Application of Active Management in Operation of Distribution Network ...

Application of Active Management in Operation of Distribution Network Using NSGA II Saeed Abapour Faculty of Electrical and Computer Engineering, University of Tabriz Tabriz, IRAN sa.abapour@gmail.com

Kazem Zare Faculty of Electrical and Computer Engineering, University of Tabriz Tabriz, IRAN Kazem.zare@tabrizu.ac.ir

Abstract— One of the main aims of distribution networks is increasing the participation of distributed generation. One of the methods that can determine the maximum penetration of DG in distribution network is use of active network management. Active network management has essential role to facilitate the connection of new generations without need of reinforcement traditional network. This paper proposes a multi- objective idea based on minimizing costs of active and reactive power of DGs and network. In addition, the mentioned objective function is includes cost of active power losses and reactive power generated (absorbed) by Reactive Power Compensators (RPC). The proposed objective function is determined based on optimal power flow and with considering the technical constraints. For active management of distribution network, coordinated control of voltage and reactive power of network equipment are implemented. Non-dominated Sorting Genetic Algorithm (NSGA II) is employed to solve the optimization problem.

I. INTRODUCTION For increasing the producing electric energy in many countries, distributed generation (DG) technology is used. However, increasing the penetration of DG encounters challenge for planning and operating of distribution networks [1]. Distribution Network Operators (DNOs) regardless of whether DG is connected to the network or not, continues operation of their network distribution with previous method. This operation scheme is consistent with passive network management so that capacity of DG connected to distribution networks not is limited by the DNO [2]. In most of cases, distribution networks are being operated radially. One of considered issues in radial networks is active power losses. Penetration of DG in appropriate range in the network reduces power losses. But if maximum available capacity of DG is utilized, it may enhance network active power losses. Active Management (AM) is an efficient method to network reinforcement for operation and connection of the DG. Therefore, applying the active management along with increase of the penetration of DG can help reduce the losses, modify the peak load, control the voltage, enhance the power quality, decrease the DG curtailment and also postpone the reinforcement of the distribution and transmission network. The proposed objective function is placed based on minimizing the costs. In this function, cost of the energy

Mehdi Abapour Faculty of Electrical and Computer Engineering, University of Tabriz Tabriz, IRAN abapour@tabrizu.ac.ir

supplied by transmission network and plus cost of energy generated by DGs and RPCs are calculated. Through executing the AM programs for optimal operation of distribution networks will dramatically reduce the power losses cost and the defined cost function. Also due to increasing the penetration of DG units in the active network, total cost of energy is reduced and the profit of DG units is increased. For the aim of comprehensive review, the load model is discussed in three scenarios. In these scenarios, the load model is considered constant and also variable form in during a day. In section II, concept of active network management and available implementation programs for optimal operation is expressed. Section III presents multi objective optimization and related constraints. The proposed algorithm and its performance are introduced in Section IV. The sample system and issue assumptions are presented in Section V. In section VI, the simulations and studies results are shown. At the last section, results obtained from applied method are discussed in detail. II. ACTIVE NETWORK MANAGEMET This Active management guarantees all medium voltage distribution network along with DGs for increasing their installable capacity to network [3]. Figure 1 shows a simple AM schematic where the control instructions are sent to circuit breakers, reactive power compensators (capacitors or STATCOMs), transformers and generators which depend on measurement of primary system parameters.

Fig 1. Schematic of AM in distribution network

In this paper, three strategies AM are offered to enhance installed capacity of DG, along with keeping the voltage in the allowable range. A. Active power management of DG Maximum operation power for any DG unit is one of used strategy to limit the nodal voltage due to high penetration of DG. If expected amount of energy curtailment is relatively low, it may be beneficial to curtail generation instead of upgrading the distribution feeder whenever the nodal voltage would exceed allowable limit. However this strategy due to heavy costs for DG owners is used rarely [4]. B. On-load-tap-changer voltage control (OLTC) This procedure is performed through management of voltage levels at the MV substation. In this method, tap changer regulation can be performed based on voltage bus information which has the most problem of voltage increment [5]. Hence using OLTC can considerably increase capacity of distributed generation that can be installed to network without triggering reinforcement costs. C. Use of reactive power compensating (RPC) equipment In this method, reactive power compensators are used to improve voltage profile also this approach reduces network active power losses. If reactive power generated by RPCs is approximately equal to load reactive power, the rate of network active power losses will be minimum. The best control strategy is based on application of coordinated and synchronized control of voltage and reactive power.

defined objective functions. Since all these functions have been considered as cost functions form, we integrate them into a parametric objective function. The general form of this equation is as follows: N ⎧ ⎪min f ( x) = min ∑ α i × fi ( x) i =0 ⎪⎪ ⎨ci ( x) = 0 j=1 . . . n ⎪h ( x) = 0 k=1 . . . m ⎪ k ⎪⎩

where N represents number of objective sub-function and is weighting coefficients of objective functions. Weighting coefficients will be a specific amount in accordance with importance of objective sub-function. To simply the issue, we assume that objective sub-functions have the same weight coefficients. Practically use of load profile for mentioned problem is difficult, thus this profile is divided into appropriate time steps and is assumed that load value is fixed at every level. Four sub-functions can be considered for the specified cost function As follows: NP N J

f1 = ∑∑ Δti × P l × ρ Pssj

•

NP NJ

f 2 = ∑∑ Δti × ( Pss × ρ Pssj + Qss × ρQssj )

•

Mode.3: Active distribution networks with AM programs: In this mode, in addition to the installed DG in distribution network, AM programs will be implemented to optimize operation network distribution.

III. MULTI-OBJECTIVE OPTIMIZATION PROBLEM For desirable operation of network, simultaneously different purposes must be considered. Often objectives are in opposed side of each other and cannot be applied conventional optimization techniques for solving a problem. The application of multi objective methods give us information that will be based on decisions results of all the

(3)

i =1 j =1 N P NC

f 3 = ∑∑ Δti × Qcj × ρQcj

Mode.1: Traditional distribution networks: power flow is unidirectional of high voltage (HV) transmission network to customers in low voltage levels. Network is not connected to any DG unit. Mode.2: Passive distribution networks with DG: In these networks, the installable capacity of DG is available. Unidirectional power flows will change to bidirectional power flows when the penetration level of DG becomes higher. AM is not applied and outputs of DG units are not controlled by DNO, therefore DG units are simply taken as negative loads.

(2)

i =1 j =1

In this paper, three modes for distribution network have been proposed: •

(1)

(4)

i =1 j =1

N P N DG

f 4 = ∑ ∑ Δti × PDGj × ρ DGj

(5)

i =1 j =1

In the above relations Np, Nj, Nc and NDG express number of time steps, load levels related to each time step, number of RPCs and DGs of installed respectively.

Δti is ith interval of time, Pl is total losses of active power, Pss and Qss are active and reactive power supplied by the transmission line,

ρ Pssj

and

ρQssj

are ith energy costs of

load level in accordance with $ unit. Also

ρ DGj and ρQcj

are

th

i energy costs of DG and RPC respectively. The constraints related to optimal load flow will be as follows: D Ptss + Pi,DG t − Pi,t =Vi,t

∑V

j,t (Gij cosδi,t + Bij sinδ j,t )

(6)

j,t (Gij cosδi,t

(7)

j

D Qtss +QiDG ,t −Qi,t =Vi,t

∑V j

− Bij sinδ j,t )

Sij,t ≤ Sijmax

(8)

Vi min ≤ Vi ,t ≤ Vi max

(9)

Pssmin ≤ Pt ss ≤ Pssmax

(10)

min ss min k

ss t

max ss max k

Q

≤Q ≤Q

T

≤ Tk ,t ≤ T

(11) (12)

Pi ,Dh and QiD,h are active and reactive loads at the ith node and tth load level. Pi ,DG is generated active power at ith node t and tth load level.

Qcj reactive power generated/absorbed by

the reactive power compensator at tth load level. setting of the kth tap-changer at tth load level, of the branch ij at tth load level and th

Tk ,t is tap

Sij ,t is load flows

Vi ,t is voltage magnitude at

th

i node and t load level. IV. NSGA II ALGORITHM Genetic algorithm with global search ability widely has been used in planning and operating of power networks problems. The aim of genetic algorithm is obtaining the optimal or near optimal solution. Genetic algorithm in different points of function uses search space and does not require derivative or other information of the function. Genetic algorithm has some advantages compared to other optimization methods. The advantages of genetic algorithm include continuous or discrete variable optimization with complex objective functions, probabilistic laws instead of deterministic laws and work ability to apply many variables. Due to certain problems in classical optimization techniques, multi-objective evolutionary algorithms have been proposed which have ability to solve complex problems in discontinuous and multicriteria spaces [6,7].

Fig.2. NSGA-II procedure [7]

Fig.2 shows an iteration process of NSGA II. NSGA-II includes the following main steps: 1. Generate a random parent population P. 2. Sort parent population based on non-domination.

3. Assign fitness and create an off-spring population Q using binary tournament, recombination and mutation. 4. Combine parent and off-spring populations and form combined population R with size of 2N (except first period). 5. Sort population R based on non-domination: R= {F1, F2, …}. 6. Form the new parent population according to nondomination and crowding distance. 7. If the maximum number of generations is reached, the process will be stopped, else go to step 2. V. SAMPLE SYSTEM AND ASSUMPTION Simulations were carried out on the 33-bus system shown in Figure 3. The assumed voltage level of substation is 12.66 kV and capacity of the feeder is 8 MVA. Peak load is 6012 kW and 3012 kvar. Two groups of DG with the same size of 200 kW at the time of connection (5%) are installed at nodes 17 and 32 respectively. STATCOM is installed with DG unit as reactive power compensators and their investment costs are considered in DG units. Network data can be found in [8]. There are two important hypotheses about the loads: 1. reactive loads have a similar load model with active loads 2. each node loads follow the same load pattern. In this paper, MATLAB software has been used for the simulation.

. Fig.3. Single-line diagram of the 33-bus

GTs output is variable between 50 to 100% of their installed capacity. DG units are connected to power system with power electronic inverters typically. Three scenarios for distribution of the network load are defined as: Scenario 1: A load level for 24 hours is assumed. Total active load and reactive load of network are 3715 kw and 2345 Kvar at the average load. Cost of active power purchased by DNO of DGs is 40 $ / MWh and benefit of DG units is considered 10 $ / MWh. Also the cost of per Mvar reactive power of capacitors and network are 16 $. The cost of active power generated by network is 40 $ / MWh. Scenario 2: Three levels of load in during a day is assumed (Table 1). As same as the previous scenario, cost of active power purchased by DNO of DGs is considered 40 $ / MWh and benefit of DG units is 10 $ / MWh. Scenario 3: load levels have been considered as 24 time steps and for each load level, its cost is listed in Table 2. Against the previous scenarios, the purchased cost of DGs has considered 42$ / MWh. In all scenarios, in AM mode, the

implemented costs of AM programs are placed in cost function. hours 7 9 8

hours 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Table 1: information related to scenario2 Cost of reactive Cost of active Status of power power network load 18 $/MVARh 45 $/MWh Full load(%125) 16 $/MVARh 40 $/MWh Average load (%100) 14 $/MVARh 35 $/MWh Low load (%75) Table 2: information related to scenario3 Cost of active Percentage load Cost of reactive power ($/MWh) (%) compared to power the average load ($/MVARh) 25 62 160 24 60 155 24 60 155 23 56 145 23 54 140 23 54 138 21 52 134 20 49 126 19 47 123 19 46 120 18 44 116 18 43 112 17 41 107 16 39 102 15 38 98 14 35 95 14 34 92 13 32 85 13 32 84 12 30 81 11 29 78 11 28 74 10 26 70 10 25 65

VI. RESULT & ANALYSES This paper focuses on three cases of below: A. Increase the penetration level of DG installed in a radial network B. Reduce the network losses as a main index. C. Suggest a cost function that shows impact of AM programs in the radial networks. The summarization results of simulations are presented Table 3. In this Table, the results for three scenarios and three proposed pattern for operation of radial network are indicated. Also profit of DG units in per MWh has been calculated and the cost of network active losses is listed in Table 3. Figures 4, 5, 6 and 7 show results obtained from NSGA II for Scenario 2.

Fig.4. the variation of the network losses level with increasing penetration levels of DG in the passive network (Scenario 2)

Fig.5. the variation of proposed cost function changes with increasing levels of DG penetration in the passive networks (scenario 2)

Fig.6. the variation of the network losses level with increasing penetration levels of DG in the active network (Scenario 2)

Fig.7. The variation of proposed cost function changes with increasing levels of DG penetration in the active networks (scenario 2)

Fig.8. the variation of the network losses level with increasing penetration levels of DG in the passive network (Scenario 3)

Figure.11. the variation of proposed cost function changes with increasing levels of DG penetration in the active networks (scenario 3)

VII. CONCLUSION In this paper, we offered a new method for calculating the total energy cost for DNO. The idea of active management is applied for distribution network operation. With implementing AM programs, the losses and their costs decrease consequently. Thus AM will be many benefits for DNO for costs reduction. Also the manner of load distribution in sample network was considered at the three scenarios. To solve this problem, NSGA II is used which is a powerful tool in solving multi-objective function. Figure.9. the variation of proposed cost function changes with increasing levels of DG penetration in the passive networks (scenario 3)

Figure 4 and 5 show reduction of losses and cost function proposed for every the penetration level of DG in network. Also figures 6 and 7 show reduction of losses level and cost function in state of active management in the network. These results are shown for scenario 3 in Figures 8, 9, 10 and 11.

REFERENCES [1]

[2]

[3]

[4]

[5] [6]

[7] Figure.10. the variation of the network losses level with increasing penetration levels of DG in the active network (Scenario 3)

Obviously, the installation of DG in radial networks will reduce the losses rate. Therefore participation of DG in radial networks is beneficial for DNO. Losses reduction and energy cost purchased from DG units have main role in reducing the proposed cost function.

[8]

S. Abapour, K. Zare, B. M. ivatloo. “Evaluation of technical risks in distribution network along with distributed generation based on active management,” IET Gener. Transm. Distrib, 2014, 8(4), pp: 609-618 S. Abapour, E. Babaei and B.Y. Khanghah, "Application of active management on distribution network with considering technical issues," In Proc. IEEE Smart Grids (ICSG), pp. 1-6, 2012. Fabio Bignucolo, Roberto Caldon, Valter Prandoni." Radial MV networks voltage regulation with distribution management system coordinated controller” Electric Power Systems Research 78 (2008) 634–645 A. Goikoetxea,J.A. Barrena,M.A. Rodriguez,G. Abad” Active Substation design to maximize DG integration” 2009 IEEE Bucharest Power Tech Conference. Zechun Hu, Furong Li. Distribution Network Reinforcement Utilizing Active Management Means. 2010 IEEE Power Tech Conference A. Moeini H. Yassami M. Owlady M.H. Sadeghi. Disco Planner Flexible DG Allocation in MV Distribution Networks Using MultiObjective Optimization Procedures. 2010, 12th International Conference on Optimization of Electrical and Electronic Equipment, K. Deb, A. Pratap, A. Agarwal, and T. Meyarivan, “A fast and elitist multi-objective genetic algorithm: NSGA II,” IEEE Trans. Evol. Comput., vol. 6, no. 2, 2002,pp. 182–197. M. A. Kashem, V. Ganapathy, G. B. Jasmon, M. I. Buhari, “A novel method for loss minimization in distribution networks,” in Proc. Int. Con.Electr. Util. Deregulation Restruct. Power Technol., Proc., Apr. 2000, pp. 251–256

Table 3: Data relating to the results of simulations How to operation

PLoss (kw)

Cost Function ($)

---

---

220

211.2

5010

2109 2450 ---

758.8 880.15 ---

121.8 78.13 220

117 74.85 211.2

4756 4635 5010

2121 2503 ---

763.56 901.1 ---

120.6 75.95 266

115.77 72.92 255.36

4783 4649 6408

2410 2801

865.7 1006.12

152.6 95.36

146.5 91.54

6024 5791

scenarios

Scenario 1 Scenario 2 Senario3

Conventional Network Passive Network Active Network Conventional Network Passive Network Active Network Conventional Network Passive Network Active Network

Benefit of DG units ($)

Cost of losses($)

PDG (kw)