Você está na página 1de 6

Distributed Reactive Power Control Methods to Avoid Voltage Rise in Grid-Connected Photovoltaic Power Generation Systems

Pau Mart , Manel Velasco, Josep M. Fuertes


Automatic Control Department Technical University of Catalonia Pau Gargallo 5, 08028 Barcelona, Spain Email: pau.marti@upc.edu

Antonio Camacho, Jaume Miret and Miguel Castilla


Department of Electronic Engineering Technical University of Catalonia Av. Victor Balaguer s/n, 08800 Vilanova i la Geltr u, Spain Email: antonio.camacho.santiago@upc.edu

AbstractA high density of grid-connected photovoltaic (PV) power generation systems is expected to occur in urban areas. The connection of distributed generation (DG) may cause serious problems in the distribution system such as voltage rise. Several voltage control strategies exist to avoid the over-voltage at the power grid. In particular, reactive power control methods have been proved to successfully bring DGs voltages within the admissible voltage range without reducing the production of active power. In addition, the possible use of communication infrastructure to link DGs opens up the development of new reactive power control methods where control actions at each DG may be decided using both local and global information. This paper presents a comparative study of existing and novel distributed reactive power control methods that have been designed exploiting the information exchange facility provided by a communication infrastructure.

I. I NTRODUCTION The rising penetration level of distributed generation poses new challenges in energy distribution [1][3]. For example, for grid-connected photovoltaic (PV) systems, when low power demand coincides with high solar radiation, voltage rise may occur, i.e. grid voltage can be outside of the statutory limits specied by the grid codes [4]. The over-voltage problem calls for a management scheme able to alleviate the excessive voltage increase by means of properly integrating real-time control and possibly information and communication technology [5][7]. Regarding the use of information and communication technology, simple voltage management methods use only local measurements at the point of common coupling (PCC) and do not require data transfer between DGs, e.g. [8][10]. Such decentralized methods may evolve to distributed or centralized approaches when information and communication technology is applied, and local actions may be based on both local and global measurements, e.g. [11][15]. In centralized control, voltage regulation may be carried out from a substation to the rest of the distribution line. The common deployment applies a master/slave model, potentially making an exhaustive use of the communication system and being the master a single point of failure. In distributed control, limited information exchange is enough

for successful voltage control and data exchange schemes can be structured using alternative communication models such as multimaster or producer-consumer models [16]. Within the existing control techniques to deal with the voltage rise problem (see e.g. [7], [17] for two condensed reviews on voltage control approaches in distributed generation), reactive power control methods have been proved to successfully bring DGs voltages within the admissible voltage range without reducing the active power of DGs. In residential neighborhoods with high PV penetration [18], PV inverters can be used for reactive power supply/absorption when the real power injection is less than the inverter rated power. By bringing together a distributed control strategy and the PV inverter capacity for reactive power control, this paper presents a comparative study of diverse distributed reactive power control methods. They have been inspired by the solution given by [12], where a dual control strategy composed by an internal local control loop and an external distributed control loop is presented. The comparative study analyzes the direct effects of each policy on reactive power absorption or injected currents at each DG, but also quanties the goodness of each method in terms of line power losses. The rest of this paper is structured as follows. Section II presents the distributed reactive power control methods. Section III presents the comparative study. And Section IV concludes the paper. II. D ISTRIBUTED C ONTROL S TRATEGIES Loosely speaking, many reactive power control policies designed to cope with voltage rise imply determining which reactive power set-points have to be given to each inverter in order to keep the line voltage within the specied range. The majority of grid codes specify that the voltage operation range should be within 0.85 and 1.10 per unit (p.u.) [4]. A. An Existing Distributed Voltage Control Strategy In [12] this is achieved by a dual control approach, which is illustrated in Figure 1. At each inverter, an internal control loop governed by a proportional/integral (PI) controller performs

Internal loop if Vi < 1.1 e=0 else e = 1.1 Vi e

Internal loop if Vi < 1.1 e=0 else e = 1.1 Vi e

Vi

PI

+ Qi +

Vi

PI

+ Qi +

External loop Qi Qj =i
1 n1

External loop Qmin PI Qi Qj =i


1 n1

MIN

AVG

Qavg

PI

Fig. 1: Existing distributed control strategy: Qmin


Internal loop

(a) Qavg policy

local voltage control. The per unit PCC voltage Vi is measured, and in the event of being above the statutory upper limit (1.1 p.u.), the control loop determines the reactive power to be absorbed in such a way that the voltage is brought at 1.1 p.u.. When the control goal is achieved (Vi 1.1), the internal loop is disconnected. Additionally, an external control loop is used for improving a balance among all of inverters. The balance is achieved by having all the inverters modifying their output reactive power to follow via a PI controller a set point which is dened to be the minimum reactive power value among all them. In terms of operation, and assuming n inverters in a distribution line topology, all inverters have to send at a given rate their reactive power value Qi measured at the PCC over the communication line. Each inverter, using the received values from the n 1 remaining inverters, selects the minimum value that is used for PI set-point control. According to [12], the dened balance helps to further reducing the voltage rise at the distribution line while enforcing all inverters to equally share the absorption of reactive power. Henceforth, this approach will be referred to as Qmin policy. It is worth noting that the underlying communication model that can be applied for implementing the approach by [12] does not demand a master/slave conguration. It can be implemented using a multimaster approach where each inverter periodically sends Qi . Within the available technologies for industrial communications, a standard Ethernet (IEEE 802.3 [19]) network could be the easiest candidate for implementing this approach even though it was not designed for critical control applications [20]. In fact, a preliminary case study for these type of distributed applications revealed that the impact of the non-determinism of Ethernet in the application performance is negligible because the external reactive power control does not require fast reaction time [21]. In addition, there are a number of techniques that can be used to adapt Ethernet for industrial processes to achieve the required real-time behavior [22]. B. Novel Distributed Voltage Control Strategies The novel policies presented in this paper put emphasis in the balance among inverters that should be achieved by the

Vi

if Vi < 1.1 e=0 else e = 1.1 Vi

PI

+ Qi +

External loop Ii Ij =i
1 n1

AVG

Iavg

PI

(b) Iavg policy

Fig. 2: Novel distributed control strategies

external control loop. The rst one, illustrated in Figure 2a, sets as a set-point for the external control loop at each inverter the average reactive power computed from the received n 1 Qi values rather than selecting the minimum. Therefore, the data exchange effort among inverters is the same as the Qmin policy, and the only difference is that each generator computes the average for PI set-point control. The expected goal is to avoid excessive reactive power absorption whenever voltages at the PCC have been already placed within the statutory voltage limits by the internal control loop. Henceforth, this policy is referred to as Qavg policy. Both Qmin and Qavg balance the effort performed by all the inverters in the absorption of reactive power. However, the injected current at each inverter may no be the same, and the difference may increase when the active power delivered by each inverter is not the same. With the goal of equalizing currents among inverters, a second policy named Iavg has been designed. It uses the external control loop to balance currents, as illustrated in Figure 2b. In this case, each inverter shares the current Ii that is injecting at the PCC by sending its value periodically. And each inverter, using the received n 1 currents, computes the average current Iavg , that is used for PI control set-point of the external control loop. This policy

grid L1 R1 L2 R2 L3 R3 L4 R4

Load 1 G1 Ethernet G2

Load 2 G3

Load 3 G4

Load 4

Fig. 3: Scheme of the PV system

guarantees that the real effort performed by all inverters is the same because all of them will inject the same current.
Voltage (p.u.)

1.2 1.1 1 0.9 0.8

C. A Non-distributed Voltage Control Policy In order to complete the comparative study, a decentralized policy is also analyzed. This policy is based on performing only local voltage control, that is, using only the internal control loop. Henceforth, this policy will be referred to as oil (only internal loop). This policy does not require data exchange between inverters. Therefore, it will help assessing among other issues whether it is worth deploying a communication infrastructure for distributed voltage control. III. C OMPARATIVE S TUDY In order to test the four policies presented in the previous section, a simple distributed PV system inspired in the example given in [6] is used. A. Distributed PV Model The distributed PV system under study is illustrated in Figure 3. Its main components are the grid, distribution lines in the form of resistance/inductance branches R1 L1 -R4 L4 , four inverters G1-G4 with constant power local loads connected in parallel, and Ethernet. Inverters are characterized by the active and reactive power that they will inject and/or absorb. Each inverter has an inner power control loop whose inputs are a) voltages of the three phases, b) active power set-point and c) reactive power set-point, and whose outputs are the three phase currents to be injected. Basic conguration for the distributed PV system shown in Figure 3 is as follows. The three-phase voltage source models both the grid and a medium-voltage/low-voltage transformer with amplitude 230 Vrms and frequency 50 Hz. Each distribution line between generators is modeled as a series connection of a resistance and an inductance, Ri Li with i = 1, . . . , 4, with 0.32 di and (0.082 di )/(2 50) H, where d1,2,3,4 = {0.3, 0.2, 0.1, 0.1} Km. All four loads are characterized by PL = 10 kW and QL = 5 kVAr. Unless otherwise stated, it is assumed that all PV panels are subject to the same solar radiation. Hence, all generators produce the same active power, P1,2,3,4 = 43 kW. Figure 4 shows the PCC voltage at each inverter without applying any control strategy (only active power Pi is injected). Two out of the four inverters, G3 and G4, lie above the upper

G1

G2 G3 Inverters no.

G4

Fig. 4: PCC voltages at each inverter without applying control

voltage limit specied in grid codes (1.1 p.u.). The second inverter, G2, is at 1.1 p.u. while the rst one, G1, is within the statutory limits, which are marked with two lines at 1.1 p.u. and 0.85 p.u.. B. Control and Communication Details The simulation of the four policies Qmin , Qavg , Iavg and oil requires specifying the control and communications parameters. All policies require tuning the internal PI controller gains. And the three rst policies also require tuning the external PI controller gains. The internal controller should be faster than the external controller, which can have an arbitrary long transient response. The internal PI gains have the same value for the four policies, which are Kp = 1000 and Ki = 10000. For the simulation purposes, the controller gains for the external PI of the Qmin , Qavg , and Iavg policies have been manually tuned to achieve the steady state value within the 20 seconds that takes each simulation. Their values are KpQmin = 1 and KiQmin = 10, KpQavg = 0.5 and KiQavg = 5, and KpIavg = 220 and KiIavg = 2200. It is worth noting that all policies have an output saturation for Qi , which is only used to avoid having any inverter injecting reactive power rather than absorbing it. Note also that it is assumed that inverters do not have any limitation in their capacity of absorbing reactive power. In terms of communication settings for the data exchange of the external control loop, Ethernet is congured at 10 Mbps and the message size is specied to be 512 bits because each message only carries the reactive power value Qi or

1.2 90 Current (Arms) Voltage (p.u.) 1.1 1 0.9 0.8 Qmin Qavg Iavg oil G1 G2 Inverter no. G3 G4 80 70 60 50 G1 Qmin Qavg Iavg oil

G2 G3 Inverters no.

G4

Fig. 5: PCC voltages at each inverter when applying the presented policies
Reactive power (VAR)

(a) Steady state currents at each inverter.


x 10 0
4

the current value Ii sent by each inverter. Note also that 512 bits is the minimum message size required for the collision detect mechanism [19]. Hence, the message transmission time is 50 s approximately. Since the whole Ethernet is only used for transmitting this message, it could be transmitted every 100 s, which would imply using a bit more than 50% of the network bandwidth. However, taking into account that the dynamics of the external control loop do not need to be fast, and considering also that Ethernet has an optimal behavior when the trafc occupies less than 30% of the bandwidth [23], the message transmission rate for each inverter in the Qmin , Qavg and Iavg policies have been nally set to 200 s. C. Main Simulation Result Figure 5 shows the PCC voltage of each inverter after applying the four policies. Remembering the voltage prole shown in Figure 4, the application of the presented policies successfully bring the four voltages within the statutory voltage limits. And no signicant difference can be appreciated among the values achieved by each policy. D. Analysis of the Steady State Values for I and Q In order to complete the analysis of the voltage results presented before, Figure 6 shows the steady state values for currents (sub-gure 6a) and absorbed reactive power (subgure 6b) for all policies at each inverter. By looking at these two sub-gures, it stands out that the behavior of the oil (only internal loop) policy is quite different with respect to the other three policies. In terms of reactive power (sub-gure 6b), only G3 and G4 absorb reactive power. This is a direct consequence of the oil policy operation, that only implements the internal control loop. Each inverter only regulates voltage whenever the measured voltage at the PCC is above the upper statutory limit 1.1 p.u.. And remembering the initial voltages shown in Figure 4, this only happens for these two inverters. As a consequence, the injected current at each inverter for the oil policy is signicantly different among inverters. In fact, G4 is injecting much more than the others. In addition, the Qmin , Qavg and Iavg policies present similar values in terms of currents and reactive power. Therefore, the objective of the external control loop, which was to achieve

2 4 6 8 G1 Qmin Qavg Iavg oil G2 G3 Inverters no. G4

(b) Steady state reactive power at each inverter.

Fig. 6: Ii and Qi when all inverters inject the same active power, P1,2,3,4 = 43kW

a balance among inverters, is met. Three main observations can be highlighted. First, the difference between both the currents and absorbed reactive power values for the three policies are small. Second, for the Qmin and Qavg policies, these values are equal. And third, the currents values of the Iavg policy lie between the upper and lower limit of the currents of the Qmin and Qavg policies. The numbers shown in Figure 6 are for the case that all inverters inject the same active power, which was specied to be P1,2,3,4 = 43kW. Henceforth, this case will be referred to as original scenario. In the following, two alternative scenarios are analyzed when the injected active power is not the same for all inverters. The rst scenario is characterized by having the rst inverter, G1, injecting P1 = 63kW rather than P1 = 43kW, while the rest remain injecting P2,3,4 = 43kW. And the second scenario is characterized by having the fourth inverter, G4, injecting P4 = 63kW rather than P4 = 43kW, while the rest remain injecting P1,2,3 = 43kW. This two alternative proles for active power injection may give more insight into the behavior of the policies under analysis. For the two new scenarios, Figures 7 and 8 show the steady state values for currents (sub-gures 7a and 8a respectively) and absorbed reactive power (sub-gure 7b and 8a respectively) for all policies. The rst observation that can be drawn is that the oil policy still shows a different pattern for Ii and Qi compared to the other three policies due to its specic design.

90 Current (Arms) 80 70 60 50 G1

Current (Arms)

Qmin Qavg Iavg oil

90 80 70 60 50 G1

Qmin Qavg Iavg oil

G2 G3 Inverters no.

G4

G2 G3 Inverters no.

G4

(a) Steady state currents at each inverter.


x 10 0 Reactive power (VAR) 2 4 6 8 G1 Qmin Qavg Iavg oil G2 G3 Inverters no. G4 Reactive power (VAR)
4

(a) Steady state currents at each inverter.


x 10 0 2 4 6 8 G1 Qmin Qavg Iavg oil G2 G3 Inverters no. G4
4

(b) Steady state reactive power at each inverter.

(b) Steady state reactive power at each inverter.

Fig. 7: Ii and Qi when all inverters inject the same active power P2,3,4 = 43kW except P1 = 63kW

Fig. 8: Ii and Qi when all inverters inject the same active power P1,2,3 = 43kW except P4 = 63kW

However, and being different from the original scenario shown in Figure 6, the new two scenarios reveal that currents and absorbed reactive power for the Qmin , Qavg and Iavg policies are no longer similar although each policy still achieves the balance that was targeting. By looking at the reactive power of the Qmin and Qavg policies in both new scenarios (sub-gure 7b and 8b respectively), it can be observed that the Qavg policy absorbs less reactive power than Qmin policy for bringing voltages within the statutory limits. This behavior corroborates its own design goal stated in sub-section II-B. The direct impact of this fact is that currents steady state values for the Qavg policy are smaller than those of the Qmin policy (sub-gures 7a and 8a). Focusing on the currents of the Iavg policy in both new scenarios (sub-gures 7a and 8a), it still holds the previous observation that the currents steady state values of the Iavg policy lie between the upper and lower limit of the currents of the Qmin and Qavg policies. Comparing currents among the three policies, the following observation holds. On one hand, when an inverter is supplying more active power than the others (G1 in sub-gure 7a or G4 in sub-gures 8a), the Iavg policy indicates that the inverter is injecting a smaller current than the case of the Qmin and Qavg policies. In other words, the Qmin and Qavg policies may force the inverter to inject a huge current. On the other hand, for the inverters that are not supplying any additional

active power (G2, G3, G4 in sub-gure 7a or G1, G2, G3 in sub-gures 8a), the Qmin and Qavg policies specify a lower current than for the case of the Iavg policy. Regarding currents, the latest observation suggests that each inverter could combine the application of different policies depending on the active power being supplied. For high active power supply, the Iavg policy seems more indicated while for low power supply, the Qavg policy should be used. E. Power Losses Analysis In order to complete the comparative study, the line losses for each policy in the three scenarios characterized by the different active power supply proles have been analyzed. Figure 9 summarized the numbers. By comparing the rst set of losses (group of bars in the left) corresponding to the original scenario with respect to the two remaining sets of losses (group of bars in the middle and in the right) corresponding to the rst and second scenario, it can be observed that as the supplied active power increases (rst and second scenario), line losses increase, as expected. In addition, the two sets of losses of the rst and second scenario show clear differences between policies that do not occur for the original scenario, where almost all policies imply the same losses. Hence, unbalanced active power supply among inverters (rst and second scenario) increases the differences in losses among the four policies.

8 6 Losses (W) 4

x 10

R EFERENCES
[1] Jenkins, N.; Allan, R.; Crossley, P.; Kirschen, D.; Strbac, G.; Embedded Generation, The Institution of Engineering and Technology, 2000. [2] Masters, C.L.; , Voltage rise: the big issue when connecting embedded generation to long 11 kV overhead lines, Power Engineering Journal , vol.16, no.1, pp.5-12, Feb. 2002 [3] OGorman, R.; Redfern, M.A.; , Voltage control problems on modern distribution systems, IEEE Power Engineering Society General Meeting, Vol.1, pp. 662- 667, 2004. [4] Photovoltaic Systems - Characteristics of the Utility Interface, Int. Electrotechn. Comm., Geneva, Switzerland, IEC61727, Dec. 2004 [5] International Energy Agency, Overcoming PV grid issues in urban areas, IEA Photovoltaic Power Systems Program, Task 10, Activity 3.3, Report IEA-PVPS T10-06, 2009 [6] Demirok, E.; Sera, D.; Teodorescu, R.; Rodriguez, P.; Borup, U.; , Clustered PV inverters in LV networks: An overview of impacts and comparison of voltage control strategies, Electrical Power Energy Conference (EPEC), 2009 IEEE , vol., no., pp.1-6, 22-23 Oct. 2009 [7] Tengku Hashim, T.J.; Mohamed, A.;, Shareef, H.; A review on voltage control methods for active distribution networks, Przglad Elektrotechniczny (Electrical Review), ISSN 0033-2097, R. 88 NR 6/2012 [8] Conti, S.; Greco, A. ; Messina, N. ; Raiti, S., Local voltage regulation in LV distribution networks with PV distributed generation, International Symposium on Power Electronics, Electrical Drives, Automation and Motion, 2006. [9] Braun, M.; Stetz, T.; Reimann, T.; Valov, B.; Arnold, G., Optimal reactive power supply in distribution networks - technological and economic assessment for PV-systems -, European Photovoltaic Solar Energy Conference and Exhibition, Sep. 2009 [10] Demirok, E.; Casado Gonzalez, P.; Frederiksen, K.H.B.; Sera, D.; Rodriguez, P.; Teodorescu, R.; , Local Reactive Power Control Methods for Overvoltage Prevention of Distributed Solar Inverters in Low-Voltage Grids, IEEE Journal of Photovoltaics, vol.1, no.2, pp.174-182, Oct. 2011 [11] Vovos, P.N.; Kiprakis, A.E.; Wallace, A.R.; Harrison, G.P.; , Centralized and Distributed Voltage Control: Impact on Distributed Generation Penetration, IEEE Trans. on Power Systems, vol.22, no.1, Feb. 2007 [12] Hojo, M.; Hatano, H.; Fuwa, Y.; , Voltage rise suppression by reactive power control with cooperating photovoltaic generation systems, Electricity Distribution - Part 2, 2009. CIRED 2009. The 20th International Conference and Exhibition on, June 2009 [13] Brenna, M.; De Berardinis, E.; Foiadelli, F.; Sapienza, G.; Zaninelli, D.; Voltage control in smart grids: an approach based on sensitivity theory, Journal of Electromagnetic Analysis and Applications, (2), 2010 [14] Kobayashi, H., Hatta, H., Reactive power control method between DG using ICT for proper voltage control of utility distribution system, 2011 IEEE Power and Energy Society General Meeting, 2011. [15] X. Liu, A. Aichhorn, L. Liu, and H. Li, Coordinated Control of Distributed Energy Storage System With Tap Changer Transformers for Voltage Rise Mitigation Under High Photovoltaic Penetration, IEEE Trans. Smart Grid, 2012 [16] Industrial Communication Systems, Editor(s): Bogdan M. Wilamowski and J. David Irwin, Auburn University, CRC Press, Feb. 2011 [17] Xu, T.; Taylor, P.; Voltage Control Techniques for Electrical Distribution Networks Including Distributed Generation, in Proceedings of the 17th IFAC World Congress, 2008. [18] Tonkoski, R.; Turcotte, D.; El-Fouly, T.H.M.; , Impact of High PV Penetration on Voltage Proles in Residential Neighborhoods, IEEE Transactions onSustainable Energy, vol.3, no.3, 2012 [19] 802.3 CSMA/CD Access Method and Physical Layer Specication, IEEE, 1985. [20] Decotignie, J.-d.; , The Many Faces of Industrial Ethernet [Past and Present], IEEE Industrial Electronics Magazine, vol.3, no.1, pp.8-19, March 2009 [21] Mart , P., Velasco, M., Castilla, M., Miret, J., Camacho, A., On the Use of Communication Infrastructure in Distributed Power Generation: a Preliminary Case Study, in 17th IEEE International Conference on Emerging Techonologies and Factory Automation, WiP Session, 2012. [22] Jasperneite, J.; Imtiaz, J. ; Schumacher, M. ; Weber, K., A Proposal for a Generic Real-Time Ethernet System, IEEE Transactions on Industrial Informatics, v.5 , n.2, 2009. [23] D. R. Boggs, J. C. Mogul, and C. A. Kent. Measured capacity of an Ethernet: myths and reality, In Symposium proceedings on Communications architectures and protocols, pp. 222-234, 1988.

Qmin 2 0 Qavg Iavg oil Original First Scenarios Second

Fig. 9: Line losses for each policy in the three scenarios characterized by different active power supply

Focusing the attention in the rst and second scenario (group of bars in the middle and in the right, respectively), the following conclusions can be drawn. For the rst scenario, the Iavg policy presents the worst losses. This could have been expected because the currents injected by the inverters (remember sub-gure 7a) for this policy are in general higher than the ones of the rest of policies. However, looking at the second scenario, this conclusion does not hold because Qmin and oil are the policies with worst losses and the best performance is achieved by the Iavg policy. Note that the currents injected by the inverters in the Iavg policy (remember sub-gure 8a) for the second scenario are still higher than the ones of the other policies, but the difference is not huge, and in G4, this tendency is broken, and currents for Qmin , Qavg and oil dramatically increase above the current given by Iavg . The latest observation corroborates the idea highlighted in the previous subsection that the best policy to be applied to deal with voltage rise could be a combination of Qavg and Iavg depending on the active power supply prole that applies. IV. C ONCLUSION This paper has presented two new distributed reactive power control policies to deal with the voltage rise problem. Their comparative study together with an existing distributed policy and a decentralized policy reveals that i) all four policies achieve their goal, which is bringing over-voltages within the statutory voltage limits, ii) the use of data exchange over a communication network permits to set additional control goals regarding the coordination among inverters such as balancing efforts, which is not possible for the decentralized policy, iii) it is difcult to identify a single distributed policy capable of delivering the best performance in different scenarios characterized by unbalanced active power supply patterns, and iv) the analysis of currents, absorbed reactive power and line losses indicates that a combination of the two presented new policies could be the best approach to cope with all scenarios. ACKNOWLEDGMENT This work was partially supported by projects CICYT DPI2010-18601 and ENE2012-37667-C02-02.

Você também pode gostar