230

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 1, JANUARY 2014

Optimal Decentralized Voltage Control for Distribution Systems With Inverter-Based Distributed Generators
Vito Calderaro, Member, IEEE, Gaspare Conio, Vincenzo Galdi, Member, IEEE, Giovanni Massa, Member, IEEE, and Antonio Piccolo, Member, IEEE
AbstractThe increasing penetration of distributed generation (DG) power plants into distribution networks (DNs) causes various issues concerning, e.g., stability, protection equipment, and voltage regulation. Thus, the necessity to develop proper control techniques to allow power delivery to customers in compliance with power quality and reliability standards (PQR) has become a relevant issue in recent years. This paper proposes an optimized distributed control approach based on DN sensitivity analysis and on decentralized reactive/active power regulation capable of maintaining voltage levels within regulatory limits and to offer ancillary services to the DN, such as voltage regulation. At the same time, it tries to minimize DN active power losses and the reactive power exchanged with the DN by the DG units. The validation of the proposed control technique has been conducted through a several number of simulations on a real MV Italian distribution system. Index TermsDistributed power generation, energy resources, photovoltaic systems, power distribution, power systems, reactive power control, voltage control, wind energy.

DG reactive power variations at the PCC. DG output voltage variations due to DG output voltage variations due to . DG converter output current. DN feeder current. MOGA Multi-objective genetic algorithm. Total number of DN branches. Total number of DGs. OR OSAB-DC PCC Operative range. Optimized SAB-DC. Point of common coupling. DG active power. DG maximum available active power. PF Power factor. Total real power losses objective on branch . PQR PV Power quality and reliability standards. Photovoltaic system. DG reactive power. constraint due to the converter current. Total amount of reactive DG PCC. at the th .

I. NOMENCLATURE AI AR CR DG DN EC Articial intelligence. Allowed range. Control range. Distributed generation power plant. Distribution network. European community. Down and upper threshold values for each DG. Maximum frequency. Active power losses objective function. Reactive power objective function. DG active power variations at the PCC.
Manuscript received January 31, 2013; revised May 26, 2013; accepted August 06, 2013. Date of publication September 17, 2013; date of current version December 16, 2013. Paper no. TPWRS-00132-2013. V. Calderaro, V. Galdi, G. Massa, and A. Piccolo are with the Department of Industrial Engineering, University of Salerno, 84084 Fisciano, Italy (e-mail: vcalderaro@unisa.it; vgaldi@unisa.it; gmassa@unisa.it; apiccolo@unisa.it). G. Conio is with Italian Vento Corporation Group, Naples 80122, Italy (e-mail: gaspare.conio@ivpc.com). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRS.2013.2280276

constraint due to the converter voltage. DG maximum reactive power capability. RES Renewable energy sources. Voltage versus active power sensitivity coefcient. Voltage versus reactive power sensitivity coefcient. SAB-DC Sensitivity analysis-based decentralized control.

0885-8950 2013 IEEE

CALDERARO et al.: OPTIMAL DECENTRALIZED VOLTAGE CONTROL FOR DISTRIBUTION SYSTEMS WITH INVERTER-BASED DISTRIBUTED GENERATORS

231

DG converter output voltage before Maximum DG converter output voltage before . DG output voltage at the PCC. Maximum allowed DG output voltage. Minimum allowed DG output voltage. DN feeder voltage. WDG Wind Distributed Generation unit.

DG transformer and grid lters reactance. DN reactance at the PCC. II. INTRODUCTION HE penetration of distributed generation power plants (DGs) into distribution networks (DNs) is rapidly increasing across the world in recent years. Various incentive programs have encouraged the adoption of renewable energy sources (RES)-based DGs in order to achieve the ambitious governments targets related to the promotion of a more sustainable development in the energy sector. For example, in 2009 the European Community (EC) has ofcially recognized [1] the need to promote RES as a priority measure. Furthermore, a signicant increase of RES units within the EC and through the world is expected in upcoming years [2], [3]. Nevertheless, the connection of DGs to DNs poses well-known technical challenges for the network management [4]. Among them, voltage rise represents one of the main issues concerning DG integration into DNs [5], especially in the case of nonprogrammable RES-DGs (e.g., wind- or solar-based ones) in weak DNs, characterized by a small value of reactance versus resistance ratio [6]. Thus, the necessity to develop proper control techniques to ensure power delivery to customers in compliance with PQR requirements and to furnish ancillary services to the network is becoming a relevant issue [7]. It is possible to distinguish two main categories through the various approaches proposed to face DGs integration issues: centralized and decentralized control strategies. The rst have been extensively investigated by the scientic community [8][10]. However, the application of centralized control strategies to the existing networks faces several drawbacks: in addition to the heavy investments necessary for devices and control systems, all centralized approaches require a highly reliable communication channel through the overall DN [11]. For these reasons, the second category is gaining relevant interest in recent years. The adoption of decentralized approaches allows the DGs to provide ancillary services to the DN, such as reserves and voltage support [12]. Moreover, it can have positive effects on both loss minimization and increased generation capacity because of their exibility [13]. Among decentralized approaches, the ones facing with DGs capability to provide reactive power support to the DN represent an emerging class of reactive dispatch technologies not yet extensively investigated. One of the most relevant features of

this type of approach is to implement the control action at the point of common coupling (PCC), allowing efcient, exible, and scalable control congurations [14]. An optimization technique aimed at minimizing power losses within the network and voltage deviation with respect to a reference signal is proposed in [15] for photovoltaic systems (PVs) and applied to the single feeder DN presented in [14]. A two-stages voltage control architecture for PV generation and Plug-in hybrid electric vehicles active load is proposed in [16]. The optimized control of the reactive power output from PV inverters to reduce voltage deviations with respect to a control reference is proposed in [17]. In [18] the Authors present an optimization strategy applied to a new algorithm for decentralized voltage control based on sensitivity analysis [19]. It is capable of maximizing active power production taking into account capability curves of wind DGs (WDGs). This paper presents the extension and completion of previous works [18] and [19]. The local voltage control strategy allows to obtain the maximum allowable active power production for each DG and tries to maximize active power production by controlling the DGs reactive/active power exchange with the DN and avoiding, as much as possible, the DG disconnection due to the infringement of voltage regulatory limits. Its main contributions and differences from the literature can be summarized as follows: 1) a modular approach that allows to implement valid local control actions even in absence of a widespread communication channel; 2) a unied strategy for RES management that permits to use the same local voltage control approach at the PCC for any type of DG connected to the grid by means of electronic converters; 3) an optimization procedure that brings together the proposed control technique and the optimization of its parameters used for voltage regulation; and 4) an articial intelligence (AI)-based optimization technique capable of avoiding local minima trapping due to the nonconvexity nature of the problem and to face with uncertainties in a better way compared with similar approaches presented within the literature [18]. The remainder of this paper is organized as follows. Section III contains the overall description of the reactive and active power control capabilities of inverter-based DGs, Section IV describes the sensitivity analysis based decentralized control (SAB-DC) method, and Section V presents the optimization approach applied to it in order to compute the optimized SAB-DC (OSAB-DC). Simulation results obtained by the OSAB-DC application to a real Italian distribution system are illustrated in Section VI, while Section VII contains nal remarks. III. REACTIVE/ACTIVE POWER CAPABILITY INVERTER-BASED DGS
OF

Typically, RES-DGs are connected to the DNs by means of electronic power converters. Among the decentralized control approaches, the capability of performing distributed control actions taking into account the converters reactive and/or active power controllable domains has recently been investigated. Such an approach could be useful to regulate voltage proles and/or to offer ancillary services to the DN, maximizing active power production at the same time [18][21]. In order

232

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 1, JANUARY 2014

Fig. 1. Control system structure.

to dene the characteristics of the overall methodology presented within the paper, the structure of the proposed control system applied to a generic schematic diagram of an inverter based RES-DG is depicted in Fig. 1. It represents the core of the OSAB-DC system and can be implemented on the on-board intelligence of each RES-DG plant connected to the DN by means of electronic converters. and are used by the OSAB-DC algorithm to perform the local control action. Concerning the inverter-based DG, the converter voltage depends on the dc-link voltage and the parameters of the adopted modulation technique. The reactance represents the total reactance of the transformers and grid lters used for the DG connection to the DN. Starting from the maximum power supplied by the primary source (e.g., PV array, wind turbine), the converter current and voltage limits impose circular constraints to the DG capability to vary and [21]. From the equations necessary to compute and it is possible to write (1) Taking into account the root mean square of the converter as follows: voltage

With an analogous approach, it is possible to prove that the circular constraint imposed by the converter current is (5) controllable domain In order to evaluate the of the DG it is necessary to take into account the converters and values. Thus, the reactive versus active power constraints can be written as [21]

(6) For each working point, the boundary of reactive power deviation available for the control action must be contained within the capability curve dened by (7) Fig. 2 shows the capability curves for different PF values, % 1.05 p.u., 0.95 whereas p.u., and 1.01 p.u, [21]. To better explain the meaning of the curve it is possible to say that, e.g., for each value of the available and each value of the admitted PF, there are upper and lower limits to the capability of varying . The upper and lower points on the curve in the gure represent these limits. Since the controllable domain equations mainly concern the grid-side converter model, this approach can be used for either full power back-to-back converter-connected WDGs and single- or double-stage inverter-connected PVs. IV. SAB-DC The proposed controller performs a SAB-DC strategy by local control of reactive and/or active power exchange between each DG and the DN at the DG connection bus. Sensitivity analysis can be very useful to evaluate the relationships among voltage proles changes and values necessary to implement a correct control action. In fact, sensitivity coefcients give information concerning the variation in the output of a model can be apportioned, qualitatively

(2) and, working on the equations, the following can be stated:

(3) is signifSince in almost all power systems applications icantly greater than the DN feeder inductance , it is possible to compute the rst circular constraint due to the converter voltage , as stated by the following equation [22]: (4)

CALDERARO et al.: OPTIMAL DECENTRALIZED VOLTAGE CONTROL FOR DISTRIBUTION SYSTEMS WITH INVERTER-BASED DISTRIBUTED GENERATORS

233

Fig. 3. Allowed, operative, and control ranges for the SAB-DC.

Fig. 2. DG capability curves.

and quantitatively, to different sources of variation. Thus, they are very useful to predict or to face losses, bus voltages, and branch ow variations due to changes in loads and/or generation proles [23], [24]. Within the paper, the sensitivity coefcients are computed only for the DN buses where DGs are connected. The evaluation of voltage versus reactive power sensitivity coefcient is implemented by xing all DN parameters, including the load and the generation proles. At each time step, only the reactive power of each bus where the sensitivity parameters must be calculated is varied in turn, as stated in [10]. The voltages at each bus are recorded at each step, and then the sensitivity coefcients are computed. An analogous procedure is used to calculate the voltage versus active power sensitivity coefcients . Thus, if time step represents the system measurement time step, for each DG, it is possible to state that (8) Taking into account the relationships among voltage variations and reactive/active power variations stated in (8), the proposed SAB-DC can be dened. It is focused on maintaining voltage levels within regulatory limits and avoiding, as much as possible, the DG disconnection. Its peculiarity is that the control action starts only if the voltage reaches warning values, dened by the introduction of two threshold levels placed around the nominal voltage value (1 p.u.). Two operational ranges are dened within the allowed range (AR) : the operative range (OR) , and the control range (CR) , as depicted in Fig. 3. The control operations begin only if the DG bus voltage level enters the CR, causing a violation of the OR limits. More precisely, when the voltage value enters the CR, a certain amount of reactive and/or active power is injected/absorbed proportionally to the difference between the voltage value within the AR and the threshold value placed between the CR and the OR. The proportionality terms are represented by the sensitivity coefcients as in (8). To better explain the operation of the SAB-DC,

the control algorithm ow chart in the case of upper threshold violation is shown in Fig. 4. At the generic time instant , the procedure begins evaluating the difference between and the threshold value , as follows: (9) The control strategy considers injection/absorption by related to (10) If reactive power injection is contained within the capability region of the DG converter and the voltage value is in the upper CR, the control system tries rst to compensate the entire amount of voltage variation by increasing reactive power injected into the DN. Thus, the amount of to be varied is computed as (11) used to bring back The maximum amount of voltage levels within the OR is limited by the DG capability coverage . For each working point, represents the maximum value of reactive power that the converter is able to absorb and/or inject into the DN. Therefore, the DG reactive power is chosen according to (7) to yield

(12) is not capable of maintaining the Then, only if voltage within the OR, an active power curtailment has to take place. The necessary amount is evaluated according to (8) and (10) as follows: (13) This active power curtailment allows to decrease voltage levels and to obtain a higher reactive power injection capability because the working point is moved leftmost on the capability curve. The right part of the owchart depicts the SAB-DC in the case of DG bus voltage contained within the OR. If a control action has been previously performed, the SAB-DC tries to recover at its maximum available

234

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 1, JANUARY 2014

Fig. 4. SAB-DC owchart.

value the DN. Thus,

and to reduce the exchanged with is modied according to

(14) is below the active power Furthermore, if amount proportional to , the residual voltage variation is used to reduce reactive power injection into the grid

The algorithm acts in a similar manner in the case of bottom OR violation. The main difference is that, in this case, only the reactive loop is available to perform the control action. In fact, it is not possible to increase active power more than its maximum available value. V. CR DEFINITION: OPTIMIZATION PROBLEM FORMULATION AND SOLUTION ALGORITHM The main issue of the SAB-DC concerns the denition of the optimal threshold values to adopt for each DG, where represents the DG index. For a proper approach, it is fundamental to consider that the value of the thresholds determines the starting point of the control action, affecting the amount of and exchanged by the DG with the DN. In fact, the SAB-DC actions take place evaluating the difference between the measured voltage and the CR/OR threshold (or ). Thus, real power losses through the DN are inuenced by the control action as well. An optimization approach nalized to the computation of the optimal thresholds is dened within the paper to prove the general validity of the SAB-DC method and to allow its applicability to every type of DN. In order to nd optimal solutions, AI-based optimization techniques can be applied. The most relevant advantages of these techniques are to not require differentiability and continuity of the objective function and to reduce the risk to be trapped in local minima compared with classical optimization strategies. Furthermore, these methods have been

(15) The reactive power injection is decreased proportionally to the entire amount of if no active power curtailment has occurred during previous time step and is shown as follows: (16) As depicted within this section, the proposed control system is characterized by two internal loops. The rst one acts by varying the reactive power up to the capability limits in order to control voltage levels, avoiding active power curtailments. If the reactive control action is not sufcient to maintain voltage levels within the AR, the second stage of the control algorithm takes place, reducing active power injection. The DG is disconnected only if neither reactive and active power control actions are able to maintain voltage levels within the AR.

CALDERARO et al.: OPTIMAL DECENTRALIZED VOLTAGE CONTROL FOR DISTRIBUTION SYSTEMS WITH INVERTER-BASED DISTRIBUTED GENERATORS

235

proved to be very suitable to solve nonlinear nonconvex problems [25], [26] like the one addressed within the paper. In fact, AI-based approaches have been proved to be more useful [18] of other ones in which the threshold estimation was based on uncertainties evaluations [27]. To drive an effective optimization procedure it is fundamental to obtain threshold values by applying them to a sample composed by variable generation and load demand proles, assuring the coverage of a wide range of working conditions. In particular, DGs thresholds are computed by solving a multi-objective optimization problem based on the denition of two objective functions: reactive power objective : it tends to control the reactive power ow through the DN. It could be useful to improve voltage proles, to allow high PF levels and to reduce power losses; power losses objective : it could help to increase system efciency and decrease power generation cost. More precisely, the objective functions are computed as follows: (17) (18) The optimization problem is implemented on an overall time interval . It consists of optimal thresholds computation, aimed at minimizing the two objective functions. Thus, the overall optimization problem formulation results in (19) (20) where (20) denes the threshold matrix containing the threshold values of each DG controller. An approach based on multi-objective genetic algorithm (MOGA) interacting with a daily power-ow routine is performed to solve the optimization problem (19). The power ow routine is applied to daily generation and load demand proles in order to maintain the problem formulation as much general as possible. Typical MOGA operators like elitism, selection, crossover, and mutation are taken into account. Several simulations with different MOGA operators are conducted in order to prove the validity of the proposed approach. For each power-ow routine iteration, the optimization problem is subject to the following system operating constraints. Constraints applied to voltage levels, power factor, and reactive and active power capabilities of DGs: (21) (22) (23) (24)

Thresholds constraints:

(25) Distribution line limits related to the maximum transmittable real power between nodes and : (26)

VI. CASE STUDY SIMULATIONS AND RESULTS In order to prove the validity of the proposed OSAB-DC, a real Italian radial DN has been used. The single-line diagram of the DN is depicted in Fig. 5. It consists of a 20-kV distribution system fed by a 132-kV, 50-Hz subtransmission system with a short-circuit level of 750 MVA through a 150/20-kV /Yg transformer with rated power 25 MVA, %, and [28]. The primary substation transformers tap is xed to 1.006 p.u., as one of the two classical strategies used in Italian distribution systems [20]. The sensitivity coefcients have been evaluated according to the procedure stated in Section IV. Due to the ofine procedure, several computations have been performed for various generation and loads values. Even if a very relevant change among the sensitivity coefcients over the various computations is not observed, those calculated for the worst case scenario have been chosen for the SAB-DC optimization phase. At the beginning of the simulations, the power factor has been set to 1. The maximum voltage variation of % around the nominal value has been imposed as stated in [29]. The application of the optimization strategy to the network has been simulated running the MOGA in interaction with a power-ow routine based on daily generation and load forecasting, computed on a 10-min base in order to assure assessment of the state changes. Two different DGs penetration scenarios applied to the DN under test have been supposed in order to obtain a general validation of the proposed strategy. In Scenario A, four WDGs have been connected to the DN. In Scenario B, four PVs have been added to the previously dened scenario. The connection buses are highlighted in Fig. 5 and Table I, where the rated power values of the DGs are specied. The same rated power has been supposed for the WDGs, while two different congurations have been used for the PVs due to the modular characteristic of PV plants. Fig. 6 shows the generation power proles for WDGs and PVs and load demand for three different load types: residential, commercial, and industrial sectors. Without loss of generality and due to the upper OR infringement shown in Fig. 7 for both scenarios, the optimization strategy has been run to calculate only the upper thresholds. Thus, several simulations of the optimization strategy have been computed varying the MOGA parameters for each scenario, in order to investigate the assessment of the optimal solution.

236

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 1, JANUARY 2014

Fig. 5. Italian radial DN under test.

TABLE I DGS RATED POWER AND PCC

Table II shows the intervals adopted for the various MOGA parameters aimed at validating the proposed methodology and the ones capable of assuring the optimal convergence of optimization procedure. As shown in the table, they correspond to ranges that assure a variability of the MOGA evolution that is not too excessive while preventing the premature ending of the simulation due to a minimum trap. Fig. 8 depicts the Pareto fronts of the optimal solution for both scenarios. It represents the tradeoff between the two conicting objective functions. Each one of the points within the fronts represents a valid solution for the optimization problem. The controlled voltage values obtained using the solution that minimizes the reactive power objective function are shown in Fig. 9. It is worth noting that, in either scenario, the voltage values are contained within regulatory limits. Fig. 10 shows the reactive power injection into the DN by the DGs participating in the control actions. Scenario A concerns only the reactive power support from WDG connected at Bus 54, while Scenario B concerns the two WDGs connected at Buses 46 and 54 and the PV at bus 47. These DGs are those that enter the CR and infringe the voltage constraints in Fig. 7. It is worth noting a sort of sharing of the control action. In fact, the reactive power from WDG connected at bus 46 tends

Fig. 6. Generation power proles and load demand.

to compensate the feeder voltage boosting due to the end of the regulation action from WDG at bus 54 around the seventh operation hour. Fig. 11 shows the capability coverage of the proposed OSAB-DC for the DGs involved with the control action. The capability curve is computed from the equations in Section II. It represents the limit to the converters capability

CALDERARO et al.: OPTIMAL DECENTRALIZED VOLTAGE CONTROL FOR DISTRIBUTION SYSTEMS WITH INVERTER-BASED DISTRIBUTED GENERATORS

237

Fig. 7. DGs bus voltages without control action.

Fig. 8. Pareto fronts of the optimal solution.

TABLE II MOGA VARIED PARAMETERS AMONG THE VARIOUS TESTS

to vary with respect to the available and the allowed PF range. The crosses in Fig. 11, Scenario A, represent the controller working points at each time step within the daily simulation interval. Similarly, the points in Fig. 11, Scenario B, show the daily working points of each one of the OSAB-DC controllers that infringe the OR limits giving rise to the reactive/active control action. It can be stated that the optimized control action does not require any curtailment of the available active power because of the absence of any capability curve constraint infringement. Thus, maximum available active power delivery is allowed by the proposed control strategy at a local level.

Fig. 9. Controlled voltage proles.

Furthermore, a different wind generation power prole has been applied to the WDG controllers optimized for the rst case

238

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 1, JANUARY 2014

Fig. 12. Modied wind generation prole for Scenario B.

Fig. 10. Reactive power sharing among DGs.

Fig. 13. DGs voltage proles (a) Without control action enabled (b) With control action enabled.

Fig. 11. Capability coverage of controlling DGs.

study in Scenario B in order to assess the robustness of the proposed solution, e.g., to unpredicted forecasting errors. Fig. 12 depicts the generation power proles. Load demands have been maintained the same as in the previous example. Fig. 13 shows the worsening of the uncontrolled voltage proles and the correct action performed by the OSAB-DC. As depicted in Fig. 14, the control action is maintained within the capability coverage of the DGs even in this case. Obviously, the PF drops below 0.95 (approximately 0.948). This value has been considered to be acceptable due to the PF Italian requirements for DGs connected by means of electronic converters (to be maintained within the range ). Taking into account more restrictive PF

CALDERARO et al.: OPTIMAL DECENTRALIZED VOLTAGE CONTROL FOR DISTRIBUTION SYSTEMS WITH INVERTER-BASED DISTRIBUTED GENERATORS

239

Fig. 14. Capability coverage for modied wind proles.

Fig. 16. DGs bus voltages without control action.

Fig. 15. Generation power proles and load demand for Winter (Wt) and Summer (Smr) Weekdays.

regulations, the control action necessary to avoid DGs disconnection could be performed by a curtailment of active power production. Considering that Scenario B represents the worst case among the two scenarios introduced previously, two further optimization procedures are presented for this DGs integration scenario. Two different sets of generation and load proles depicted in Fig. 15, concerning a Winter Weekday and a Summer Weekday, have been considered in order to prove the advantages of the proposed optimization methodology. Fig. 16 shows the uncontrolled voltage proles at the DGs PCC for the two case studies. It is worth noting that the optimization procedure conducts to the calculation of thresholds capable of performing a correct control action for either of the case studies, as depicted in Fig. 17. Fig. 18 shows the reactive power contribution to the control action by each DG. For each case study, it is worth noting that the control action is performed by the DGs that infringe the AR, just as it did for the

Fig. 17. Controlled voltage proles.

previous scenarios. Either the Winter and Summer Weekdays scenarios allow to perform a control action that avoid the DGs disconnection while maximizing active power production. It is

240

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 1, JANUARY 2014

ages communications. No continuously detailed information is required to implement the correct control action. The OSAB-DC is well suited for large DG penetration cases due to its local nature that reduces the complexity of the control system compared with a completely centralized approach. It is characterized by a unied and modular conguration, and its simplicity and effectiveness would allow its practical implementation. In fact, its extension due to new DGs connections only requires, for example, that the DSO adds the capability specications to the constraints section and two variables per each DG to the chromosome that represents the optimal solution to nd. By run the MOGA on the DN model it would be possible to evaluate the optimal thresholds. Then, the online application of the OSAB-DC would be implemented by sensing voltage levels at the DG connection bus. These features allow the OSAB-DC to be suitable for smart grid development scenarios because of its intrinsic modular and scalable architecture. APPENDIX The MOGA used within the paper is implemented as a controlled elitist genetic algorithm. It is a variant of the Nondominated Sorting Genetic Algorithm II. Due to its elitist nature, it always tends to prefer individuals with better tness value. It also controls the elite members of the population as the algorithm progresses in order to support the individuals that can help to increase the diversity of the population necessary to converge to an optimal Pareto front. The MOGA has been run several times on the validation scenarios during the validation phase of the proposed control methodology. These computations have been conducted by varying the MOGA parameters within the ranges indicated in the test range column in Table II. It has been noted that the Pareto front found by the MOGA tended to the same convergence region if the setting parameters were contained within the optimal range presented in Table II in each one of the validation scenarios. Thus, this parameter range represents the optimal set to assure a global convergence of the optimization procedure. REFERENCES
[1] Directive 2009/28/EC of the European Parliament and of the Council, Ofcial J. Eur. Union, pp. 1662, 2009. [2] Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, Energy roadmap 2050, in Proc. COM, 2011, pp. 120. [3] Int. Energy Agency, World Energy Outlook 2012, Executive Summary, 2012, pp. 19. [4] R. A. Walling, R. Saint, R. C. Dugan, J. Burke, and L. A. Kojovic, Summary of distributed resources impact on power delivery systems, IEEE Trans. Power Del., vol. 23, no. 3, pp. 16361644, Jul. 2008. [5] C. Masters, Voltage rise the big issue when connecting embedded generation to long overhead 11 kV lines, Power Eng. J., vol. 16, no. 1, pp. 512, 2002. [6] A. E. Kiprakis and A. R. Wallace, Maximising energy capture from distributed generators in weak networks, Proc. Inst. Elect. Eng., Gen., Transm., Distrib., vol. 151, no. 5, pp. 611618, 2004. [7] E. J. Coster, J. M. A. Myrzik, B. Kruimer, and W. L. Kling, Integration issues of distributed generation in distribution grids, Proc. IEEE, vol. 99, no. 1, pp. 2839, Jan. 2011. [8] A. Casavola, G. Franz, D. Menniti, and N. Sorrentino, Voltage regulation in distribution networks in the presence of distributed generation: A voltage set-point reconguration approach, Electr. Power Syst. Res., vol. 81, pp. 2534, 2011.

Fig. 18. Reactive power sharing among DGs.

important to point out that these control actions also take place without any active power curtailment. VII. DISCUSSION AND CONCLUSION This paper has presented an optimization strategy applied to an SAB-DC for RES DGs, able to maintain voltage levels within regulatory limits while producing the maximum available active power. It avoids the DG disconnection due to violation of voltage limits at the connection bus and allows the DGs to offer ancillary services to the DN. A unied approach to the controller modeling in the case of RES DGs connected to the DN by means of electronic converters and the validity of the optimization strategy have been proposed and proved through several simulations run with different DGs penetration levels. All of the solutions presented by the Pareto fronts are valid thresholds for the optimization problem. The choice of one of them represents the optimal tradeoff between active losses and reactive power minimization. It could be useful for the distribution system operator (DSO) in order to determine the amount of the independent power producer (IPP) participation to the ancillary service operations. Thus, it allows to face the conicting interests between DN losses minimization or reactive power action reduction aimed at maximum power production by the IPP. Results obtained by the OSAB-DC application to different DGs scenarios have been shown. Its robustness with respect to unpredicted changes in generation power proles has been proved through the simulations as well. This property allows reducing the communication channel reliability requirements that mainly concern only the optimized thresholds sharing and out-

CALDERARO et al.: OPTIMAL DECENTRALIZED VOLTAGE CONTROL FOR DISTRIBUTION SYSTEMS WITH INVERTER-BASED DISTRIBUTED GENERATORS

241

[9] M. S. P. Carvalho, P. F. Correia, and Lu s A. F. M. Ferreira, Distributed reactive power generation control for voltage rise mitigation in distribution networks, IEEE Trans. Power Syst., vol. 23, no. 2, pp. 766772, May 2008. [10] A. Keane, L. F. Ochoa, E. Vittal, C. J. Dent, and G. P. Harrison, Enhanced utilization of voltage control resources with distributed generation, IEEE Trans. Power Syst., vol. 26, no. 1, pp. 252260, Feb. 2011. [11] H. E. Farag and E. F. El-Saadany, A novel cooperative protocol for distributed voltage control in active distribution systems, IEEE Trans. Power Syst., vol. 28, no. 2, pp. 16451656, May 2013. [12] A. G. Madureira and J. A. Pecas Lopes, Coordinated voltage support in distribution networks with distributed generation and microgrids, IET Renew. Power Gener., vol. 3, no. 4, pp. 439454, 2009. [13] L. F. Ochoa and G. P. Harrison, Minimizing energy losses: Optimal accommodation and smart operation of renewable distributed generation, IEEE Trans. Power Syst., vol. 26, no. 1, pp. 198205, Feb. 2011. [14] K. Turitsyn, P. Sulc, S. Backhaus, and M. Chertkov, Options for control of reactive power by distributed photovoltaic generators, Proc. IEEE, vol. 99, no. 6, pp. 10631073, Jun. 2011. [15] H.-G. Yeh, D. F. Gayme, and S. H. Low, Adaptive VAR control for distribution circuits with photovoltaic generators, IEEE Trans. Power Syst., vol. 27, no. 3, pp. 16561663, Aug. 2012. [16] B. A. Robbins, C. N. Hadjicostis, and A. D. Dom nguez-Garc a, A two-stage distributed architecture for voltage control in power distribution systems, IEEE Trans. Power Syst., vol. 28, no. 2, pp. 14701482, May 2013. [17] K. Tanaka, M. Oshiro, S. Toma, A. Yona, T. Senjyu, T. Funabashi, and C. H. Kim, Decentralised control of voltage in distribution systems by distributed generators, IET Gener. Transm. Distrib., vol. 4, no. 11, pp. 12511260, 2010. [18] V. Calderaro, V. Galdi, G. Massa, and A. Piccolo, Distributed generation management: An optimal sensitivity approach for decentralized power control, in Proc. 3rd Conf. Exhib. Innovative Smart Grid Technol., 2012, pp. 18. [19] V. Calderaro, V. Galdi, G. Massa, and A. Piccolo, Distributed generation and local voltage regulation: An approach based on sensitivity analysis, in Proc. 2nd Conf. Exhib. Innovative Smart Grid Technol., 2011, pp. 18. [20] V. Calderaro, G. Conio, V. Galdi, and A. Piccolo, Reactive power control for improving voltage proles: A comparison between two decentralized approaches, Electr. Power Syst. Res., vol. 83, no. 1, pp. 247254, 2012. [21] N. R. Ullah, K. Bhattacharya, and T. Thiringer, Wind farms as reactive power ancillary service providersTechnical and economic issues, IEEE Trans. Energy Conv., vol. 24, no. 3, pp. 661672, Sep. 2009. [22] M. Ohshima and E. Masada, The P,Q controllable domain of a single phase PWM converter to preserve sinusoidal AC current waveform, IEEE Trans. Power Electron., vol. 15, no. 3, pp. 485494, . [23] D. K. Khatod, V. Pant, and J. Sharma, A novel approach for sensitivity calculations in the radial distribution system, IEEE Trans. Power Del., vol. 21, no. 4, pp. 20482057, Oct. 2006. [24] H. M. Ayres, W. Freitas, M. C. De Almeida, and L. C. P. Da Silva, Method for determining the maximum allowable penetration level of distributed generation without steady-state voltage violations, IET Gener. Transm. Distibr., vol. 4, no. 34, pp. 495508, 2010. [25] J. G. Vlachogiannis and K. Y. Lee, A comparative study on particle swarm optimization for optimal steady-state performance of power systems, IEEE Trans. Power Syst., vol. 21, no. 4, pp. 17181728, Nov. 2006. [26] G. K. Viswanadha Raju and P. R. Bijwe, Reactive power/voltage control in distribution systems under uncertain environment, IET Gener. Transm. Distrib., vol. 2, no. 5, pp. 752763, 2008. [27] F. Bignucolo, R. Caldon, and V. Prandoni, Radial MV network voltage regulation with distribution management system coordinated controller, Electr. Power Syst. Res., vol. 78, pp. 634645, 2008. [28] S. Conti and A. M. Greco, Innovative voltage regulation method for distribution networks with distributed generation, in Proc. Int Conf. Electricity Distrib., 2007, pp. 14. [29] IEEE Application Guide for IEEE Std. 1547, IEEE Standard for Interconnecting Distributed Resources With Electric Power Systems, 1547.2-2008, 2009, pp. 1-207.

Vito Calderaro (M10) received the M.Sc. degree in electronic engineering and Ph.D. degree in information engineering from the University of Salerno, Fisciano, Italy, in 2001 and 2006, respectively. He is currently a Research Assistant with the Department of Industrial Engineering, University of Salerno, Fisciano, Italy. His current research focuses on integration of distributed generation systems on electrical distribution networks, soft computing methodologies in power system applications, and protection and diagnostic of complex systems. He has served as reviewer for many international journals and conferences.

Gaspare Conio received the M.Sc. degree in electronic engineering from the University of Salerno, Fisciano, Italy, in 2008. Since 2009, he has worked in cooperation with the University of Salerno on grid integration of wind power plants. In 2010, he joined the Italian Vento Power Corporation Group, Naples, Italy, where he is involved with the development, operation, and maintenance of large wind farms. His main research activities concern the grid integration of renewable energy power plants, power systems control, and energy storage.

Vincenzo Galdi (M98) received the M.Sc. degree in electronic engineering from the University of Salerno, Fisciano, Italy, in 1994, and the Ph.D. degree in electrical engineering from the University of Naples, Naples, Italy, in 1999. Since 2003, he has been an Associate Professor with the University of Salerno, Fisciano, Italy, now with the Department of Industrial Engineering. His current research interests include application of ICT to power systems and to power management and energy efciency in transportation systems, smart grids, and ITS.

Giovanni Massa (M12) received the M.Sc. degree in electronic engineering and Ph.D. degree in information engineering from the University of Salerno, Fisciano, Italy, in 2009 and 2013, respectively. His main research interests focus on smart-grid development issues, including integration of distributed generation systems on electrical distribution networks, soft computing methodologies in power system applications, energy saving in electric transportation, and Intelligent Transportation Systems.

Antonio Piccolo (M06) received the M.Sc. degree in electrical engineering from the University of Naples, Naples, Italy, in 1974. Since 1990, he has been a Full Professor with the University of Salerno, Fisciano, Italy, currently as a founding member of the Department of Industrial Engineering. His main elds of interest are the application of ICT to electric energy infrastructures and transportation systems. He has served as a reviewer and session chairman for many international conferences. He has authored or coauthored more than 130 papers mainly in international journals and conferences in the eld of power systems. Prof. Piccolo is a Chartered Engineer.