CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 7, NO.

4, DECEMBER 2022 451

An Improved Constant Power Generation Algorithm

for Photovoltaic Systems
Tangwu YANG, Xingshuo LI, and Shuye DING

Abstract—High penetration of photovoltaic power generation 80

irr:100 W/m 2
will bring some new challenges to the grid, such as overvoltage irr:1000 W/m 2 CPG mode MPP
and overload. To address these problems, a constant power gen- 60

Power (W)
eration (CPG) algorithm is proposed. However, the conventional Plimit
CPG algorithm cannot balance tracking speed and accuracy. 40
A B
Therefore, an improved CPG algorithm with rapid tracking
speed and high accuracy is proposed in this paper. Finally, the 20 MPPT mode V oc
superiority of the algorithm is verified in simulation and experi- V oc
ments. 0
0 5 10 15 20 25
Voltage (V)
Index Terms—Active power control, constant power generation
(CPG), fast-tracking, photovoltaic system, power oscillations. Fig. 1. The operating point may fall into the open-circuit condition during the
fast decreasing irradiance.

I. Introduction 80

A
W/m22
irr:800W/m
irr:800
S fossil fuel is going to exhaust, renewable energy like 60
photovoltaic (PV) has drawn more attention in recent MPP
MPP
Power (W)

Plimit
years. In the past decades, the photovoltaic power plants 40
(PVPPs) usually are operated at the maximum power point A B
V step
step
(MPP) to maximize energy harvest efficiency [1]. However, it 20
V pv * V
has been found that the high penetration rate of PV has brought V old
many problems, including grid overload and frequency devia- 0
0 5 10 15 20 25
tion [2], [3]. Voltage (V)
To solve these above problems, stricter regulations have
been adopted to transform the PVPPs [4]. In some cases, the Fig. 2. The schematic diagram of CPG control algorithm during the Ppv > Plimit
period.
active power from the PVPPs should be limited to a limited
constant power [5], [6]. As shown in Fig. 1, if the maximum
power output from the PVPPs is lower than this limited power, (1)
it will work in the MPPT mode. Otherwise, it will maintain at
this limited active power [7].
where is the determined reference voltage, Ppv is the active
So far, several CPG algorithms have been proposed to track
power generation, Plimit is the limited constant power genera-
this limited power by perturbing the operating point towards
tion, Vpv is the measured PV voltage, Vstep is the perturbation
point A or B. Since the perturbation at point B may exceed the step size, and VMPPT refers to the reference volatage derived by
open-circuit voltage in case of rapid changes in solar irradi- MPPT algorithm.
ance, point A is usually chosen as the perturbation point in the While this CPG scheme is popular thanks to the straight-
previous research [8]-[10]. As illustrated in Fig. 2, the pertur- forward structure, its performance highly depends on the
bation point can be tracked by using the reference PV voltage selection of the step size Vstep. The traditional method adopts
： a fixed step size (FSS), which leads to a trade-off between the
tracking speed and accuracy[11], [12]. This means that a larger
Manuscript received January 31, 2022; revised April 18, 2022; accepted May 28, step size can result in faster tracking but lower accuracy, and
2022. Date of publication December 30, 2022; date of current version December
a small step size has the opposite effect. To resolve the dilem-
22, 2022. This work was supported in part by the 2021 Carbon Emission Peak
and Neutrality (Industry Prospects and Key Core Technologies) Project of Jiangsu ma, the tracking method with variable step size (VSS) has
Province under Grant SBE2021100009 and in part by the Graduate Science and been proposed in [13]. This VSS method divides the Vstep into
Innovation Projects in Jiangsu Province under Grant SJCX22_0554. (Corresponding steady-state and transient state. During the transient and steady
author: Xingshuo Li.) operation, the perturbation step size keeps to be adjusted based
All authors are with School of Electrical and Automation Engineering, Nanjing
on the position of the current operating point. This algorithm
Normal University, Nanjing, China (e-mail: [email protected]; xingshuo.
[email protected]; [email protected]). achieves both high tracking accuracy and speed and can usu-
Digital Object Identifier 10.24295/CPSSTPEA.2022.00041 ally function well with relatively stable irradiance vibration.
452 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 7, NO. 4, DECEMBER 2022

L 5
I pv MPPT→
MPPT → CPG
CPG Power
Loss
D 4
+ Cpv +

Current (A)
V pv V out R 3 MPP
- C
- PWM
PWM B
2 A
V pv Plimit
CPG/MPPT 1
I pv
0
Flag
Flag K Plimit
limit 0 5 10 15 20 25
Voltage (V)
(a)
Fig. 3. System configuration of CPG procedure.
5
CPG→
CPG CPG
→ CPG
However, the effectiveness degrades when it works under ex- 4
treme weather conditions. C B
MPP

Current (A)
To address the above issue, a novel CPG algorithm is pro- 3
A
posed in this paper. The proposed method introduces the esti- 2
mation of the voltage reference and the selection of the voltage Plimit
1
step, which make it possible to estimate the future steady
state’s reference voltage even in the transient state with rapid 0
0 5 10 15 20 25
irradiation changes. Moreover, the step size can be additionally Voltage (V)
modified based on the present power generation and its limited (b)
power in the steady-state. Therefore, the power oscillation in
the steady-state is reduced, and the tracking speed in the tran- Fig. 4. Mode transition diagram for the conventional CPG control under the
irradiation enhancement. (a) From MPPT mode to CPG mode. (b) From CPG
sient state is increased. mode to CPG mode.
The remainder of this paper is organized as follows: the con-
5
ventional CPG control compared as a benchmark is introduced CPG→
CPG CPG
→ CPG
in Section II. Section III demonstrates the derivation and basic 4 A
operation principle of the proposed method. Simulation and B MPP
Current (A)

3
experiment tests are implemented, and the results are compared C
and analyzed in Sections IV and V to verify the effectiveness 2
of the proposed strategy. Finally, the conclusions are drawn in Plimit
1 Power
Section VI.
Loss
0
0 5 10 15 20 25
Voltage (V)
II. Conventional CPG Control
(a)
The basic operation principle of the conventional CPG con-
5
trol is demonstrated in Fig. 3. The PVPP’s output voltage and CPG→
CPG → MPPT
MPPT
current are measured to calculate the power generation Ppv. 4
Ppv is constantly compared with the power limitation Plimit to A
Current (A)

3 MPP
update the duty cycle D regulating the converter’s conversion
ratio. More detailed analysis will be given in this section to 2 B C
show this method results in slow response in transient state Plimit
1
and significant power oscillation in steadystate. The analysis is
based on both increasing and decreasing irradiance scenarios. 0
0 5 10 15 20 25
Voltage (V)
A. Increasing Irradiance Scenario (b)

When the irradiation is increased, there will be a transition Fig. 5. Mode transition diagram for the conventional CPG control under the
between two different modes. Fig. 4 shows the conventional irradiation reduction. (a) From CPG mode to CPG mode. (b) From CPG mode
to MPPT mode.
CPG’s operation process when the solar irradiance goes up. As
shown in Fig. 4(a), the PVPP will switch from MPPT mode to
will continue CPG mode and move towards point C to avoid
CPG mode once it detects the power generation at B exceeds
overwhelming power.
the limitation. In this condition, the operation point will move
away from the MPP track with fixed step size to curtail the
power output actively. In another case, the PVPP may initially B. Decreasing Irradiance Scenario
work at A with the CPG operation mode. When it estimates Similar operation mode switch process can also be identified
that the power generation at MPP B will exceed the limit, it in the decreasing irradiance scenario. Fig. 5 shows the con-
T. YANG et al.: AN IMPROVED CONSTANT POWER GENERATION ALGORITHM FOR PHOTOVOLTAIC SYSTEMS 453

Start Determination of
operating point

Measure I and V
Control action block
Yes (Plimit
limit >P&&Pold >Plimit
limit) || No
Flag
Flag=1,K=2,
= 1, K = 2, (P >Plimit
limit&&Plimit
limit >Pold)?
DDold
old=
=45
45

Irradiation change determination block

Calculate
dP, dP*and dV
dP,dP Flag
Flag=Flag+1
= Flag + 1 Flag
Flag=Flag+0
= Flag + 0

Yes No
K==1? Yes No
|dP*||<
< dP*th2
Yes No Judgment of disturbance direction
KK==0?
= = 0?
Yes |dP| ≥ dPth1&& No K
K=1;
= 1; Zero oscillations Yes No
dP* >
>00˛
˛
|dP*| ≥ dP*th1 Flag
Flag=1
=1
Yes
|dP| ≥ dPth2
No Ŀ
ķ ŀ
ĸ
Conventional
Fast tracking Step
Step=0
=0 Step
Step=2^(2-Flag)
= 2^(2 Flag)
P&O Algorithm
DD=Eq.(11)
= Eq.(11) DD=Eq.(15)
= Eq.(15) Fast tracking
Determination Determination
of operating of operating Step
Step=
= ±2^(2-Flag)
±2^(2 Flag) |dP/dV|
|dP/dV|<e
< eth?
K
K=2;
= 2; point K
K=2;
= 2; point Yes
Flag
Flag=1
=1 Flag
Flag=1
=1
DD=D
= D old+
+Step
Step K=0
K =0

Dold = D,Vold = V, Iold = I

Dold = D, Vold = V, Iold = I
Return
Return

Fig. 6. Flowchart of the CPG method.

ventional CPG’s operation process when the solar irradiance 80

irr:800 W/m2
goes down. As shown in Fig. 5(a), the PVPP will execute a
60
rightward perturbation operation once it detects the power P2
Power (W)

generation at B falls the limitation. In this condition, the op-

40 MPP
eration point will move towards point C with fixed step size. Plimit P1 A B
When it comes to the vicinity of point C, there will be a point P = 20
Plimit. Therefore it no longer perturbs to the right, but oscillates P3
around point C. In another case, it is found that it is already at 0
the maximum point when it comes to the vicinity of point C. It 0 5 10 15 20 25
Voltage (V)
is no longer satisfied that the existence point P is greater than (a)
the point Plimit, so oscillates around MPP C.
Based on above analysis, it can be concluded that the con- 80
ventional CPG method with FSS has two major drawbacks. irr:800 W/m2
one is that the tracking speed is relatively slower during the ir- 60
Plimit
radiance transition period. Another is that the power oscillation
Power (W)

during the steady stage can result in significant power loss. 40 MPP
P1 P2
20
III. The Proposed CPG Algorithm
As shown in Fig. 6, the proposed control scheme consists of 0
0 5 10 15 20 25
two major function blocks named irradiation change determi- Voltage (V)
nation block and control action block. The proposed method (b)
aims at speeding up the tracking process and reducing the
power loss caused by oscillation. In Fig. 6, there is only one Fig. 7. Judgment of disturbance direction.
parameter, dP*th2 , which is used to judge the changes of the ex-
ternal environment. Here, the variable parameter dP* is used to disturbance.
judge. That is, when dP* is greater than dP*th2 , it is considered
that the external environment has changed, so it is necessary
1) dP* > 0
to initialize the step size and jump the target duty cycle. On the
According to algorithm ① in Fig. 6, the operating points will
contrary, when dP* is less than dP*th2 , it is considered to be still
move to the left side with fixed steps. In this case, both point P2
in a steady state. Where dP* refers to the difference between the
and point B move to point A.
current power P and the limited power Plimit.
2) dP* < 0
A. The Judgment of Disturbance Direction
Based on the operation principle ② of the proposed algo-
Due to the boost converter’s electrical characters, the direc- rithm , operating points on both sides will climb to point CPP in
tion of its operating point voltage is opposite to the duty cycle Fig. 7(a) if power generation beyonds the limitation. Otherwise,
454 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 7, NO. 4, DECEMBER 2022

5
MPPT→
MPPT → CPG
C
CPG
C
4 (5)
D
Current (A)

3 MPP
E B
2
According to the paper, the formula for calculating the duty
A
Plimit cycle of point D of the Boost circuit is:
1

0 (6)
0 5 10 15 20 25
Voltage (V)
(a) where V and Vold refer to the current voltage and past voltage
respectively, I refers to the current current and Voc refers to the
5
CPG→
CPG CPG
→ CPG
open circuit voltage.
4 E D Therefore, after the next iteration, point B jumps directly to
MPP point D, and then traces to point E, which can absolutely im-
Current (A)

C B
3 prove the tracking speed.
A
2 Besides, there will be another transition modes from CPG
Plimit to CPG under the irradiation enhancement. when irradiance
1 increases, point A moves towards point B on the load line.
0 From Fig. 8(b), in order to reduce tracking time, it is nec-
0 5 10 15 20 25 essary to allow point B to jump to point D which is close to
Voltage (V)
point E. Taking into account that the current at point B is ap-
(b)
proximately equal to the current at Point E, the following idea
Fig. 8. Mode transition diagram for the proposed CPG control under the comes up.
irradiation enhancement. (a) From MPPT mode to CPG mode. (b) From CPG
mode to CPG mode.
(7)
the PVPPs will keep following the MPPs to maximize erner-
gy harvest (Fig. 7(b)). It should be noted that dP* denotes the (8)
differece between the power and power limit in the algorithm
flowchart, dP is the power change over time. P, Pold and PMPP
indicates the PVPP’s power generation at current time instant, (9)
power output at last time instant and the available maximum
power. Furthermore, the duty cycle of point D of the Boost circuit
is:
B. Changes in Irradiation Conditions

1) Irradiation Enhancement (10)

In order to increase the tracking speed, the following schemes
are proposed. After the next iteration, point B jumps directly to point D,
From Fig. 8(a), in consideration of jumping to the vicinity of and then traces to point E.
the target point in one step as much as possible, the principle of
similar triangles is used [14]. 2) Irradiation Reduction
Similarly, when irradiance increases, point A moves towards Similar to the enhancement of irradiation, the following
point B on the load line. Then, by applying the principle of sim- section briefly analyzes the transition state under the irradiation
ilar triangles, point C can be derived as: reduction.
It is critical to take one step to jump to the vicinity of the
(2) target point E to reduce the tracking time. As can be seen in
Fig. 9(a) and (b), point D can be constructed with the help of
(3) inverse proportional curve and point B. The specific idea is as
follows:
Rearrange (3), and obtain (4) (11)

(12)
(4)

(13)
Then, the desired point C is determined by
T. YANG et al.: AN IMPROVED CONSTANT POWER GENERATION ALGORITHM FOR PHOTOVOLTAIC SYSTEMS 455

5 80
CPG→
CPG CPG
→ CPG irr:800 W/m2
4 A
60 step size = 2
B C MPP step size = 1
Current (A)

MPP

Power (W)
3 Plimit A
D E 40
2
Plimit 20
1
0
0 0 5 10 15 20 25
0 5 10 15 20 25
Voltage (V) Voltage (V)
(a) (a)

5 80
CPG→ MPPT
M PT
irr:800 W/m2
4 step size = 1/2
60
A
Current (A)

step size = 1/4 MPP

Power (W)
3 Plimit A
B E C 40
y
2 plif
D
Plimit Am
20
1

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Voltage (V) Voltage (V)
(b) (b)
Fig. 9. Mode transition diagram for the proposed CPG control under the Fig. 10. The schematic diagram of the proposed zero oscillation method.
irradiation reduction. (a) From CPG mode to CPG mode. (b) From CPG mode
to MPPT mode. 200 W around Plimit. That is, when the formula (18) is satisfied,
the step size is selected as 0, and the Flag is reset to 1. The
So, the duty cycle of point D is that: variable K is set to 1 as the state description parameter, which
means that it is currently in the stable state of the CPG mode
[17], [18].
(14)
(15)
(16)
C. The Reduction of Power Oscillations
(17)
To reduce power oscillations at point A in Fig. 7(a), an ad-
vanced method is proposed to solve it. (18)
When the operating point comes near point A, as shown in
Fig. 10, the step size will be shortened spontaneously as the
direction of the disturbance changes. In Fig. 10(a), point P is
IV. Simulation Results
greater than point Plimit. From the above algorithm ① , it can be To prove the superiority of the proposed algorithm, it has
seen that point P is perturbed to the left with a certain step (the been simulated in the MATLAB Simulink. The proposed
initial step size is 2). If the P point crosses the Plimit point, the scheme is tested with three different weather conditions when
Flag is added by one; if not, the Flag remains unchanged and the PV model of EN50530 is applied. The model consists of 8
the leftward perturbation continues to be executed. That is, in PV (EN50530) Cells connected parallel and 20 PV (EN50530)
the judgment link in Fig. 6: When the formula (15) is satisfied, cells connected in series to make solar PV array. The relevant
it is considered that point P has crossed the Plimit point from parameters of PV model EN50530 are in Table I, and the sys-
above; For the same reason, when the formula (16) is satisfied, tem data of simulation and experiment are in Table II.
it is considered that point P has crossed the Plimit point from be- Next, two different time scales will be analyzed and dis-
low [15], [16]. cussed, namely a short time scale of 10 seconds and a long
Since the selected variable step-size formula is (17), the time scale of 400 seconds derived from local live solar irradi-
step-size can be reduced by half each time the point P cross- ation from 8 hours. In order to better highlight the advantages
es the point Plimit. As the point P repeatedly disturbs near the of the proposed method, this paper compares the traditional
Plimit point, the Flag keeps increasing, and the step size keeps method with the improved method. It should be emphasized
decreasing. Therefore, the final operating point gets closer and that both are based on the real-time changes of voltage and
closer to the point Plimit. In this paper, the steady-state is consid- current, and the duty cycle regulation is used to replace the PID
ered to be reached when the current power P is plus or minus controller.
456 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 7, NO. 4, DECEMBER 2022

TABLE I 12000
Main Parameters for EN50530 PV Modules Conventional method
10000 Improved method
Plimit = 7000 W

Power (W)
8000
6000
4000
2000
0
0 2 4 6 8 10
Time (s)
(a)

12000
Conventional method
10000 Plimit = 8000 W Improved method
TABLE II

Power (W)
8000
Main Parameters of the PV System
6000
4000
2000
F F 0
0 2 4 6 8 10
F F
Time (s)
(b)

Fig. 11. Simulation results during short periods of time. (a) Comparison
diagram of power waveform under power limit of 7000 W. (b) Comparison
diagram of power waveform under power limit of 8000 W.

A. Short-Term Simulation 1100

Solar Irradiation (W/m2)

The short-term simulation results are shown in Fig. 11. Take 1000
the power limit of 7000 W as an example for specific expla- 900
nation, it can be seen that PV system operates in MPPT mode 800
with maximum power output under the condition of 600 W/m2 700
irradiance. When the irradiance suddenly increases to 700 W/m2, 600
800 W/m2, 900 W/m2 and 1000 W/m2, the power generation 500
will exceed the power limitation if the system still operates in
100 200 300 400
MPPT mode. Thus, it has to switch to CPG mode with con- Time (s)
stant power output. (a)
Fig. 11(a) and (b) describe the results of the proposed algo-
rithm under the limited power of 7000 W and 8000 W, respec- 1100
Solar Irradiation (W/m2)

tively. It is obvious that in this environment, the two disadvan- 1000

tages of traditional algorithm have been well improved. Firstly, 900
the steady-state power oscillation is significantly reduced. Sec- 800
ondly, when the solar irradiance changes at any time, the pro- 700
posed algorithm successfully speeds up the tracking process.
600
As shown in Fig. 11, when the irradiance suddenly changes at
500
8 s, the tracking speed has been significantly improved, that is,
it can reach the steady state faster. 100 200 300 400
Time (s)
(b)
B. Long-Term Simulation
Fig. 12. Variation diagram of solar irradiation during long periods of time. (a)
The long-term scale can be divided into steady-state rise and Stable solar irradiance. (b) Unstable solar irradiance.
drastic change according to the change of solar irradiation, and
the change diagrams are shown in Fig. 12.
in 0-50 s and 350-400 s while it switches to the CPG mode
1) Stable Solar Irradiance during 50-350 s. The proposed alorathim significantly reduces
Fig. 13 shows the simulation results with one day meteoro- the power ripple when the PVPP works in CPG mode. Thus,
logical data. In this case, the solar irradiation sampling time is the power quality is improved with smoother generation.
400 s, the size interval is from 550 W/m2 to 900 W/m2, and the
solar irradiation is changed once every second. Based on Fig. 2) Unstable Solar Irradiance
13(c) and (d), it is clear that the PVPP operate in MPPT mode To furthermroe prove the effectiveness of the proposed
T. YANG et al.: AN IMPROVED CONSTANT POWER GENERATION ALGORITHM FOR PHOTOVOLTAIC SYSTEMS 457

600 600
Conventional method Conventional method
500 Improved method 500 CPG mode Improved method

Voltage (V)
400 400
Voltage (V)

300 300
200 200
CPG mode
100 MPPT MPPT
MPPT 100
0 mode mode
mode 0
0 100 200 300 400 0 100 200 300 400
Time (s) Time (s)
(a) (a)

600 600
Conventional method Conventional method
500 500 CPG mode Improved method
Improved method Plimit = 7 000 W

Voltage (V)
400 400
Voltage (V)

300 300

200 MPPT MPPT 200

CPG mode
100 mode mode 100

0 0
0 100 200 300 400 0 100 200 300 400
s
Time (s)
Time (s)
(b) (b)

12000 12000
Conventional method Conventional method
10000 Improved method 10000 Improved method
Plimit = 7000 W Plimit = 7 000 W
Power (W)

8000
Power (W)

8000
6000 6000
4000 4000
2000 2000
0 0
0 100 200 300 400 0 100 200 300 400
Time (s) Time (s)
(c) (c)

12000 12000
Conventional method Conventional method
10000 Improved method 10000 Improved method
Power (W)

8000
Power (W)

8000
6000 6000
Plimit = 8000 W Plimit = 8000 W
4000 4000
2000 2000
0 0
0 100 200 300 400 0 100 200 300 400
Time (s) Time (s)
(d) (d)

Fig. 13. Simulation results of slow changes in solar irradiation during long Fig. 14. Simulation results of sharp changes in solar irradiation during long
periods of time. (a) Comparison diagram of voltage waveform under power periods of time. (a) Comparison diagram of voltage waveform under power
limit of 7000 W. (b) Comparison diagram of voltage waveform under power limit of 7000 W. (b) Comparison diagram of voltage waveform under power
limit of 8000 W. (c) Comparison diagram of power waveform under power limit of 8000 W. (c) Comparison diagram of power waveform under power
limit of 7000 W. (d) Comparison diagram of power waveform under power limit of 7000 W. (d) Comparison diagram of power waveform under power
limit of 8000 W. limit of 8000 W.

scheme, it is tested under fastchanging solar irradiance (Fig. V. Experimental Evaluation

14). Clearly, the proposed algorithm is robust enough to handle
the extreme weather by successfully regulating the power gen-
A. Experimental Setup
eration within the limitation. In comparison to the traditional
method, it has better steady state performance with power os- The experiment platform is built to test the performance of
cillation. the proposed scheme. Fig. 15 shows the basic experimental
458 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 7, NO. 4, DECEMBER 2022

Real -Time Model

Ppv (2000 W/div), Vpv (200 V/div)

Host PC Ppv
Simulator
(MT 6016)

Osilloscope

Interface board
Vpv

Fig. 15. Experimental setup of the PV system.

Time (2 s/div)
(a)
setup of the experiment system where the hardware-in-the-
loop (HIL) is used. Ppv
The HIL system, MT 6016, is a high-performance semiph-

Ppv (2000 W/div), Vpv (200 V/div)

ysical simulation platform based on StarSim small-step simu-
lation technology and advanced FPGA hardware. The interface
board is connected to the MT 6016 and the oscilloscope gets 7000 W
the waveform through the port connection on the interface
board. The specification of the experiment’s parameters are
summarized in Table II while many are same to the simula-
tion’s configuration. Vpv

B. Experimental Results
Time (2 s/div)
Fig. 16 compares the proposed scheme with the conven- (b)
tional method in the short-period scenario with fast-changing
solar irradiance. The results show that with the increase of
Ppv (2000 W/div), Vpv (200 V/div)

irradiance, the power oscillation caused by traditional method

becomes more and more obvious. In contrast, whether the
power is limited to 7000 W or 8000 W, the proposed improved Ppv
method can always achieve ideal results. The experimental
results show that the proposed one possesses better dynamic
tracking characteristics in the transient state and eliminate the
power oscillation in the steady state.
Fig. 17(a)-(h) show the experimental results of the long-pe- Vpv
riod scenario. The time interval is the 40 s per div for a total of
400 s and the irradiation is set to change every seconds in this
Time (2 s/div)
case. Fig. 17(a) and (b) show the comparison between the tra- (c)
ditional method and the improved method under the condition
of power limitation of 7000 W in the case of long-time scale
with slow irradiance change. At the same time, Fig. 17(e) and
Ppv (2000 W/div), Vpv (200 V/div)

(f) are the comparison diagrams in the case of long-time scale

with drastic changes in irradiance. Fig. 17(c), (d), (g) and (h) 8000W Ppv
are the comparison diagrams of the power limit of 8000 W
under the above two different environments. It is clear that the
proposed scheme can always achieve fast tracking and zero os-
cillation in either stable or fast-changing irradiance conditions.
Vpv
VI. Conclusion
This paper has proposed a novel CPG control scheme to Time (2 s/div)
significantly speed up the tracking process in the transient-state (d)
and eliminate the power oscillation in the steady state. There-
Fig. 16. Experiment results with the proposed control during short periods of
fore, the proposed scheme can improve the PVPP’s overall en- sharp changes in irradiation: (a) Conventional CPG algorithm; (b) Proposed
ergy harvest efficiency and the power quality. The claimed im- CPG algorithm.
T. YANG et al.: AN IMPROVED CONSTANT POWER GENERATION ALGORITHM FOR PHOTOVOLTAIC SYSTEMS 459

Ppv (2000 W/div), Vpv (200 V/div)

Ppv (2000 W/div), Vpv (200 V/div)

Ppv
Ppv
7000 W

V pv Vpv

Time (40 s/div) Time (40 s/div)

(a) (b)

Ppv (2000 W/div), Vpv (200 V/div)

Ppv (2000 W/div), Vpv (200 V/div)

Ppv 8000 W Ppv

Vpv Vpv

Time (40 s/div) Time (40 s/div)

(c) (d)
Ppv (2000 W/div), Vpv (200 V/div)

Ppv (2000 W/div), Vpv (200 V/div)

Ppv 7000 W Ppv

Vpv Vpv

Time (40 s/div) Time (40 s/div)

(e) (f)
Ppv (2000 W/div), Vpv (200 V/div)

Ppv (2000 W/div), Vpv (200 V/div)

Ppv 8000 W Ppv

Vpv V pv

Time (40 s/div) Time (40 s/div)

(g) (h)

Fig. 17. Experiment results with the proposed control during long periods of solar irradiation.
460 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 7, NO. 4, DECEMBER 2022

provements are proved by both the simulation results and the [14] X. Li, H. Wen, and W. Xiao, “A modified MPPT technique based on the
experimental test in this article. As the proposed method may MPP-locus method for photovoltaic system,” in Proceedings of IECON
2017 - 43rd Annual Conference of the IEEE Industrial Electronics Soci-
result in a small mismatch between the tracked point and the
ety, Beijing, China, Oct. 2017, pp. 2338–2343.
actual MPP, future work will additionally optimize the scheme [15] A. Xl, B. Hw, B. Bc, A. Sd, and C. Wx, “A cost-effective power ramp
by improving the tracking accuracy. rate control strategy based on flexible power point tracking for photovol-
taic system - sciencedirect,” in Solar Energy, vol. 208, pp. 1058–1067,
Sept. 2020.
References [16] Y. Zhu, H. Wen, G. Chu, Y. Hu, and J. Ma, “High-performance photo-
[1] R. A. Mastromauro, M. Liserre, and A. D. Aquila, “Control issues in sin- voltaic constant power generation control with rapid maximum power
gle-stage photovoltaic systems: MPPT, current and voltage control,” in point estimation,” in IEEE Transactions on Industry Applications, vol.
IEEE Transactions on Industrial Informatics, vol. 8, no. 2, pp. 241–254, 57, no. 1, pp.714–729, Jan. –Feb. 2020.
May 2012. [17] X. Li, Q. Wang, H. Wen, and W. Xiao, “Comprehensive studies on op-
[2] A. Sangwongwanich, Y. Yang, and F. Blaabjerg, “High-performance erational principles for maximum power point tracking in photovoltaic
constant power generation in grid-connected PV systems,” in IEEE systems,” in IEEE Access, vol. 7, pp. 121407–121420, Aug. 2019.
Transactions on Power Electronics, vol. 31, no. 3, pp. 1822–1825, Mar. [18] H. -K. Yang, Z. Aleem, J. Lee, and J. -W. Park, “Constant power genera-
2015. tion method for grid-connected photovoltaic systems with fast response
[3] G. C. Mahato, T. R. Choudhury, B. Nayak, D. Debnath, S. B. Santra, and under dynamic irradiance condition,” in Proceedings of 2020 IEEE En-
B. Misra, “A review on high PV penetration on smart grid: Challenges ergy Conversion Congress and Exposition (ECCE), Detroit, MI, USA,
and its mitigation using FPPT,” in Proceedings of 2021 1st International Oct. 2020, pp. 2131–2135.
Conference on Power Electronics and Energy (ICPEE), Bhubaneswar,
India, Jan. 2021, pp. 1–6.
[4] H. D. Tafti, A. Sangwongwanich, Y. Yang, G. Konstantinou, and F.
Blaabjerg, “A general algorithm for flexible active power control of
photovoltaic systems,” in Proceedings of 2018 IEEE Applied Power Tangwu Yang received the B.S. degree in building
Electronics Conference and Exposition (APEC), San Antonio, TX, USA, electricity and intelligence from Suzhou University of
Mar. 2018. Science and Technology, China, in 2019.
[5] Y. Yang, F. Blaabjerg, and H. Wang, “Constant power generation of He is currently pursuing the M.S. degree in electri-
photovoltaic systems considering the distributed grid capacity,” in Pro- cal engineering from Nanjing Normal University, Nan-
ceedings of 2014 IEEE Applied Power Electronics Conference and Ex- jing, China. His research interests include renewable
position – APEC 2014, Fort Worth, TX, USA, Mar. 2014, pp. 379–385. energy technology and grid-connected in photovoltaic
[6] R. I. Yuwanda, E. Prasetyono, and R. P. Eviningsih, “Constant power system.
generation using modified MPPT P&O to overcome overvoltage on so-
lar power plants,” in Proceedings of 2020 International Seminar on In-
telligent Technology and Its Applications (ISITIA), Surabaya, Indonesia,
Jul. 2020, pp. 392–397.
[7] G. C. Mahato, T. Roy Choudhury, and B. nayak, “Study of MPPT and
FPPT: A brief comparison,” in Proceedings of 2020 IEEE 17th India Xingshuo Li received the B.S. degree in computer
Council International Conference (INDICON), New Delhi, India, Dec. science from Zhengzhou University, Zhengzhou,
2020, pp. 1–7. China, in 2012, and the M.S. degree in sustainable
[8] H. D. Tafti, A. I. Maswood, G. Konstantinou, J. Pou, and F. Blaabjerg, energy technology with distinction from Xi’an
“A general constant power generation algorithm for photovoltaic sys- Jiaotong-Liverpool University, Suzhou, China, in
tems,” in IEEE Transactions on Power Electronics, vol. 33, no. 5, pp. 2015. In 2019, he received the Ph.D. degree from of
4088–4101, May 2018. Liverpool, Liverpool, U.K. He currently is Associate
[9] A. Sangwongwanich, Y. Yang, F. Blaabjerg, and H. Wang, “Bench- Professor at School of Electrical and Automation
marking of constant power generation strategies for single-phase grid Engineering, Nanjing Normal University, Nanjing,
connected photovoltaic systems,” in IEEE Transactions on Industry China.
Applications, vol. 54, no. 1, pp. 447–457, 2016. His research interest includes renewable energy technology and distributed
generation, especially ancillary service, power forecasting and fault diagnosis
[10] A. Sangwongwanich, Y. Yang, and F. Blaabjerg, “A sensorless power
in photovoltaic systems.
reserve control strategy for two-stage grid-connected PV systems,” in
IEEE Transactions on Power Electronics, vol. 32, no. 11, pp. 8559–
8569, 2017.
[11] Y. Yang, H. Wang, F. Blaabjerg, and T. Kerekes, “A hybrid power con-
trol concept for PV inverters with reduced thermal loading,” in IEEE
Shuye Ding was born in 1978. He received the B.S.,
Transactions on Power Electronics, vol. 29, no. 12, pp. 6271–6275, Dec.
M.S., and Ph.D. degrees in electrical machinery and
2014.
appliance from the Harbin University of Science and
[12] H. D. Tafti, G. Konstantinou, C. D. Townsend, G. G. Farivar, S. Cebal-
Technology, Harbin, China, in 2001, 2004, and 2008,
los, J. Pou, and J. E. Fletcher, “Comparative analysis of flexible power respectively.
point tracking algorithms in photovoltaic systems,” in Proceedings of He is currently a Professor with the School of
2020 IEEE Energy Conversion Congress and Exposition (ECCE), De- Electrical and Automation Engineering, Nanjing
troit, MI, USA, Oct. 2020, pp. 110–115. Normal University, Nanjing, China. He is the author
[13] H. D. Tafti, A. Sangwongwanich, Y. Yang, J. Pou, G. Konstantinou, and F. or co-author of more than 80 published peer-re-
Blaabjerg, “An adaptive control scheme for flexible power point tracking viewed papers and holds more than ten patents. His
in photovoltaic systems,” in IEEE Transactions on Power Electronics, research interests include synthesis physical fields of large electrical machines
vol. 34, no. 6, pp. 5451–5463, Jun. 2019. and theoretical study of special electrical machines.