1398
`
`IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006
`
`Overview of Control and Grid Synchronization for
`Distributed Power Generation Systems
`
`Frede Blaabjerg, Fellow, IEEE, Remus Teodorescu, Senior Member, IEEE, Marco Liserre, Member, IEEE,
`and Adrian V. Timbus, Student Member, IEEE
`
`Abstract—Renewable energy sources like wind, sun, and hydro
`are seen as a reliable alternative to the traditional energy sources
`such as oil, natural gas, or coal. Distributed power generation
`systems (DPGSs) based on renewable energy sources experience
`a large development worldwide, with Germany, Denmark, Japan,
`and USA as leaders in the development in this field. Due to the
`increasing number of DPGSs connected to the utility network,
`new and stricter standards in respect to power quality, safe run-
`ning, and islanding protection are issued. As a consequence, the
`control of distributed generation systems should be improved to
`meet the requirements for grid interconnection. This paper gives
`an overview of the structures for the DPGS based on fuel cell,
`photovoltaic, and wind turbines. In addition, control structures of
`the grid-side converter are presented, and the possibility of com-
`pensation for low-order harmonics is also discussed. Moreover,
`control strategies when running on grid faults are treated. This
`paper ends up with an overview of synchronization methods and a
`discussion about their importance in the control.
`
`Index Terms—Control strategies, distributed power generation,
`grid converter control, grid disturbances, grid synchronization.
`
`I. INTRODUCTION
`
`N OWADAYS, fossil fuel is the main energy supplier of the
`
`worldwide economy, but the recognition of it as being a
`major cause of environmental problems makes the mankind to
`look for alternative resources in power generation. Moreover,
`the day-by-day increasing demand for energy can create prob-
`lems for the power distributors, like grid instability and even
`outages. The necessity of producing more energy combined
`with the interest in clean technologies yields in an increased
`development of power distribution systems using renewable
`energy [1].
`Among the renewable energy sources, hydropower and wind
`energy have the largest utilization nowadays. In countries with
`hydropower potential, small hydro turbines are used at the
`distribution level to sustain the utility network in dispersed or
`remote locations. The wind power potential in many countries
`
`Manuscript received January 31, 2006. Abstract published on the Internet
`July 14, 2006. This work was supported in part by Risø National Laboratory,
`in part by Eltra, and in part by the Danish Research Councils under Contract
`2058-03-0003.
`F. Blaabjerg, R. Teodorescu, and A. V. Timbus are with the Institute of
`Energy Technology, Aalborg University, 9220 Aalborg, Denmark (e-mail:
`fbl@iet.aau.dk; ret@iet.aau.dk; avt@iet.aau.dk).
`M. Liserre is with the Department of Electrotechnical and Electronic Engi-
`neering, Polytechnic of Bari, 70125 Bari, Italy (e-mail: liserre@poliba.it).
`Digital Object Identifier 10.1109/TIE.2006.881997
`
`Installed capacity at the end of 2004. (a) Wind energy in Europe [2].
`Fig. 1.
`(b) PV power in the world [3].
`
`around the world has led to a large interest and fast development
`of wind turbine (WT) technology in the last decade [2]. A total
`amount of nearly 35-GW wind power has been installed in
`Europe by the end of 2004, as shown in Fig. 1(a).
`Another renewable energy technology that gains acceptance
`as a way of maintaining and improving living standards without
`harming the environment is the photovoltaic (PV) technology.
`As shown in Fig. 1(b), the number of PV installations has an
`exponential growth, mainly due to the governments and utility
`companies that support programs that focus on grid-connected
`PV systems [3], [4].
`Besides their low efficiency, the controllability of the distrib-
`uted power generation systems (DPGSs) based on both wind
`and sun are their main drawback [5]. As a consequence, their
`connection to the utility network can lead to grid instability
`or even failure, if these systems are not properly controlled.
`Moreover, the standards for interconnecting these systems to
`the utility network are stressing more and more the capability of
`the DPGS to run over short grid disturbances. In this case, both
`synchronization algorithm and current controller play a major
`role. Therefore, the control strategies applied to distributed
`systems become of high interest.
`This paper gives an overview of the main DPGS structures,
`PV and fuel cell (FC) systems being first discussed. A clas-
`sification of WT systems with regard to the use of power
`electronics follows. This is continued by a discussion of control
`structures for grid-side converter and the possibilities of imple-
`mentation in different reference frames. Further on, the main
`characteristics of control strategies under grid fault conditions
`are discussed. The overview of grid synchronization methods
`and their influence in control conclude this paper.
`
`0278-0046/$20.00 © 2006 IEEE
`
`GE 2011
`Vestas v. GE
`IPR2018-01015
`
`

`

`BLAABJERG et al.: OVERVIEW OF CONTROL AND GRID SYNCHRONIZATION FOR DPGSs
`
`1399
`
`Fig. 3. Hardware structure for a PV system using a dc–dc stage to boost the
`input voltage.
`
`III. HARDWARE TOPOLOGIES FOR DPGS
`
`A detailed description of the hardware structure for many
`types of DPGSs is given in [5]. Noticeable is that the PV and
`FC systems have a similar hardware structure, whereas different
`hardware topologies can be found for WT systems, depending
`on the type of the generator used. A brief introduction into the
`structure of these systems is given below.
`
`A. PV and FC Systems
`
`As aforementioned, the hardware structures of PV and FC
`systems are quite similar. Although both FC and PV systems
`have a low-voltage input provided by the FC and PV panels,
`more such units can be connected together to obtain the re-
`quired voltage and power. Usually, power conditioning systems,
`including inverters and dc–dc converters, are often required to
`supply normal customer load demand or send electricity into
`the grid, as shown in Fig. 3. The voltage boosting can be done
`in the dc or ac stage of the system [5]–[11]. For smoothing the
`output current, an LCL filter is normally used between these
`systems and the utility network. In addition, isolation between
`the input and output powers is required in many countries where
`such systems are installed. Again, there are two ways to achieve
`isolation, namely: 1) using the dc–dc converter and 2) using an
`isolation transformer after the dc–ac stage.
`
`B. WT Systems
`
`In this section, a classification of WT systems in those using
`and those not using power electronics as interface to the utility
`network is given. Hardware structures in each case will be
`illustrated to distinguish the systems.
`1) WT Systems Without Power Electronics: Most of these
`topologies are based on squirrel-cage induction generator
`(SCIG), which is directly connected to the grid. A soft starter
`is usually used to reduce the inrush currents during start up
`[5], [12], [13]. Moreover, a capacitor bank is necessary to
`compensate for the reactive power necessary to the machine,
`as shown in Fig. 4(a).
`2) WT Systems With Power Electronics: By adding power
`electronics units into the WT systems, the complexity of the
`system is increased. In addition, the solution becomes more
`expensive. In any case, better control of the input power and
`grid interaction is obtained. For example, maximum power for
`a large interval of wind speeds can be extracted while control
`of both active and reactive powers into the grid is achieved by
`means of power electronics.
`
`Fig. 2. General structure for distributed power system having different input
`power sources.
`
`II. DPGS STRUCTURE
`
`A general structure for distributed systems is illustrated in
`Fig. 2. The input power is transformed into electricity by means
`of a power conversion unit whose configuration is closely
`related to the input power nature. The electricity produced
`can be delivered to the local loads or to the utility network,
`depending where the generation system is connected.
`One important part of the distributed system is its control.
`The control tasks can be divided into two major parts.
`
`1) Input-side controller, with the main property to extract
`the maximum power from the input source. Naturally,
`protection of the input-side converter is also considered
`in this controller.
`2) Grid-side controller, which can have the following tasks:
`• control of active power generated to the grid;
`• control of reactive power transfer between the DPGS
`and the grid;
`• control of dc-link voltage;
`• ensure high quality of the injected power;
`• grid synchronization.
`
`The items listed above for the grid-side converter are the ba-
`sic features this converter should have. Additionally, ancillary
`services like local voltage and frequency regulation, voltage
`harmonic compensation, or active filtering might be requested
`by the grid operator.
`As previously pointed out, the power conversion unit has
`different hardware structures, which are closely related to the
`input power nature. The following section presents the revision
`of the technologies mostly used today in FC and PV systems as
`well as WT systems.
`
`

`

`1400
`
`IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006
`
`only delivered into the rotor from the grid. A speed variation of
`60% around synchronous speed may be obtained by the use of
`a power converter of 30% of nominal power [5].
`By implementing a full-scale power converter between the
`generator and the utility grid, additional technical performances
`of the WT system can be achieved, with the payback in losses
`in the power conversion stage. Normally, as shown in Fig. 4(c),
`SCIG is used in this configuration, but an advantage to elim-
`inate the gearbox can be obtained by using multipole wound-
`rotor synchronous generator or permanent-magnet synchronous
`generator, as depicted in Fig. 4(d).
`It could be noticed that for interacting with the power system,
`all the structures presented above use two-level pulsewidth-
`modulation (PWM) voltage-source inverters (VSI) because this
`is the state-of-the-art technology used today by all manufactur-
`ers of wind systems. The possibility of high switching frequen-
`cies combined with a proper control makes these converters
`suitable for grid interface in the case of distributed generation,
`which has a large contribution to the improvement of generated
`power quality.
`Yet, three-level neutral-point-clamped VSI is an option for
`high-power WT systems (5 MW) to avoid high-voltage power
`devices. Attempts of using multilevel [14] or matrix converters
`[15], [16] have been made, but the use of these technologies is
`not validated yet in the field of distributed generation.
`Therefore, the next section presents discussion on the con-
`trol structures and strategies applied to two-level VSI PWM-
`driven converters, focusing on the grid-side converter control.
`Control structures implemented in different reference frames
`are presented, and the possibility of compensating for low-order
`harmonics is discussed. Moreover, control strategies when grid
`faults occur are considered.
`
`IV. CONTROL STRUCTURES FOR GRID-CONNECTED DPGS
`
`The control strategy applied to the grid-side converter con-
`sists mainly of two cascaded loops. Usually, there is a fast
`internal current loop, which regulates the grid current, and
`an external voltage loop, which controls the dc-link voltage
`[17]–[22]. The current loop is responsible for power quality
`issues and current protection; thus, harmonic compensation and
`dynamics are the important properties of the current controller.
`The dc-link voltage controller is designed for balancing the
`power flow in the system. Usually, the design of this controller
`aims for system stability having slow dynamics.
`In some works, the control of grid-side controller is based
`on a dc-link voltage loop cascaded with an inner power loop
`instead of a current loop. In this way, the current injected into
`the utility network is indirectly controlled [23].
`Moreover, control strategies employing an outer power loop
`and an inner current loop are also reported [24].
`In the following, a division of the control strategies in respect
`to the reference frame they are implemented in is given, and the
`main properties of each structure are highlighted.
`
`A. Synchronous Reference Frame Control
`
`Synchronous reference frame control, also called dq control,
`uses a reference frame transformation module, e.g., abc → dq,
`
`Fig. 4. WT systems using power electronics. (a) Minimum electronics unit.
`(b) Partial power converter. (c) Full-scale power converter structure with
`gearbox. (d) Full-scale power converter structure without gearbox and using
`multipole synchronous generator.
`
`The usage of power electronics into WT systems can be
`further divided into two categories, namely: 1) systems using
`partial-scale power electronics units and 2) systems using full-
`scale power electronics units. A particular structure is to use an
`induction generator with a wounded rotor. An extra resistance
`controlled by power electronics is added in the rotor, which
`gives a variable speed range of 2% to 4%. The power converter
`for the rotor resistance control is for low voltage but high
`currents. In any case, this solution also needs a soft starter and
`a reactive power compensator [5].
`Additionally, another solution is to use a medium-scale
`power converter with a wounded rotor induction generator, as
`shown in Fig. 4(b). In this case, a power converter connected
`to the rotor through slip rings controls the rotor currents.
`If the generator is running supersynchronously, the electrical
`power is delivered through both the rotor and stator. If the
`generator is running subsynchronously, the electrical power is
`
`

`

`BLAABJERG et al.: OVERVIEW OF CONTROL AND GRID SYNCHRONIZATION FOR DPGSs
`
`1401
`
`Fig. 5. General structure for synchronous rotating frame control structure.
`
`Fig. 6. General structure for stationary reference frame control strategy.
`
`to transform the grid current and voltage waveforms into a ref-
`erence frame that rotates synchronously with the grid voltage.
`By means of this, the control variables become dc values; thus,
`filtering and controlling can be easier achieved [25].
`A schematic of the dq control is represented in Fig. 5. In
`this structure, the dc-link voltage is controlled in accordance to
`the necessary output power. Its output is the reference for the
`active current controller, whereas the reference for the reactive
`current is usually set to zero, if the reactive power control is not
`allowed. In the case that the reactive power has to be controlled,
`a reactive power reference must be imposed to the system.
`The dq control structure is normally associated with
`proportional–integral (PI) controllers since they have a satisfac-
`tory behavior when regulating dc variables. The matrix transfer
`function of the controller in dq coordinates can be written as
`
`(cid:2)
`
`(dq)
`PI (s) =
`
`G
`
`Kp + Ki
`0
`
`s
`
`0
`Kp + Ki
`
`s
`
`(cid:3)
`
`(1)
`
`where Kp is the proportional gain and Ki is the integral gain of
`the controller.
`Since the controlled current has to be in phase with the grid
`voltage, the phase angle used by the abc → dq transformation
`module has to be extracted from the grid voltages. As a solution,
`filtering of the grid voltages and using arctangent function
`to extract the phase angle can be a possibility [26]–[28]. In
`addition, the phase-locked loop (PLL) technique [29]–[33]
`became a state of the art in extracting the phase angle of the
`grid voltages in the case of distributed generation systems.
`
`For improving the performance of PI controller in such a
`structure as depicted in Fig. 5, cross-coupling terms and voltage
`feedforward are usually used [17], [19], [25], [34], [35]. In any
`case, with all these improvements, the compensation capability
`of the low-order harmonics in the case of PI controllers is
`very poor, standing as a major drawback when using it in grid-
`connected systems.
`
`B. Stationary Reference Frame Control
`
`Another possible way to structure the control loops is to
`use the implementation in stationary reference frame, as shown
`in Fig. 6. In this case, the grid currents are transformed into
`stationary reference frame using the abc → αβ module. Since
`the control variables are sinusoidal in this situation and due
`to the known drawback of PI controller in failing to remove
`the steady-state error when controlling sinusoidal waveforms,
`employment of other controller types is necessary. Proportional
`resonant (PR) controller [36]–[39] gained a large popularity in
`the last decade in current regulation of grid-tied systems.
`In the PR case, the controller matrix in the stationary refer-
`ence frame is given by
`
`(cid:2)
`
`(αβ)
`PR (s) =
`G
`
`Kp + Kis
`s2+ω2
`0
`
`0
`Kp + Kis
`s2+ω2
`
`(cid:3)
`
`(2)
`
`where ω is the resonance frequency of the controller, Kp is the
`proportional gain, and Ki is the integral gain of the controller.
`
`

`

`1402
`
`IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006
`
`Fig. 7. General structure for natural reference frame control strategy.
`
`Characteristic to this controller is the fact that it achieves
`a very high gain around the resonance frequency, thus being
`capable to eliminate the steady-state error between the con-
`trolled signal and its reference [38]. The width of the frequency
`band around the resonance point depends on the integral time
`constant Ki. A low Ki leads to a very narrow band, whereas a
`high Ki leads to a wider band.
`Moreover, high dynamic characteristics of PR controller have
`been reported in different works [39], [40].
`
`C. Natural Frame Control
`
`The idea of abc control is to have an individual controller
`for each grid current; however, the different ways to connect
`the three-phase systems, i.e., delta, star with or without isolated
`neutral, etc., is an issue to be considered when designing
`the controller. In the situation of isolated neutral systems, the
`phases interact to one another; hence, only two controllers are
`necessary since the third current is given by the Kirchhoff
`current law. In any case, the possibility of having three inde-
`pendent controller is possible by having extra considerations in
`the controller design as usually is the case for hysteresis and
`dead-beat control.
`Normally, abc control is a structure where nonlinear con-
`trollers like hysteresis or dead beat are preferred due to their
`high dynamics. It is well known that the performance of these
`controllers is proportional to the sampling frequency; hence,
`the rapid development of digital systems such as digital signal
`
`processors or field-programmable gate array is an advantage for
`such an implementation.
`A possible implementation of abc control is depicted in
`Fig. 7, where the output of dc-link voltage controller sets the
`active current reference. Using the phase angle of the grid
`voltages provided by a PLL system, the three current references
`are created. Each of them is compared with the corresponding
`measured current, and the error goes into the controller. If
`hysteresis or dead-beat controllers are employed in the current
`loop, the modulator is not anymore necessary. The output of
`these controllers is the switching states for the switches in the
`power converter. In the case that three PI or PR controllers are
`used, the modulator is necessary to create the duty cycles for
`the PWM pattern.
`1) PI Controller: PI controller is widely used in conjunction
`with dq control, but its implementation in abc frame is also
`possible as described in [35]. The transfer function of the
`controller in this case becomes (3), shown at the bottom of the
`page, and the complexity of the controller matrix in this case,
`due to the significant off-diagonal terms representing the cross
`coupling between the phases, is noticeable.
`2) PR Controller: The implementation of PR controller in
`abc is straightforward since the controller is already in station-
`ary frame and implementation of three controllers is possible
`as illustrated in (4), shown at the bottom of the page. Again,
`in this case, the influence of the isolated neutral in the control
`has to be accounted; hence, the third controller is not neces-
`sary in (4). However, it is worth noticing that the complexity
`
`⎡⎢⎢⎢⎣
`
`·
`
`2 3
`
`⎡⎢⎣
`
`(abc)
`G
`PI
`
`(s) =
`
`(abc)
`PR (s) =
`G
`
`(3)
`
`(4)
`
`⎤⎥⎥⎥⎦
`
`2 − Kis−√
`− Kp
`3Kiω0
`2·(s2+ω2
`0)
`√
`
`− Kp2 − Kis+
`3Kiω0
`2·(s2+ω2
`0)
`Kp + Kis
`s2+ω2
`0
`
`√
`− Kp2 − Kis+
`
`3Kiω0
`2·(s2+ω2
`0)
`Kp + Kis
`s2+ω2
`2 − Kis−√
`0
`− Kp
`3Kiω0
`2·(s2+ω2
`0)
`0
`0
`Kp + Kis
`s2+ω2
`0
`
`⎤⎥⎦
`
`Kp + Kis
`s2+ω2
`2 − Kis−√
`0
`− Kp
`3Kiω0
`2·(s2+ω2
`0)
`√
`2 − Kis+
`− Kp
`3Kiω0
`2·(s2+ω2
`0)
`Kp + Kis
`0
`s2+ω2
`0
`0
`Kp + Kis
`s2+ω2
`0
`0
`0
`
`

`

`BLAABJERG et al.: OVERVIEW OF CONTROL AND GRID SYNCHRONIZATION FOR DPGSs
`
`1403
`
`TABLE I
`DISTORTION LIMITS FOR DISTRIBUTED GENERATION SYSTEMS
`SET BY IEC STANDARD [50]
`
`In addition, in the case of abc control, two modalities of
`implementing the PLL are possible. The first possibility is to
`use three single-phase PLL systems [33]; thus, the three phase
`angles are independently extracted from the grid voltages. In
`this case, the transformation module dq → abc is not anymore
`necessary, with the active current reference being multiplied
`with the sine of the phase angles. The second possibility is to
`use one three-phase PLL [31], [32], [48], [49]. In this case, the
`current references are created, as shown in Fig. 7. A discussion
`about the influence of the PLL in the control loop is given in
`Section VII.
`
`D. Evaluation of Control Structures
`
`The necessity of voltage feedforward and cross-coupling
`terms is the major drawbacks of the control structure imple-
`mented in synchronous reference frame. Moreover, the phase
`angle of the grid voltage is a must in this implementation. In the
`case of control structure implemented in stationary reference
`frame, if PR controllers are used for current regulation, the
`complexity of the control becomes lower compared to the
`structure implemented in dq frame. Additionally, the phase
`angle information is not a necessity, and filtered grid voltages
`can be used as template for the reference current waveform.
`In the case of control structure implemented in natural frame,
`the complexity of the control can be high if an adaptive band
`hysteresis controller is used for current regulation. A simpler
`control scheme can be achieved by implementing a dead-beat
`controller instead. Again, as in the case of stationary frame
`control, the phase angle information is not a must. Noticeable
`for this control structure is the fact that independent control
`of each phase can be achieved if grid voltages or three single-
`phase PLLs are used to generate the current reference.
`
`V. POWER QUALITY CONSIDERATIONS
`
`One of the demands present in all standards with regard
`to grid-tied systems is the quality of the distributed power.
`According to the standards in this field [13], [50]–[53], the
`injected current in the grid should not have a total harmonic
`distortion larger than 5%. A detailed image of the harmonic
`distortion with regard to each harmonic is given in Table I.
`As it was mentioned previously, one of the responsibilities of
`the current controller is the power quality issue. Therefore, dif-
`ferent methods to compensate for the grid harmonics to obtain
`an improved power quality are addressed in the following.
`
`Fig. 8. Structure of the dead-beat controller using an observer to compensate
`for the delay introduced by the controller.
`
`of the controller in this case is considerably reduced com-
`pared to (3).
`3) Hysteresis Controller: It is worth noticing that in the
`case of hysteresis control implementation, an adaptive band
`of the controller has to be designed to obtain fixed switching
`frequency. In [41]–[44], different methods and algorithms to
`obtain fixed switching frequency are presented.
`Since the output of the hysteresis controller is the state of
`the switches, considerations about the isolated neutral are again
`(cid:4)
`term is introduced in the formula of
`necessary. In [43], an a
`the hysteresis band (HB) to account for the load (transformer)
`connection type, i.e.,
`(cid:4)
`0.25a
`Udc
`fswLT
`
`(cid:11)
`
`(cid:10)
`1 − L2
`(cid:2)2Udca
`
`
`T
`
`∗
`+ di
`dt
`
`Ug
`LT
`
`(cid:13)
`
`(cid:12)2
`
`.
`
`(5)
`
`HB =
`
`In [45], a similar approach is used, but the current error is split
`into its noninteracting part ζ and the interacting part γ to resolve
`the equation for the variable HB.
`4) Dead-Beat Controller: The dead-beat controller attempts
`to null the error with one sample delay. The controller in its
`digital implementation is as follows:
`
`(6)
`
`(7)
`
`(cid:12)
`
`.
`
`· 1 − az
`−1
`1 − z−1
`where a and b are denoted as follows:
`− RT
`a = e
`LT
`b = − 1
`RT
`
`T s
`
`(cid:11)
`
`− RT
`e
`LT
`
`T s − 1
`
`1 b
`
`(abc)
`DB =
`G
`
`Since dead-beat controller regulates the current such that it
`reaches its reference at the end of the next switching period,
`the controller introduces one sample time delay. To compensate
`for this delay, an observer can be introduced in the structure of
`the controller, with the aim to modify the current reference to
`compensate for the delay [46], as shown in Fig. 8.
`The discrete transfer function of the observer is given by
`1
`1 − z−1
`thus, the new current reference becomes
`∗ − i).
`∗(cid:2) = F
`(abc)
`DB (i
`i
`
`(8)
`
`(9)
`
`(abc)
`DB =
`
`F
`
`As a consequence, a very fast controller containing no delay
`is finally obtained. Moreover, the algorithms of the dead-beat
`controller and observer are not complicated, which is suitable
`for microprocessor-based implementation [47].
`
`A. Harmonics Compensation Using PI Controllers
`
`Since PI controllers typically are associated with dq control
`structure, the possibilities for harmonic compensation are based
`
`

`

`1404
`
`IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006
`
`Fig. 9. Method for compensating the positive sequence of the fifth harmonic in dq control structure.
`
`on low-pass and high-pass filters [54]. If the current controller
`has to be immune to the grid voltage harmonic distortion
`(mainly fifth and seventh in three-phase systems), harmonic
`compensator for each harmonic order should be designed.
`Fig. 9 shows the dq control structure having a harmonic com-
`pensator for the positive sequence of the fifth harmonic. In
`addition, under unbalanced conditions, harmonic compensators
`for both positive and negative sequences of each harmonic order
`are necessary. As a consequence, four compensators like the
`ones depicted in Fig. 9 are necessary to compensate for the
`fifth and seventh harmonics. The complexity of the control
`algorithm is noticeable in this case.
`
`B. Harmonics Compensation Using PR Controllers
`
`In the case of PR control implementation, things are dif-
`ferent. Harmonic compensation can be achieved by cascading
`several generalized integrators tuned to resonate at the desired
`frequency. In this way, selective harmonic compensation at dif-
`ferent frequencies is obtained. In [38], the transfer function of
`a typical harmonic compensator (HC) designed to compensate
`the third, fifth, and seventh harmonics is given as follows:
`
`Kih
`
`s
`s2 + (ω · h)2 .
`
`(10)
`
`(cid:14)
`
`h=3,5,7
`
`Gh(s) =
`
`In this case, it is easy to extend the capabilities of the
`scheme by adding harmonic compensation features simply with
`more resonant controllers in parallel to the main controller, as
`illustrated in Fig. 10. The main advantage in this situation is
`given by the fact that the harmonic compensator works on both
`positive and negative sequences of the selected harmonic; thus,
`only one HC is necessary for a harmonic order.
`An interesting feature of the HC is that it does not affect the
`dynamics of the PR controller, as it only reacts to the frequen-
`cies very close to the resonance frequency. This characteristic
`
`Fig. 10. Structure of the harmonic compensator attached to the resonant
`controller of the fundamental current.
`
`makes the PR controller a successful solution in applications
`where high dynamics and harmonics compensation, especially
`low-order harmonics, are required, as in the case of a DPGS.
`
`C. Harmonics Compensation Using Nonlinear Controllers
`
`Since both hysteresis and dead-beat controller have very fast
`dynamics, there is no concern about the low-order harmonics
`when the implemented control structure uses such controllers.
`In any case, it should be noticed that the current waveform will
`contain harmonics at switching and sampling frequencies order.
`Another issue is the necessity of fast sampling capabilities of
`the hardware used.
`
`D. Evaluation of Harmonic Compensators
`
`The necessity of using two filters, two transformation mod-
`ules, and one controller to compensate for the positive sequence
`of only one harmonic makes the harmonic compensator imple-
`mented in dq frame to be not a practical solution. On the other
`
`

`

`BLAABJERG et al.: OVERVIEW OF CONTROL AND GRID SYNCHRONIZATION FOR DPGSs
`
`1405
`
`capabilities for a DPGS, the influence of the unbalance should
`be minimized when running under faulty conditions.
`With regard to the control strategy under faults, four major
`possibilities are available.
`
`A. Unity Power Factor Control Strategy
`
`Fig. 11. Distributed generation system connected through a Δ/Y transformer
`to the utility network.
`
`hand, easier implementation is observed in the situation when
`the control structure is implemented in stationary reference
`frame since the structure of the compensator is reduced and it
`acts on both positive and negative sequences.
`
`VI. CONTROL STRATEGY UNDER GRID FAULTS
`
`Due to the large amount of distributed power generation con-
`nected to the utility networks in some countries, instability of
`the power system may arise. As a consequence, more stringent
`demands for interconnecting the DPGS to the grid are issued.
`Among all the requests, more and more stress is put on the
`ability of a DPGS to ride through short grid disturbances such
`as voltage and frequency variations.
`The grid faults can be classified in two main categories [55].
`1) Symmetrical fault is when all three grid voltages register
`the same amplitude drop but the system remains balanced
`(no phase shifting is registered). This type of fault is very
`seldom in the power systems.
`2) Unsymmetrical fault is when the phases register an un-
`equal amplitude drop together with phase shifting be-
`tween them. This type of fault occurs due to one or two
`phases shorted to ground or to each other.
`By considering the DPGS connected to the utility network
`as shown in Fig. 11, where a distribution transformer is used
`by the generation system to interface the power system, the
`propagation of a voltage fault occurring at bus 1 appears
`different at bus 2. For example, if a severe grid fault like
`single phase shorted to ground takes place at bus 1, two of the
`voltages at the DPGS terminals (after the Δ/Y transformer)
`experience a voltage drop that is dependent on the impedance
`of the line between the fault and DPGS transformer value. As a
`consequence, the voltages at bus 2 will register both amplitude
`and phase unbalance [55].
`the negative
`Since this case is an unsymmetrical fault,
`sequence appears in the grid voltages. This creates second-
`harmonic oscillations that propagate in the system, which ap-
`pear in the dc-link voltage as a ripple [56]. Moreover, the
`control variables are also affected by this phenomenon. In
`[57]–[59], it has been shown that the PLL system can be
`designed to filter out the negative sequence, which produces
`a clean synchronization signal. If the three-phase PLL system
`is not designed to be robust to unbalanced, second-harmonic
`oscillations will appear in the phase angle signal, thus in the
`current reference.
`In addition, the second-harmonic ripple present in the dc-link
`voltage will also have a negative influence in generation of the
`current reference. As a consequence, to provide ride-through
`
`(11)
`
`i = gv,
`
`One of the control strategies that a DPGS can adopt on grid
`faults is to maintain unity power factor during the fault. The
`most efficient set of currents delivering the instantaneous active
`power P to the grid can be calculated as follows:
`g = P|v|2
`where g is the instantaneous conductance seen from the inverter
`output, and |v| denotes the module of the three-phase voltage
`vector v. Its value is constant in balanced sinusoidal condi-
`tions, but under grid faults, the negative sequence component
`gives rise to oscillations at twice the fundamental frequency
`in |v|. Consequently, the injected currents will not keep their
`sinusoidal waveform, and high-order components will appear
`in their waveform. Current vector of (11) is instantaneously
`proportional to the voltage vector and, therefore, does not have
`any orthogonal component in relation to the grid voltage, hence
`giving rise to the injection of no reactive power to the gr