A novel single phase PhaseLocked Loop (PLL) is proposed in this paper to accurately and rapidly estimate the instantaneous phase angle of a grid. A conjugate rotating vector pair is proposed and defined to synthesize the single phase signal in the stationary reference frame. With this concept, the proposed PLL innovatively sets one phase input of the PARK transformation to a constant zero. By means of a proper cancellation, a zero steady state phase angle estimation error can be achieved, even under magnitude and frequency variations. The proposed PLL structure is presented together with guidelines for parameters adjustment. The performance of the proposed PLL is verified by comprehensive experiments. Satisfactory phase angle estimation can be achieved within one input signal cycle, and the estimation error can be totally eliminated in four input cycles for the most severe conditions.
I. INTRODUCTION
Due to the popularity and wide acceptance of distributed generation systems, the last decade has become an explosive period for gridconnected power electronics converter research
[1]
. A variety of converter topologies have been proposed and wellstudied. Later, different control methods were proposed
[2]
,
[3]
. In addition, a variety of applications for three phase gridconnected power electronics converters and single phase gridconnected power electronics converters have become standard configuration in many applications, such as photovoltaics, UPSs, energy storage technologies, aerospace applications and FACTS
[4]

[9]
. In such systems, the control performance is greatly dependent on accurate grid information
[5]
,
[7]
,
[9]
. Both phase and magnitude have been used to generate references for the converters output, for example, the current command. Phase information has also been utilized to transform particular state variables into the synchronous reference frame
[10]
,
[11]
. Islanding detection can be achieved by monitoring the grid frequency
[12]
,
[13]
. Therefore, the grid synchronization method is a key method for the control of gridconnected power converters.
A lot of work has already been done on this topic. Zerocrossing point detection suffers from low dynamic transients and sensitivity to jitters at around the zerocrossing point. The Discrete Fourier Transformation (DFT) requires more computational resources, and it also features a onecycle detection delay
[14]
. Among all the exiting synchronization methods, the PhaseLocked Loop is the most widely accepted method.
Different types of single phase PLLs have been proposed in the literature. They are usually constructed in terms of structure by the Phase Detector (PD), Loop Filter (LF) and Voltage Controlled Oscillator (VCO). Most creative work on PLL research focused on new PDs. The original PLL used a mixer as the PD
[15]
. However, multiplication generates an undesired ripple simultaneously at twice the input frequency under the phase locked state. This ripple contaminates the phase angle estimation, and further influences the output of the gridconnected converter. Therefore, either a Low Pass Filter (LPF) or a reduced bandwidth LF has to be implemented to guarantee the phase angle estimation accuracy
[8]
,
[16]
. A Modified Mixer PD is proposed to reduce such a ripple, but accurate phase angle estimation can only be achieved within a limited input magnitude range around unity value
[17]
.
In three phase systems, synchronous rotating transformation is widely used in control and phase angle estimations
[18]

[20]
. These PLLs are referred to as Synchronous Reference Frame PLL (SRFPLL). This transformation is also adopted in single phase gridconnected converter control and PLLs. Many single phase SRFPLLs have been proposed with a synchronous rotating transformation PD, referred to as the PARK PD
[18]
. These PLLs generate a clean phase angle estimation during the phaselocked state. However, the sole source of information sensed at the input makes the PARK PD unable to be implemented directly in single phase PLLs. Almost all the proposed methods focused on the generation of the required signal, which is orthogonal to the input. This process is also called Quadrature Signal Generation (QSG). Transport delay
[21]
, differentiation
[11]
and certain kinds of filters
[22]

[25]
can be used for QSG. However, the transport delay QSG introduced an inevitable undesired delay. The differentiation QSG amplifies the noise of the input. Filtering the input can also generate an orthogonal signal without delay, and additional attenuation of the undesired harmonics can be also achieved. However, such QSG methods are not frequencyadaptive. In order to be frequencyadaptive, an additional frequency loop is required.
A single phase SRFPLL which does not equip QSG has also been proposed. This type of SRFPLL sets the orthogonal signal as zero constantly
[14]
,
[26]
. Although it is declared in
[14]
that it sets a constant zero orthogonal signal, the cascaded delay signal cancellation operators in the stationary reference frame generates two signals in orthogonal for the PARK PD.
It can be concluded from all of the single phase PLLs in the literature that, the single phase PLL with a mixer PD is inherently frequencyadaptive. However, accurate phase angle estimation is hard to obtain. Accurate phase angle estimation can be achieved by single phase SRFPLLs, but frequencyadaptive phase angle estimation requires an additional frequencylocked loop. The removal of QSG from a single phase SRFPLL is novel but not fully achieved in the literature. In this paper, a novel single phase PLL is proposed to achieve both accurate phase angle estimation and inherent frequencyadaption without QSG.
This paper is organized as follows. In section II, the conjugate rotating vector pair is defined and described to synthesize a single phase signal. In section III, a novel single phase SRFPLL is proposed and presented. The design issues in the proposed PLL are discussed in detail with both mathematical analysis and numeric simulation. In section IV, a comprehensive experiment verified the static and dynamic performances of the proposed PLL. The paper is concluded in Section V.
II. SYNTHESIS OF A SINGLE PHASE SIGNAL
Fig. 1
shows a commonly used power converter along with the related control systems for singlephase gridconnected applications. The PLL is used to estimate the line voltage phase angle, or the voltage angle at the point of common connection. It has to be mentioned that the PLL estimates the line current phase angle in series connected converters, such as the distributed series compensator. Therefore, the PLL estimates either the voltage or current phase angle, depending on the application. In this paper, the variables are represented by
u
for example.
Common single phase distributed generation system.
In the wellknown stationary reference frame, a pulsing vector can be generated by setting the value of the αaxis as the single phase signal, and by setting the value of the βaxis as zero constantly. In this case, the single phase signal is expressed in (1). The end of the pulsing vector is on the αaxis, and it moves back and forth within the range of
α
∈ [
A
,
A
] . The vector can be represented mathematically in (3).
Where:
It is straight forward that such a pulsing vector can be synthesized by two rotating vectors, which are called rotating vector pairs. They are referred to as the positive rotating vector
and the negative rotating vector
. If these two vectors are represented as (4) and (5), the relations within (3), (4) and (5) can be easily obtained as follows:
A
^{+}
=
A
^{}
= 0.5·
A
,
θ
^{+}
(
t
) =
θ
(
t
) , and
θ
^{}
(
t
) = 
θ
(
t
) . This means that the magnitude of the rotating vector pair is the same as half of the input signal magnitude, and the phase angle with respect to the αaxis is opposite. This rotating vector pair is defined as cconjugate rrotating vector pair. A vector synthesis diagram is shown in
Fig. 2
. The synthesized input vector is also plotted.
Vector synthesis with Conjugate Rotating Vector Pair and time domain representation.
It should be noted that the positive rotating vector carries the instantaneous phase angle information of the single phase signal. The phase angle estimation of the single phase signal can be achieved by suitably obtaining the angle between the positive rotating vector and the αaxis.
In addition, as shown in (2), the frequency is not limited at a unique constant. All of the frequency components in
u_{i}
(
t
) can be synthesized by a series of rotating vector pairs. It should be mentioned that the DC component in
u_{i}
(
t
) cannot be synthesized by the sum of a rotating vector pair. It can be synthesized by the static vector pair shown in (7) and (8), while the DC component in
u_{i}
(
t
) is represented in (6). In this way, any single phase signal can be synthesized by a series of conjugate rotating vector pairs.
III. THE PROPOSED PLL
 A. Description
In order to simplify the following analysis, the input single phase signal is redefined in (9). Harmonics are not considered for the sake of simplicity. By using the PARK transformation rotating at the estimated phase angle
θ_{e}
(
t
), the daxis and qaxis output can be found in (10), after setting
u_{i}
as α, and maintaining the β at zero. In this case, the PARK transformation rotating at
θ_{e}
(
t
) is defined as (11).
It is noted from (10) that the first term of the qaxis output carries the phase angle error between the input signal and the phase angle estimation. It is supposed that the error is well controlled by a certain feedback mechanism and that the error goes to zero. This means that
θ_{e}
(
t
) ≈
θ
_{1}
(
t
). Under this phaselocked state, the daxis and qaxis outputs are shown in (12).
Where,
ε
is a small value, i.e. nearly zero, and it is expressed in (13).
It can be stated from (12) that under phase lock, the daxis output consists of a DC component and a second order AC component. The DC component is nearly half of the input magnitude, and the magnitude of the second order AC component is also half of the input. Corresponding to the cconjugate rotating vector pair, it is true that the projection of one vector became a DC quantity in the SRF synchronous to it, and the other is kept rotating in that SRF at twice the angular frequency. The module of the vectors is unchanged and they are still equal to each other in that SRF. By making use of this characteristic, the AC components in (12) can be expressed as (14). In other words, the AC components can be obtained by a proper transformation of the DC components in (12). The 2x2 matrix in (15) is just the PARK transformation matrix with a rotating phase angle of 2·
θ_{e}
(
t
) .
In this way, (10) can be rewritten as (15). (15) indicates that the second order AC component in (10) can be cancelled by a certain feedback of the PARK transformation of the DC component in (10) at twice the estimated phase angle.
The derived PLL is presented in
Fig. 3
. The outputs of the proposed PLL are the phase angle estimation, the frequency estimation and the magnitude estimation. The frequency estimation is a useful state variable which can be a measure of the PLL dynamics.
Structure of the proposed PLL.
The PD can be seen inside the dashed box in
Fig. 3
. It contains 2 PRAK transformation blocks (i.e. PARK1 and PARK2), 2 identical low pass filters, 2 adders and 2 gains. The single phase input is set as the αinput of PARK1, and the βinput of PARK1 is kept at a constant zero. The estimated phase angle is fed back to PARK1 to accomplish the rotating transformation. The outputs of PARK1 are then added to the outputs of PARK2. The sums (i.e.
u_{d}
and
u_{q}
) are filtered by LPF1 and LPF2 to obtain their DC components, as shown in (10). The filtered
u_{d}
and
u_{q}
are then served as the input of PARK2, which is rotating at 2·
θ_{e}
(
t
) . These procedures accomplished the extraction of the DC components in (10) by means of a kind of feedback instead of a simple filtering.
The rest of the parts of the proposed PLL are same as those of a conventional PLL. The output of PD
u_{q}
is fed to the PItype LP to obtain an accurate phase angle estimation. In addition, the filtered
u_{d}
, which carries the magnitude information of the input, is used to estimate the magnitude of the input with a gain of 2.
There are only two blocks to be designed in the proposed PLL. The design issues will be discussed in the following parts. It has to be mentioned that there are two LPFs in the proposed PLL. However, they are not responsible to the attenuation of the characteristic ripple created by the negative rotating vector. Consequently, simple firstorder LPFs are well suited to this purpose.
 B. Phase Detector Evaluation
It is well known that most PLLs differ from the PD, and the proposed PLL is not different. Hence, the dynamic and static behavior of the PD is very important for the PLL performance. The PD of the proposed PLL is indicated by the dashed box in
Fig. 3
. It is obvious that the PD behavior depends largely on the LPFs. Since a firstorder LPF can serve this purpose satisfactorily, it can be said that the cutoff frequency is dominant in the PD performance.
In order to simplify the analysis, PARK1 is constantly rotating at
ω
_{1}
rad/s. In addition, the cutoff frequency
ω_{c}
is selected so that is it proportional to
ω
_{1}
with the coefficient k. The transfer function of the LPF is expressed in (16), using the variables of LPF1 for example.
Where:
Because two PARK transformations are included in the PD, it is hard to derive the transfer function in the sdomain. Therefore, the LPF is expressed in the time domain as (18).
Based on (18), the relations of LPF1 and LPF2 can be derived and written in matrix (19).
The computation of the two sums in
Fig. 3
is written in (20), and PARK2’s output is expressed in (21).
In addition,
u
_{1d}
and
u
_{1q}
are the PARK transformation outputs of the single phase input. For the sake of simplicity, its phase angle is assumed to be
θ
, and the PARK1 outputs are expressed in (22).
From (19)(22), the following differential equations model can be written:
The differential equation model in (23) states that the PD is a linear time variant system. Since the analytical solution is highly complex and hard to obtained, a computeraided method is proposed to evaluate the performance of the PD. This method is efficient and fast in guiding the selection of
ω_{c}
and is expressed in
Fig. 4
. The cosine function in the bottom left generates the cosine unity input to the PD, and the phase angle error to each PARK block is zero. As a result, when a step input occurs in the PD, the PD output
u_{df}
converges to 0.5 in the steady state since the input is a unity input, while the PD output
u_{qf}
goes back to zero. This method is implemented in MATLAB/SIMULINK,
u_{q}
with different values of the coefficient
k
shown in
Fig. 5
(a) and
Fig. 5
(b), respectively.
Performance evaluation method for Phase Detector.
u_{q} for different k with unity input (a: k<1; b: k>1).
It can be observed from
Fig.5
that an LPF with a low
ω_{c}
attenuates the ripple well. However, it can also be seen that the convergence takes a longer time. When
k
is lower than 0.3, it takes more than 2 input signal cycles. A high
k
, which is over 1, brings
u_{q}
into the extended steady state in almost half an input signal cycle. However, the ripple is underdamped, which causes periodic errors. By selecting a value for
k
that is between 0.5~1, a satisfactory result can be obtained in
u_{q}
. Considering the total settling time,
k
is selected to be 0.707. In the following tests of the paper,
k
is equal to 0.707 in both the simulation and the experiment.
 C. Loop Filter Design
The LF is commonly designed with the PLL’s linear small signal model, as shown in
Fig. 6
. Most studies select a PI controller as the LF in order to track the ramped phase angle without errors. The VCO is model as an integrator. Usually, only the static gain of the PD
G_{PD}
is considered in the LF design procedure.
Smallsignal model of PLL.
Based on the smallsignal model, the closed loop transfer function of the PLL can be easily derived in (24). Certain relations can be derived by rewriting (24) in the general form of a second order system.
The PItype LF can be designed with the desired system performance expectations with (26), i.e.
ω_{n}
and
ξ
. The damping factor The
ξ
is commonly selected as 0.707 for sufficient tracking speed and suppression of the overshoot. Because
G_{PD}
is usually constant, the LF parameters are dominated by
ω_{n}
. For certain values of
ω_{n}
, the relation between
k_{p}
and
k_{i}
is expressed in (27). The special case of the proposed PLL is expressed in (28) when
ξ
= 0.707 and the input is 1.5 times the perunit value. In most PLLs,
G_{PD}
is equal to the input magnitude. However, based on the analysis in Section II, it should be pointed out that
G_{PD}
is equal to half of the single phase input signal. Attention should be paid to this when calculating the LF parameters.
In this way,
ω_{n}
becomes the only parameter of concern while
ξ
is already selected as 0.707. The natural frequency
ω_{n}
is decided by the desired bandwidth
ω_{b}
. Let the module of (25) equal
, and when
s
=
j
·
ω_{b}
, the coefficient between
ω_{n}
and
ω_{b}
can be obtained, i.e.
ω_{b}
≈ 2.1 ·
ω_{n}
.
The dynamic performances of the PLL under common conditions are evaluated with different values of
ω_{n}
, and the results are shown in
Fig. 7
. In this case,
m
is the ration between the natural frequency
ω_{n}
and the input fundamental frequency
ω
_{1}
.
Frequency estimation with different ω_{n} under (a) 90 degree phase angle jump, (b) 50% sag and (c) 2 Hz frequency step.
It can be observed that with a high
m
, which means a high bandwidth, the overshoot of the frequency estimation is considerably high. Large variations in the frequency estimation also mean a large mismatch in the phase angle estimation during transients. Especially in
Fig. 7
(c), the transient overshoot is undesirable with a high
m
value. What is even worse, the frequency estimation fails to converge back to the new frequency of the input, when
m
=0.7. This is not acceptable in PLL applications.
The ±1 Hz band around the actual frequency is also plotted in dotdash lines in all of the simulation results. A low bandwidth LF brings a small frequency estimation overshoot but at the expense of a longer settling time.
The phase angel comparison in
Fig. 8
, with
m
=0.2, states that: 1) after a sudden 90 degree phase angle jump, the estimated phase angle mismatch is well limited after 1 input signal cycle, and the mismatch is totally eliminated after 3 input signal cycles. 2) The input sag brings a minor influence to the estimated phase angle. 3) The frequency step is the most critical condition for the proposed PLL. The proposed PLL limits the phase angle mismatch in a small range after one input cycle. The phase angle estimation coincides with the actual phase angle totally after 4 input cycles.
Phase angle comparison with m=0.2 under (a) 90 degree phase angle jump, (b) 50% sag and (c) 2 Hz frequency step.
It is well accepted that the antiharmonics performance of a PLL can be enhanced by reducing the bandwidth. Because the PLL is a low pass filter, which can be known from (25), it provides attenuation at a slope of 20dB/Dec at a frequency higher than
ω_{b}
. However, the bandwidth is related to the lock range of the PLL. Too small a
ω_{b}
will cause a lock failure under certain conditions. It order to deal with this failure, the value of
ω
_{1}
in
Fig. 3
should be set to the frequency of concern. The abandoned set of
ω
_{1}
should be avoided.
IV. EXPERIMENTS AND COMPARISONS
The proposed PLL is verified and evaluated by comprehensive tests. The test platform is sketched in
Fig. 9
. During the test, a function/arbitrary waveform generator (Agilent 33220A) is used to generated a set of particular single phase inputs. The PLL algorithm is implemented in a DSP development board based on a TMS320F28069, whose clock frequency is 90MHz. The estimated phase angle and frequency are visualized by an onboard digitalanalog converter.
Experiment setup for PLL evaluation.
 A. Sampling Frequency Selection
It is important to select a suitable sampling frequency when implementing algorithms in a digital signal processor. However, when any algorithm is implemented in a digital control IC, there are 2 time constants which need considerations, i.e. the sampling period T
_{s}
and the computation delay T
_{d}
. A diagram of a PLL digital implementation considering these 2 time constants is shown in
Fig. 10
.
PLL implementation diagram considering discrete time constants.
T
_{s}
is introduced by the AD conversion, which translates the continuous input into a sampled data signal with a constant update rate. T
_{d}
is a time interval where the algorithm completes the computation and makes the output update. After the PLL algorithm is implemented in a TMS320F28069, it can be measured that the maximum execution time of the proposed PLL algorithm for each phase angle estimation update is 16.756μs, i.e. 59 kHz. In other words, a T
_{s}
which is smaller than 17μs is unnecessary. Because T
_{d}
is about 17μs in the PLL program, the extra sampled data will be dropped by the controller IC since it will not read new samples before the computation is complete.
A lower sampling frequency results in a delay and an excessive overshoot as shown in
Fig. 11
. It can be seen that the overshoot and the settling time become small when the sampling frequency is increased. However, there are very small distinctions between the results when using 5 kHz sampling frequency and 50 kHz sampling frequency. Therefore, an extremely high sampling frequency is also unnecessary. It brings less of a performance contribution but a higher thermal stress to the controller IC. The selection of a sampling frequency should follow the target of the application. If only the steady state results of fundamental frequency are of concern, an extreme low sampling frequency such as 500Hz (10 samples in each input cycle) can be selected. For grid connected converters, such as distributed generation and laptop adapters, the sampling frequency should be set as high as the switching frequency. If the digital Pulse Width Modulation (PWM) is configured in double update mode, it is better to set the sampling frequency at twice the switching frequency. The premise of such a selection is that all of the control algorithm computations can be completed within one selected sampling period, i.e.
T_{d}
≤
T_{s}
.
Frequency estimation under 90 degree phase jump with different sampling frequency.
 B. Static and Dynamic Performance Evaluation
The evaluation follows the procedure below. During the tests, the sampling frequency is set at 10 kHz.

1) The input signal is zero;

2) The input is switched to a sinusoidal source;

3) 300ms later (hereafter called steptime) certain variations are performed in the input signal;

4) The results of 200ms (10 input cycles) are recorded and shown in the following parts.
In the experiments, the proposed PLL is carried out with the following parameters to cope with all of the serious tests. The LPF1 and the LPF2 have the same cutoff frequency as 0.707 times of the input fundamental frequency, i.e. 35.35Hz. In addition, a PItype LF is selected with
k_{p}
=124.4 and
k_{i}
=5803. This means that the system natural frequency is 10Hz, the damping factor is 0.707, and the PLL bandwidth is 21Hz.
The experiment results are shown in
Fig. 12
to
Fig. 14
, corresponding to a 90 degree phase angle jump, a 50% input sag and a 2 Hz frequency jump, respectively. The largest overshoot in the frequency estimation during transition happens in the phase angle jump test. The phase angle estimation slope bends in the first 20 ms. After another 20 ms, it restores a high linearity. This phenomenon is shown in
Fig. 12
. The input sag has a minimal influence on the proposed PLL. Although a 50% sag is an unusual case, the phase angle estimation maintains a high linearity during transition. This is due to the fact that the observed fluctuation in the frequency estimation is quite small. In the above 2 tests, the phase estimation is well limited within one input cycle.
Phase angle jump experiment results.
Input sag experiment results.
Frequency jump experiment results.
In the 2 Hz frequency step test, the excitation contains both a 2 Hz frequency increase and a 3.77 rad (216.7 degree) phase jump. The phase jump is introduced by the generation of the input’s instantaneous phase angle
θ_{i}
(
t
) . Since the value of
t
in (29) is not zero when the 2 Hz step is added, a certain phase jump is also added.
It can be observed from
Fig. 14
that the transition is smooth, and that the overshoot in the frequency estimation is much smaller than in the case of the phase angle jump test. However, the proposed PLL takes 4 input cycles to fully eliminate the estimation error.
In order to experimentally evaluate the behavior of the proposed PLL when the input is distorted, the PLL is implemented in a cascaded Hbridge multilevel static VAR generator prototype. the grid voltage is distorted by 20% of the 5
^{th}
order harmonics which is provided by a programmable voltage source. The results are shown in
Fig. 15
. The ripples in the frequency estimation are well limited to below 0.5 Hz. This guarantees the phase angle estimation with a high linearity. As mention in the LF design, decreasing the PLL bandwidth helps attenuate such ripples at the cost of settling time.
Experiment results with 20% 5th harmonics distortion.
 C. Comparisons with Existing PLLs
As introduced in section I, a lot of PLLs have been proposed in the literature. As a result, it is necessary to make comparisons among the proposed PLL and the existing PLLs. The SOGIPLL
[13]
,
[22]
,
[23]
and the delayed signal PLL
[14]
,
[21]
are selected for the comparison because of their widespread use in single phase converters. The loop filters in all of the PLLs are identical, i.e.
k_{p}
=124.4 and
k_{i}
=5803. For the SOGIPLL, the SOGI coefficient is 1. Additionally, because the original structures of the SOGIPLL and the delayed signal PLL are not frequency adaptive, a certain frequency feedback is added in order to compare their performance under frequency variations. In addition, the sampling frequency is 20 kHz in order to obtain enough data for the delayed signal PLL. The results are shown in
Fig. 16
and
Fig. 17
.
Comparative results under 90 degree phase angle jump. (a) the proposed PLL, (b) delayed signal PLL and (c) SOGIPLL.
Comparative results under 2 Hz frequency step. (a) proposed PLL, (b) delayed signal PLL and (c) SOGIPLL.
As can be seen from
Fig. 16
, i.e. the phase angle error under a 90 degree phase angle jump, the delayed signal PLL features the largest overshoot. However, the transition of the delayed signal PLL is minimal. After less than 2 input cycles, the phase angle error is well limited. Comparatively, the overshoot of the proposed PLL is minimal. After one and half input cycles the phase angle error is well limited. At the same time, the SOGIPLL features a long transition and a remarkable overshoot. Therefore, the proposed PLL is the most convenient of the PLLS under the phase angle jump test.
A 2 Hz frequency step test is also concluded with these 3 PLLs, and the results are shown in
Fig. 17
. It is shown that all of the PLLs can operate in the new operation point, and that the phase angle error is much smaller than in the cases of phase angle jump test. The proposed PLL has the minimum overshoot in terms of the phase angle error. Ripples can be observed in the other 2 PLLs. In the delay signal PLL, the number of necessary samples for the delay is not an integer after the frequency jump. This results in the undesired small ripple shown in
Fig. 17
(b). An even bigger ripple is founded in the SOGIPLL. It is introduced by the simple forward Euler discretization of the SOGI block in the comparison. A discussion on the discretization can be found in
[22]
. Although ripples can be found, they do not bring undesired effects because they are too small. It has to be mentioned that the proposed PLL shows no ripples, even though the simple forward Euler discretization is employed in all of the comparisons.
The comparative results state that the proposed PLL shows more convenient performance when the input has a single frequency. Unlike the SOGIPLL, which provides attenuation of the harmonics in the QSG block, the proposed PLL can only provide harmonic attenuation by decreasing its bandwidth. However, as shown in
Fig. 15
, convenient estimation can also be obtained in the steady state by the proposed PLL even under a distorted grid.
V. CONCLUSIONS
The issues related to the single phase PLL are systematically discussed in this paper. By introducing conjugate rotating vector pair, the pulsing single phase input vector in the stationary reference frame is synthesized with both the positive rotating vector and the negative rotating vector. A single phase SRFPLL is proposed in this paper by utilizing the characteristics of the conjugate rotating vector pair. It is distinguished by a constant zero β input. The design of the PLL parameters is divided into the LPF design in the PD and the loop filter design. A time domain phase detector model of the proposed PLL is derived, and a numerical simulation method is proposed to evaluate the performance of the PD with different LPF cutoff frequencies. The LF design is revised, and a brief guideline is described. Comprehensive tests are executed to verify and evaluate the proposed PLL performance. The bandwidth of the PLL, which is between 0.21 and 1.05 times the input frequency, is acceptable. Experimental results show that the PLL with suitable parameters provides satisfactory phase angle estimation in one input cycle. They also show that the estimation error can be eliminated in four input cycles for the most severe conditions.
Acknowledgements
This work was supported in part by National Science Foundation of china under project No. 50707025 and by the National High Technology Research and Development Program ("863"Program) of China under Project 2012AA050206. Most of the instruments in the work are supported by the State Key Laboratory of Electrical Insulation and Power Equipment, China
BIO
Ming Li (S’08) received his B.S. and M.S. degrees from Xi’an Jiaotong University, Xi’an, China, in 2005 and 2008, respectively. He is currently working towards his Ph.D. degree at the same university. From 2009, he became a member of the State Key Laboratory of Electrical Insulation and Power Equipment, China. From 2010 to 2011, he was in charge of the Xi’an Jiaotong University Power Electronics Chapter of the IEEE. He has published more than 20 technical papers and holds 2 patents. His current research interests include distributed FACTS, renewable power generation, high power IGBT gate drives, energy storage and power quality.
Yue Wang (M’05) received his B.S. degree from Xi’an Jiaotong University, Xi’an, China, in 1993, his M.S. degree from Beijing Jiaotong University, Beijing, China, in 1999, and his Ph.D. degree from Xi’an Jiaotong University, in 2003. He is currently an Associate Professor at Xi’an Jiaotong University. He received a National S&T Progress Award of China, in 2011. His current research interests include power quality, wind power generation, motor drives, multilevel converters, and flexible ac power transmission.
Xiong Fang received his B.S. degree from the Huazhong University of Science and Technology, Wuhan, China, in 2008, and his M.S. degree from Xi’an Jiaotong University, Xi’an, China, in 2011. He is presently working at FSPPowerland Technology Inc., Nanjing, China. His current research interests include the design and control of single phase inverters and rectifiers.
Yuan Gao received his B.S. degree from the Department of Electrical Engineering, Xi’an Jiaotong University, Xi’an, China, in 2012. He is currently working toward his M.S. degree at the same university. His current research interests include bidirectional isolated DC–DC converters and digital control techniques for power electronics converters.
Zhaoan Wang (SM’98) received his B.S. and M.S. degrees from Xi’an Jiaotong University, Xi’an, China, in 1970 and 1982, respectively. He received his Ph.D. degree from Osaka University, Osaka, Japan, in 1989. From 1970 to 1979, he was an Engineer at Xi’an Rectifier Factory, Xi’an, China. Since 1982, he has been a Lecturer at Xi’an Jiaotong University, where he is currently an honored Professor. He is also an appointed expert in State Key Laboratory of Electrical Insulation and Power Equipment, China. He is the author or coauthor of more than 200 technical papers and has led numerous government and industrysponsored projects in the areas of power and industrial electronics. He received the National Teaching Award, in 2009, and the National S&T Progress Award of China, in 2011. His current research interests include power conversion systems, power quality, motor drives and power electronic integration.
Roncero J. R.
2008
“Integration is key to Smart Grid management,”
2008 IETCIRED Seminar in Smart Grids for Distribution
1 
4
Guoqiao S.
,
Dehong X.
,
Luping C.
,
Xuancai Z.
2008
“An improved control strategy for gridconnected voltage source inverters with an LCL filter,”
IEEE Trans. Power Electron.
23
(4)
1899 
1906
DOI : 10.1109/TPEL.2008.924602
Cortes P.
,
Rodriguez J.
,
Quevedo D. E.
,
Silva C.
2008
“Predictive current control strategy with imposed load current spectrum,”
IEEE Trans. Power Electron.
23
(2)
612 
618
DOI : 10.1109/TPEL.2007.915605
Ruxi W.
,
Wang F.
,
Boroyevich D.
,
Burgos R.
,
Rixin L.
,
Puqi N.
,
Rajashekara K.
2011
“A high power density singlephase PWM rectifier with active ripple energy storage,”
IEEE Trans. Power Electron.
26
(5)
1430 
1443
DOI : 10.1109/TPEL.2010.2090670
Divan D.
,
Johal H.
2007
“Distributed FACTS – A new concept for realizing grid power flow control,”
IEEE Trans. Power Electron.
22
(6)
2253 
2260
DOI : 10.1109/TPEL.2007.909252
Wiik J. A.
,
Wijaya F. D.
,
Shimada R.
2009
“Characteristics of the magnetic energy recovery switch (MERS) as a series FACTS controller,”
IEEE Trans. Power Del.
24
(2)
828 
836
DOI : 10.1109/TPWRD.2008.2005879
Zhihui Y.
,
de Haan S. W. H.
,
Ferreira J. B.
,
Cvoric D.
2010
“A FACTS device: Distributed powerflow controller (DPFC),”
IEEE Trans. Power Electron.
25
(10)
2564 
2572
DOI : 10.1109/TPEL.2010.2050494
Santos Filho R. M.
,
Seixas P. F.
,
Cortizo P. C.
,
Torres L. A. B.
,
Souza A. F.
2008
“Comparison of three singlephase PLL algorithms for UPS applications,”
IEEE Trans. Ind. Electron.
55
(8)
2923 
2932
DOI : 10.1109/TIE.2008.924205
Namadmalan A.
,
Moghani J. S.
2013
“Singlephase current source induction heater with improved efficiency and package size,”
Journal of Power Electronics
13
(2)
322 
328
DOI : 10.6113/JPE.2013.13.2.322
Dong D.
,
Thacker T.
,
Burgos R.
,
Boroyevich D.
,
Wang F.
2009
“On zero steadystate error of singlephase PWM inverters voltage control and phaselocked loop system.”
IEEE Energy Conversion Congress and Exposition
892 
899
Roshan A.
,
Burgos R.
,
Baisden A. C.
,
Wang F.
,
Boroyevich D.
2007
“A DQ frame controller for a fullbridge single phase inverter used in small distributed power generation systems,”
IEEE Applied Power Electronics Conference
641 
647
Thacker T.
,
Burgos R.
,
Wang F.
,
Boroyevich D.
2009
“Singlephase islanding detection based on phaselocked loop stability,”
IEEE Energy Conversion Congress and Exposition
3371 
3377
Ciobotaru M.
,
Agelidis V. G.
,
Teodorescu R.
,
Blaabjerg F.
2010
“Accurate and lessdisturbing active antiislanding method based on PLL for gridconnected converters,”
IEEE Trans. Power Electron.
25
(6)
1576 
1584
DOI : 10.1109/TPEL.2010.2040088
Wang Y. F.
,
Li Y. W.
2013
“A grid fundamental and harmonic component detection method for singlephase systems,”
IEEE Trans. Power Electron.
28
(5)
2204 
2213
DOI : 10.1109/TPEL.2012.2214445
Gardner F. M.
2007
Phase Lock Techniques
Wiley
Chap. 1
Awad H.
,
Svensson J.
,
Bollen M. J.
2005
“Tuning software phaselocked loop for seriesconnected converters,”
IEEE Trans. Power Del.
20
(1)
300 
308
DOI : 10.1109/TPWRD.2004.837823
Thacker T.
,
Boroyevich D.
,
Burgos R.
,
Fei W.
2011
“Phaselocked loop noise reduction via phase detector implementation for singlephase systems,”
IEEE Trans. Ind. Electron.
58
(6)
2482 
2490
DOI : 10.1109/TIE.2010.2069070
Blaabjerg F.
,
Teodorescu R.
,
Liserre M.
,
Timbus A. V.
2006
“Overview of control and grid synchronization for distributed power generation systems,”
IEEE Trans. Ind. Electron.
53
(5)
1398 
1409
DOI : 10.1109/TIE.2006.881997
Rodriguez P.
,
Pou J.
,
Bergas J.
,
Candela J. I.
,
Burgos R. P.
,
Boroyevich D.
2007
“Decoupled double synchronous reference frame PLL for power converters control,”
IEEE Trans. Power Electron.
22
(2)
584 
592
DOI : 10.1109/TPEL.2006.890000
Choi H. J.
,
Song S. H.
,
Jeong S. G.
,
Choi J. Y.
,
Choy I.
2011
“Enhanced dynamic response of SRFPLL system for high dynamic performance during voltage disturbance,”
Journal of Power Electronics
11
(3)
369 
374
DOI : 10.6113/JPE.2011.11.3.369
Sakamoto S.
,
Izumi T.
,
Yokoyama T.
,
Haneyoshi T.
2002
“A new method for digital PLL control using estimated quadrature two phase frequency detection.”
Power Conversion Conference, PCCOsaka 2002, Vol. 2
671 
676
Ciobotaru M.
,
Teodorescu R.
,
Blaabjerg F.
2006
“A new singlephase PLL structure based on second order generalized integrator,”
IEEE Power Electronics Specialists Conference
1 
6
Rodriguez P.
,
Luna A.
,
Etxeberria I.
,
Hermoso J. R.
,
Teodorescu R.
2009
“Multiple second order generalized integrators for harmonic synchronization of power converters,”
IEEE Energy Conversion Congress and Exposition
2239 
2246
Bojoi R. I.
,
Limongi L. R.
,
Roiu D.
,
Tenconi A.
2011
“Enhanced Power Quality Control Strategy for SinglePhase Inverters in Distributed Generation Systems,”
IEEE Trans. Power Electron.
26
(3)
798 
806
DOI : 10.1109/TPEL.2010.2103572
KarimiGhartemani M.
,
Khajehoddin S. A.
,
Jain P. K.
,
Bakhshai A.
,
Mojiri M.
2012
“Addressing DC component in PLL and notch filter algorithms,”
IEEE Trans. Power Electron.
27
(1)
78 
86
DOI : 10.1109/TPEL.2011.2158238
Li M.
,
Wang Y.
,
Fang X.
,
Wang K.
,
Lei W.
,
Wang Z.
,
Liu J.
2010
“A novel single phase synchronization method for DFACTS controller based on symmetrical components synthesis,”
IEEE Energy Conversion Congress and Exposition
1519 
1524