate. Saturation is a real hazard because any DC component in the injected voltage, whether from sensor offset, modulation asymmetry, or an unbalanced transient, drives flux in one direction; control loops therefore include explicit flux estimation or DC-component rejection.

Protection and Bypass

A short circuit downstream of the conditioner drives fault current through the series transformers, and the series converter cannot supply or withstand it. A fast bypass across the series injection path is therefore mandatory, typically an anti-parallel thyristor pair for sub-cycle action backed by a mechanical contactor that closes to carry current continuously and remove the thyristor conduction loss. The bypass must also close on converter failure, on loss of control power, and on sustained overload, and it must be verified by routine testing, since a bypass that fails to close converts a conditioner fault into a load outage.

Protection coordination requires care in both directions. Downstream overcurrent devices must still see enough fault current to clear in their expected time, which the bypass permits, and the conditioner must not mask a developing fault by compensating the voltage depression that would otherwise reveal it.

Losses and Efficiency

A UPQC processes only a fraction of load power in normal operation, which is its main efficiency argument against a double-conversion uninterruptible supply. The standing losses are nonetheless real: conduction loss in the series transformer windings and any series semiconductors, switching and conduction loss in the shunt converter as it produces compensation current, magnetizing loss in the transformers, and auxiliary loads such as cooling and control. Total standing loss in the low single-digit percentages of load power is typical, and it accrues continuously whether or not a disturbance ever occurs. Any economic case must weigh those hours against the disturbances avoided.

Commissioning and Verification

Because a UPQC has many functions, verifying it means testing each independently and then together. Sag and swell response is confirmed with a sag generator that produces calibrated disturbances of specified depth, duration, and point-on-wave. Harmonic compensation is verified against measurements taken to the class A requirements of IEC 61000-4-30, comparing source-side distortion before and after enabling the shunt stage. Reactive compensation is checked by measuring displacement power factor at the point of common coupling. Bypass operation, DC-link precharge sequencing, and behavior on control-power loss all warrant explicit testing, since these are the paths by which protective equipment causes the outage it was installed to prevent.

Where the Combination Pays

The UPQC suits sites that both suffer from and cause power quality problems, where the two are coupled, and where a single point of connection serves the affected equipment. Semiconductor fabrication, precision machining, continuous plastics and paper processing, and pharmaceutical manufacturing all combine sag-sensitive controls with large populations of drives and switch-mode supplies. Medium-voltage feeders serving mixed industrial parks present a similar profile at a different scale. Interest has also grown at the interface between distribution networks and inverter-based generation, where a conditioner can regulate voltage at a weak point of connection while simultaneously filtering the harmonics that clustered converters produce.

Deployment nonetheless remains far narrower than the research literature might suggest. Standalone active filters and dynamic voltage restorers are mature commercial products with large installed bases, while integrated conditioners appear mostly in pilot installations, utility demonstration projects, and academic testbeds. The pattern was set at the beginning: Fujita and Akagi validated the concept on a twenty-kilovolt-ampere laboratory model, and the published record in the decades since has been dominated by simulation studies and bench prototypes of comparable rating rather than by field installations. Reviews of the subject name the same obstacles repeatedly, namely capital cost against a benefit realized only during disturbances, the absence of a product-level standard that would let conditioners be specified and compared the way drives and filters are, and the specialized expertise that commissioning and maintaining a coordinated two-converter control system demands. The commercial logic points the same way: a facility that needs both functions can usually buy them separately from established suppliers, retire each independently, and avoid a single control system whose failure removes both protections at once.

That imbalance between literature and practice should shape how published performance claims are read. Compensation results quoted for a UPQC, whether a distortion figure reduced from one value to another or a sag of a stated depth ridden through, almost always come from a simulation or from a bench with a known load, a chosen source impedance, and a synthesized disturbance. They are demonstrations of a control strategy, not equipment ratings, and none of them transfers to a site without restating the load, the supply, the disturbance profile, and the converter ratings that produced the result.

Comparison with Other Equipment

Against a dynamic voltage restorer alone, the UPQC adds current-side compensation and gains the ability to draw sag energy from the supply through the shunt converter rather than from storage. Against a shunt active power filter alone, it adds voltage-side protection. Against separate restorer and filter units, it saves one DC link and one enclosure but concentrates risk and complicates control. Against a double-conversion uninterruptible power supply, it offers better efficiency and lower storage cost for sag ride-through while offering nothing at all during a complete interruption unless storage is added to the DC link. Against a static transfer switch, it works where only one source exists and where the disturbance would affect both feeds anyway.

The honest comparison is often with doing less. Many sag-related trips originate in inexpensive relays, contactors, and drive control supplies, and hardening those components or adding small point-of-use ride-through modules can eliminate most of the exposure for a fraction of the cost of a facility-scale conditioner. A power quality survey that identifies which specific components trip, and at what depth and duration, is the necessary first step before any conditioner is specified.

Standards and Performance Assessment

UPQC performance is specified and judged against the same standards that govern the disturbances it addresses. IEEE 1159 supplies the classification of sags, swells, interruptions, and transients by magnitude and duration, which is the vocabulary in which a compensation envelope is stated. IEEE 519 sets the harmonic limits at the point of common coupling that the shunt stage exists to meet. IEC 61000-4-30 defines the measurement methods, and its class A requirements in particular, that make before-and-after comparisons defensible. IEC 61000-4-11 and IEC 61000-4-34 define the voltage dip and short interruption immunity tests that establish what the protected equipment can withstand unaided, the first covering equipment drawing up to sixteen amperes per phase and the second covering equipment above that. The ITI (CBEMA) curve remains the common graphical reference for judging whether a recorded event should have caused a trip: its acceptable region admits a sag to seventy percent of nominal for up to half a second and to eighty percent for up to ten seconds, and plotting a site's measured events against it is the clearest way to state what a conditioner must actually accomplish. The curve describes the tolerance of information technology equipment rather than of industrial plant generally, so it is a starting point for a specification and not a substitute for measuring what the protected equipment actually withstands.