Topological Structure of Large-scale Photovoltaic ... - Science Direct

Available online at www.sciencedirect.com

ScienceDirect Energy Procedia 105 (2017) 113 – 118

The 8th International Conference on Applied Energy – ICAE2016

Topological structure of large-scale photovoltaic array and its MPPT controlling method Ruo-li Tanga,*, Zhou Wub, Yan-jun Fangc a

School of Energy and Power Engineering, Wuhan University of Technology, Wuhan, Hubei, China b School of Automation, Chongqing University, Chongqing, China c Department of Automation, Wuhan University, Wuhan, Hubei, China

Abstract

The “hot spot” problem, which is usually caused by the partial shading environmental conditions, threatens the safety operation of a photovoltaic system seriously. In this paper, a novel topological structure and its maximum power point tracking (MPPT) model and algorithm are researched to solve the “hot spot” problem in a large-scale photovoltaic system. Numerical result shows that with the application of the proposed structure and MPPT method, the large-scale photovoltaic system can achieve the minimum control unit (MCU) level MPPT, conquer the “hot spot” problem, and improve output power under complex environmental conditions significantly.

©©2017 by Elsevier Ltd. This an open access 2016Published The Authors. Published byisElsevier Ltd. article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Selection and/or peer-review under responsibility of ICAE Peer-review under responsibility of the scientific committee of the 8th International Conference on Applied Energy.

Keywords: Large-scale photovoltaic array; Maximum power point tracking; Large-scale global optimization

1. Introduction As one of the most important sustainable sources, solar energy has become extremely important for replacing the fossil fuel. The photovoltaic cell is usually regarded as an unstable power supplier due to the varying environmental conditions. In a photovoltaic system, the photovoltaic array is usually composed of many modules in series and parallel connections. Under partial shading conditions, shading of even a single module can reduce the efficiency of the entire system significantly, and possibly cause the socalled “hot spot” problem [1]. To overcome this weakness, the conventional method is to add a bypass diode across each module in parallel connections reversely, and to add a blocking diode in series connections at each branch. The added diodes eliminate the shaded module or even a branch to ensure operational safety of the system, however, it also causes an additional reduction of the output power.

1876-6102 © 2017 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the 8th International Conference on Applied Energy. doi:10.1016/j.egypro.2017.03.288

114

Ruo-li Tang et al. / Energy Procedia 105 (2017) 113 – 118

In this study, a novel topological structure of large-scale photovoltaic array and its MCU-level MPPT controlling method is proposed. Specifically, the large-scale array is decomposed into many photovoltaic MCUs at first, and each with a specific topological structure. Then, a bidirectional Cuk converter as the “photovoltaic module control device (PMCD)” is utilized to control every two MCUs in a series branch and a boost converter as the “branch voltage stabilization device (BVSD)” is utilized to control terminal voltage of each parallel branch. Finally, the MCU-level MPPT model is studied and formulized as a large-scale global optimization (LSGO) problem, and the AM-CCPSO algorithm in our previous work is applied to solve this problem. 2. The MCU-based topological structure of large-scale photovoltaic array In a large-scale photovoltaic array with thousands of modules, all the modules are divided into several parts at first, and the structure of each part is the conventional series and parallel connection of MCUs. Each MCU consists of several photovoltaic modules connected in a certain topology, and the structures of these MCUs are decided by the engineering situation and even can be different from each other. To control the operating point of each MCU independently, the bidirectional Cuk converter is introduced as the PMCD, and the boost converter is introduced as the BVSD in this new structure. Specifically, the bidirectional Cuk converter and photovoltaic module is connected as Fig. 1. Under uniform illumination, the two switches S1 and S2 keep open (off) and the converter does not work. Under partial shading conditions, e.g., PV2 is shaded but PV1 is not, S1-D2 can be activated by a PWM wave imposed on S1. As a result, the converter is equivalent to a unidirectional Cuk circuit (V1 is the input and V2 the output), in which power is transferred from the V1 terminal to the V2 terminal. The duty cycle of S1 can be controlled by a common proportional-integral (PI) controller, in which the reference values (control targets) are always set to the MPPs given by a certain MPPT algorithm. '&%XV L1

I

39

PV1

39

$QRWKHUEUDQFK

+ S1

C1

D1

V1

Cn S2

C2

D2 L2

bidirectional Cuk converter

V2

+ PV2

D

L

C1

S

C2

39Q Boost converter

-

I

Fig. 1. Structure of bidirectional Cuk converter based PMCD

39Q

$QRWKHUEUDQFK

Fig. 2. Structure of boost converter based BVSD

The boost converter based BVSD and photovoltaic branch is connected as Fig. 2. Each branch is connected in parallel through the BVSD, in which the terminal of each branch is the input of BVSD, and outputs of all the BVSDs are connected in parallel. With the application of PMCD and BVSD, the MCUlevel MPPT is applicable only if the theoretical MPP of each MCU is given to the corresponding device under certain environmental conditions. The structure of PMCD-BVSD-based photovoltaic array can be designed flexibly according to the real engineering applications. For example, when the system is equipped in a large ship, structure of each

115


MCU can be designed according to the ship’s layout and the illumination characteristics of the ship. The topology of large-scale photovoltaic array, and the structure of each MCU when the system is equipped in a large ship are shown as Fig. 3. (b)

(a)

0&8˖ 0&8˖

AC Bus DC Bus I

M C U

...

B V S D

M C U M C U

M C U

P M C D

M C U

P M C D

M C U

B V S D

M C U

M C U

P M C D

39

39

39

39

39

39

39

39

39

M C U

~

...

Inverter

39 39

Part II

P M C D

=

~

Inverter

...

M C U

P M C D

...

... M C U

39

39

= M C U

39

P M C D

=

~

...

0&8˖ 0&8˖ 39

Part I

Part N

39

39

39

39

Inverter

39

39

39

39

39

39

39

39

39

39

39

39

39

Fig. 3. (a) Topological structure of large-scale photovoltaic array based on PMCD and BVSD; (b) Different structures of each MCU

3. The LSGO-based MCU-level MPPT model A lot of intelligent algorithms are utilized to solve the MPPT problem, e.g., particle swarm optimization (PSO) [2] and genetic algorithm (GA) [3]. However, complexity of the problem increases exponentially as dimensions increase, and most of the algorithms fail to find the MPP in time as the scale of photovoltaic array becomes extremely large. Furthermore, in terms of the large-scale MCU-level MPPT, P-V (or P-I) curve of the shaded MCUs become multimodal under partial shading conditions, which will increase the complexity of MPPT further. In the following simulation experiments, an engineering analytical model of silicon solar cells proposed in reference [4] is utilized. Based on the model, to regard the current of each MCU as optimizing variable, and the output power of the entire system as fitness function value, the MCU-level MPPT problem can be transformed into a LSGO problem, which is formulized as min

1 i nMCU nsc

¦ ¦I C V i

i 1

j 1

s.t. I i [ I l I u ]

ij ij 2 OC

*log[

I 1 (1 i i ij ) 1] C1ij n pc * I SC

(1)

116


where nMCU denotes the number of MCUs. Ii (i =1, 2, …, nMCU) is current of the ith MCU, Il and Iu are the lower and upper bound of Ii. In general, Il is set to 0, and Iu is usually set to the short-circuit current of ij ij each MCU under extreme conditions. C1ij , C2ij , VOC and I SC denote the corresponding parameters: C1, C2, VOC and ISC of the photovoltaic module PVjt (j=1, 2,…, nsci ; t=1,2,…, nipc ) in the ith MCU [4]. nsci and nipc denote the modules in series and in parallel connections within the ith MCU. 4. Simulation and analysis 4.1. Comparison of the PMCD-BVSD based system and bypass- blocking-diode based system A small-scale photovoltaic array with 4 modules is simulated, and the bypass-blocking diode system is studied for comparison. Simulation results are shown in Table 1, in which system-diode denotes the bypass-blocking diode based system and system-device denotes the PMCD-BVSD based system. As shown in Table 1, under complex environmental conditions, the system-device outperforms the common system-diode evidently. Specifically, in the system-device, each module can operate around its own theoretical MPP, however, certain deviations exist in system-diode. As a result, the output power of system-device is 112.19W, which is exceeds 78.25W of system-diode significantly, and the power promotion is about 43.4%. Table 1. Comparison of the two systems under complex environmental conditions Irradiation (W/m2)

Operating point of system-diode (V)

Operating point of system-device (V)

Theoretical MPP (V)

S11=700

V11=18.19

V 11=15.27

V r11=15.28

S12=300

V 12=14.19

V 12=14.05

V r12=14.04

S 21=1000

V 21=18.46

V 21=15.78

V r21=15.78

S 22=500

V 22=13.92

V 22=14.77

V r22=14.78

Output power of system-diode (W)

Output power of system-device (W)

Power promotion

78.25

112.19

43.4%

4.2. Comparison of different MPPT methods In this section, the AM-CCPSO algorithm in our previous work is applied to solve the MCU-level MPPT problem of a large-scale photovoltaic array, and some other state-of-the-art algorithms are applied for comparison as well. The number of MCUs nMCU is set to 1000, and the structures of these MCUs are different. MCU 1 to 100: each contains 6 modules, 2 in parallel and 3 in series. MCU 101 to 500: each contains 10 modules, 2 in parallel and 5 in series. MCU 501 to 800: each contains 9 modules, 3 in parallel and 3 in series. MCU 801 to 1000: each contains 4 modules in series. Note that the entire photovoltaic array contains 8100 modules. The basic PSO with constriction coefficient is utilized for comparison, in which c1 and c2 are set to 2.05, and F = 0.729 [5]. In addition, some state-of-the-art algorithms are also utilized for comparison, including two PSO variants: CCPSO-SK [6] and CCPSO2 [7], and a DE variant: JADE [8]. Parameter settings of JADE are the same as its original paper, and parameters of CCPSO-SK and CCPSO2 are as follows: z CCPSO-SK: the number of subcomponents K is set to 10. z CCPSO2: the dynamic group size is set to S={20, 50, 100}. The parameters of photovoltaic module in reference [4] are set as: ISCref=4.515A, VOCref=44.852V, Imref=3.989A, and Vmref=36.895V. For each compared algorithm, the average results of 25 independent runs are recorded. Population size of all the algorithms is set to 15, and maximum number of fitness

117


evaluations (Max_FES) is set to 5.0e+05. In AM-CCPSO, the dynamic group size is set to S={20, 50, 100}, and the number of context vectors is set to 5. To consider the terrible environmental conditions, decision space of Ii (i=1,2, …, 1000) in Eq. (1) is set to [0 15] A. The complex environmental conditions are set as Table 2, and the theoretical MPP of the entire array is about 2.95378e+05W. The average results of 25 independent runs are shown as Table 3. Shown as Table 3, the MCU-level MPPT of a large-scale photovoltaic system can be solved by optimizing a large-scale multimodal problem with the application of an appropriate algorithm. Specifically, the PSO and CCPSO-SK have the worst performance with relative errors 48.49% and 21.18% respectively, which are far away from the global optimum. CCPSO2 and JADE obtain relative errors 3.13% and 14.24% respectively, which are much better than that of PSO and CCPSO-SK. However, the AM-CCPSO achieves the best performance with the relative error 1.02%, and can be regarded as a relatively more appropriate MPPT algorithm of the MCU-based large-scale photovoltaic system. Table 2. Settings of the complex environmental conditions MCU index

1~100

101~500

501~800

801~1000

Modules

Irradiation (W/m2)

Temperature (oC)

PV11 and PV12

1000

25

PV21 and PV22

800

22

PV31 and PV32

600

20

PV11 and PV12

200

16

PV21 and PV22

100

15

PV31 and PV32

800

22

PV41 and PV42

600

20

PV51 and PV52

400

18

PV11 , PV12 and PV13

400

18


200

16


100

15

PV1

1000

25

PV2

600

20

PV3

400

18

PV4

100

15

Table 3. Comparison of different methods on MCU-level MPPT

5. Conclusion

Algorithms

Result (W)

AM-CCPSO

2.92354e+05

Theoretical MPP (W)

Error (%) 1.02%

CCPSO2

2.86125e+05

CCPSO-SK

2.32811e+05

3.13%

JADE

2.53322e+05

14.24%

PSO

1.52149e+05

48.49%

2.95378e+05

21.18%

118


A new topological structure of large-scale photovoltaic array and its MPPT controlling method is proposed in this study. Firstly, the large-scale photovoltaic array is divided into a lot of parts and MCUs. Secondly, a bidirectional Cuk converter is utilized to control the operating point of each MCU, and a boost converter is utilized to control the terminal voltage of each branch. Thirdly, the MCU-level MPPT of a large-scale photovoltaic system is modelled as the LSGO problem, and some state-of-the-art largescale optimizing algorithms, especially the AM-CCPSO are applied to solve this MPPT problem. According to the simulation results, each MCU can operate around its reference value stably on the new structure. The AM-CCPSO is also successfully applied to the MCU-level MPPT problem of a largescale photovoltaic array with 1000 MCUs (8100 PV modules). In a word, with the application of the proposed topological structure and its MPPT controlling method, the large-scale photovoltaic system can avoid "hot spot" problem and achieve maximum output power under complex environmental conditions. References [1] Moreton, R., Lorenzo, E., Narvarte, L., 2015. Experimental observations on hot-spots and derived acceptance/rejection criteria. Solar Energy 118: 28-40. [2] Shi, J.Y., Zhang, W., 2014. MPPT for PV systems based on a dormant PSO algorithm. Electric Power Systems Research 123, 100- 107. [3] Shaiek, Y., Smida, M.B., Sakly, A., Mimouni, M.F., 2013. Comparison between conventional methods and GA approach for maximum power point tracking of shaded solar for PV generators. Solar Energy 90 (4), 107-122. [4] Su, J.H. et al., 2001. Investigation on engineering analytical model of silicon solar cells. Acta Energiae Solaris Sinica 22 (4), 409-412. [5] Eberhart, R.C., Shi, Y., 2000. Comparing inertia weights and constriction factors in particle swarm optimization. In: Proceedings of the IEEE Congress on Evolutionary Computation, pp. 84–89. [6] Van den Bergh, F., Engelbrecht, A., 2004. A cooperative approach to particle swarm optimization. IEEE Transactions on Evolutionary Computation 8 (3), 225-239. [7] Li, X.D., Yao, X., 2012. Cooperatively coevolving particle swarms for large scale optimization. IEEE Transactions on Evolutionary Computation 16 (2), 210-224. [8] Zhang J.Q., Sanderson A.C., 2009. JADE: Adaptive differential evolution with optional external archive. IEEE Trans Evol Comput, 13(5): 945-958.

Biography Dr. Ruo-li Tang is lecturer in School of Energy and Power Engineering, Wuhan University of Technology, Wuhan, China. He got his Ph.D. in 2016 from Wuhan University in China. He has published some papers in the area of intelligent algorithm and its applications in energy and power engineering.