- Email: [email protected]
Real-time reconfiguration of distribution network with distributed generation
Electric Power Systems Research 107 (2014) 59–67
Contents lists available at ScienceDirect
Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr
Real-time reconfiguration of distribution network with distributed generation D.P. Bernardon ∗ , A.P.C. Mello, L.L. Pfitscher, L.N. Canha, A.R. Abaide, A.A.B. Ferreira UFSM – Federal University of Santa Maria, RS, Brazil
a r t i c l e
i n f o
Article history: Received 30 July 2013 Received in revised form 18 September 2013 Accepted 18 September 2013 Keywords: AHP Automatic reconfiguration Distribution network Distributed generation Remote controlled switches Smart grid
a b s t r a c t This paper presents a new methodology to perform the automatic reconfiguration of distribution networks incorporating distributed generation in normal operation. The power generation availability of wind turbines, solar photovoltaic panels and small hydropower are considered in the reconfiguration process. The real-time reconfiguration methodology is based on a heuristic method to determine the best settings. The method assumes that only remote controlled switches are considered in the analysis. The multicriteria analysis AHP (Analytic Hierarchy Process) method is employed to determine the best sequence of switching. The developed algorithms are integrated into a supervisory system, which allows real-time measurements and commands to the equipment. The proposed methodology is tested in a real network of a power utility and results are presented and discussed. To evaluate the performance and efficiency of the proposed method, different network reconfiguration scenarios with distributed generation were tested. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The need to improve the quality and reliability of power systems, as well as environmental needs with the quest for reducing greenhouse gas emissions and replacing fossil fuels in power generation, insert Distributed Generation (DG) using renewable energy sources in discussions about the future of distribution of electric power systems. New methodologies and analysis tools for distribution systems with distributed generators are necessary and have been receiving great attention from researches of Smart Grids and Microgrids. Smart grids are characterized by a number of technologies, methodologies and integrated procedures, including the ability for self-reconfiguration under changing operating conditions [1]. The reconfiguration of the network can be responsible for promoting more efficient use of distributed generators of renewable primary sources, such as solar radiation and wind energy, by analyzing the availability of each source for injecting energy into the system. As a result, it is possible to obtain a significant reduction of losses and to improve the reliability of power supply. Several researches are related to the reconfiguration of distribution network with distributed generators, based on mathematical methods, heuristics and artificial intelligence techniques [2].
∗ Corresponding author. Tel.: +55 55 32208344; fax: +55 55 32208030. E-mail addresses: [email protected], [email protected] (D.P. Bernardon). 0378-7796/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.epsr.2013.09.011
In recent years, new methodologies of reconfiguration of distribution network have been presented, exploring the greater capacity and speed of computer systems, the increased availability of information and the advancement of automation, particularly, the SCADA (Supervisory Control and Data Acquisition). With the increased use of SCADA and distribution automation through remote controlled equipment, the reconfiguration of distribution network became more viable as a tool for planning and control in real-time. Most of recent researches [3–5], however, do not take into account reconfiguration in real-time. The works of Wang et al. [6] and López et al. [7] present an online reconfiguration approach, the first with emphasis on the reduction of load flow calculation time and the last on the demand characterization. The work of Vargas and Samper [8] presents fast algorithms for a smart distribution management system, which include load estimation, load flow calculation and optimal feeder reconfiguration considering DG. Rao et al. [9] present a meta-heuristic algorithm to reduce losses of distribution networks and the allocation of DG units simultaneously. In these works, however, the optimization approach is for monocriterial analysis, and some feasibility aspects for real-time application are not taken into account, such as a proper definition of the frequency of reconfiguration and the coordination of protection devices. A multicriteria analysis is discussed in the work of Martins and Borges [10], using the reconfiguration of network as an alternative to system expansion. In general, the vast majority of works which involve DG, focuses on the planning of distribution systems and allocation of DG, and does not apply to the reconfiguration under normal operation and real-time.
60
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
The work of Celli et al. [11] considers the short-term variations in generation due to meteorological conditions. In real-time applications it is necessary consider this information because the generation profiles can change significantly. In this work, a methodology and a computational tool for automatic reconfiguration of distribution network considering DG in real-time are developed from the standpoint of Smart Grid. The methodology uses the information and functionality of remote controlled equipment installed in distribution systems, applying them in a computer system that allows the reconfiguration of the network in normal operation. The analysis consider the demand curves of feeders and generation curves based on Wind, Solar Photovoltaic and Small Hydropower energies, highlighting that these systems are interconnected to the distribution network. The developed methodology is an evolution of the works [12,13], but with significant contributions: (i) analysis of generation profiles related to distributed generators of different technologies; (ii) specific constraints for distributed generation; (iii) heuristic technique for selecting configurations incorporating DG; (iv) case studies considering real data of power utilities; (v) integration of the developed tool with SCADA system, allowing the automatic network reconfiguration. The proposed methodology is tested in a real distribution network, ensuring the practical applicability of the heuristic optimization algorithm employed. 2. Problem formulation The reconfiguration of distribution network can be characterized as an optimization problem. To improve network performance, one or more objectives (e.g. reduce losses and improve reliability) are established, and then one verifies which configuration produces
the best result, without violating constraints on proper and safe operation of the network. This configuration is defined as the optimal solution for the system. When more than one objective is set, the analysis should incorporate methods for multicriteria decision making, which may include expert opinion on the definition of a preference for one objective over another. The main problem is that optimization of real networks allows a number of configurations quite high due to the number of switching devices on the network. In general, it may be unfeasible to test all possible combinations and perform, for each one, the necessary calculations – such as load flow – in order to identify the configuration that results in best performance. To solve this problem, optimization methods that reduce the search space of the optimal solution are commonly used. In this work, the developed optimization algorithm is based on a heuristic method and on the AHP multicriteria decision making method to identify the best network configuration. In the sequence, some fundamental considerations are predefined to perform automatic reconfiguration in real-time. 2.1. Real-time characteristics and requirements Reconfiguration of distribution network in normal operation usually has as main objective the reduction of energy losses. When considering automatic reconfiguration in real-time, some aspects have to be included in the study: a) it is necessary to establish a cost-benefit relationship to determine the necessity and the effectiveness of the reconfiguration; b) the network must be flexible to allow the reconfiguration; c) the technical feasibility of the switching should be studied in real-time, considering the availability of current measurements of the network. The diagram in Fig. 1 shows the architecture employed in this work to meet these premises.
Fig. 1. Architecture of the proposed system.
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
900
0,6
800
5
0,5
700
4 3 2
P(kW)
600
0,4
Cp
500
0,3
400 300
0,2
200
1
0,1
100 0
0 Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Power Coefficient (Cp)
b
6
Power (kW)
Wind Speed (m/s)
a
61
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Dec
Wind Speed (m/s)
Month
Fig. 2. (a) Wind resource (b) power curve for wind turbine Enercon E53.
The real-time information access is needed to implement the methodology, such as the data and communication status of the devices controlled remotely, which is obtained from SCADA. The reconfiguration software evaluates the load rates and generation scenarios of DG, and it performs the optimization method considering the heuristic search and multicriteria analysis based on AHP. The proposed system also sends commands to open and close the switches by SCADA. The steps of the presented architecture are discussed in Section 3.
3.2.2. Photovoltaic power systems The profile of photovoltaic generation was defined based on indices of sunshine and solar radiation incidence in the South region of Brazil. Fig. 4 illustrates the typical profile throughout the year. It was considered a set of PV modules with a total power of 0.5 MW. Fig. 5 shows the photovoltaic generation curve for a typical day with clear sky normalized to maximum generation (0.5 MW).
3. Methodology for automatic reconfiguration in real-time 3.1. Demand rate evaluation
3.2. Profile of distributed generation The DG incorporation considered in this paper refers to a moderate insertion of midrange DG with installed power of a few MW from renewable sources. The DG is assumed to be connected to the medium voltage of the distribution feeders. During the reconfiguration occurs the simultaneous analysis of the demand and generation curves. The program verifies the availability of DG power for injection into the system.
3.2.1. Wind power (W) In this work, it was considered a small wind farm and a typical generation scenario from historical data of average wind speed obtained by database of a weather station located in the South region of Brazil, with hourly measurements. Fig. 2(a) shows the monthly incidence of wind during the study period. The wind turbine model used was the Enercon E53, with rated power of 800 kW, as shown in the power curve in Fig. 2(b). From the meteorological parameters and the power curve of the turbine, it is possible to obtain the daily curve of the potential for wind power generation. Fig. 3 illustrates the wind power curve for a typical day with good wind incidence, normalized to the maximum generation value (800 kW).
Fig. 3. Wind generation curve.
8
Daily Radiation (kWh/m²/d)
The first step on the proposed methodology is to determine when a change in network configuration is actually needed. This avoids frequent switching which may cause equipment degradation. The strategy employed in this work is based on demand rates and load forecasting using typical curves for residential, industrial, commercial and rural consumers. In order to obtain a good discretization of the demand curve and to avoid frequent reconfigurations in the network, it is employed a set of six demand rates, as presented in work [13].
7 6 5 4 3 2 1 0 Jan
Feb
Mar
Apr
May
Jun Jul Aug Daily Radiation
Sep
Oct
Nov
Fig. 4. Monthly radiation profile throughout the year.
Fig. 5. Photovoltaic generation for a typical day with clear sky.
Dec
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
Subject to
1
|Ii | ≤ Ii max Vj min ≤ Vj ≤ Vj max
0,8
(2)
Pmin ≤ PDGn ≤ Pmax
0,6 0,4 0,2 0
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00
Small Hydroeletric Power (p.u)
62
Period (hour) Fig. 6. SHP curve generation.
3.2.3. Small hydroelectric power plant (SHP) The power plant operation depends on the river flow and on the reservoirs capacity, so there is also considerable uncertainty regarding this type of generation and the share that can be considered in the studies. The flow data were extracted from the database of an SHP, in the South region of Brazil. From the knowledge of the generation characteristics such as local water availability and SHP machines availability, it was possible to obtain the hydraulic power curve generated by the SHP. Fig. 6 illustrates the generation curve normalized to the maximum generation value (1 MW), for a typical day. Several researches that analyze DG interconnected with distribution networks consider that the DG unit is able to keep generating throughout of the peak demand period. However, this does not always happen because there are some uncertainty factors, such as changes in radiation and wind speed levels, hydrological risks and possibility of machines breakdowns due to the mechanical problems, causing the unavailability of generation. This work considers the behavior of generation sources, using the expected values for each period of analysis. In the worst-case scenarios with low rate of DGs generation availability (for example, solar energy in the period of light load) these ones do not participate in the reconfiguration process.
3.3. Heuristic reconfiguration technique using the potential generation of DG 3.3.1. Objective functions and constraints The objective function (OF) used in this work includes the minimization of three distinct functions: (i) Normalized Expected Loss of Energy in the primary network (ELossesi *) (ii) Normalized Expected System Average Interruption Frequency Index in the system (ESAIFIi *) and (iii) Normalized Expected value of Energy Not Supplied (EENSi *). The total gain obtained in each configuration combines the three functions and it is presented in (1): OF = min(ELossesi ∗ · w1 + ESAIFIi ∗ · w2 + EENSi ∗ · w3 )
(1)
where i corresponds to the period of analysis: 1. . .6, w1 . . .w3 are weights of the criteria in multicriteria method, Ii is current magnitude of each element must lie within its permissible limits (A), Vj is voltage magnitude of each node must lie within its permissible ranges (kV) and PDGn is active power delivered by the distributed generator n. The values of Pmin and Pmax depend on the DG technology considered (kW). In addition to these constraints the system should always keep radiality after reconfiguration of feeders and should not allow islanded operation of DG units. To calculate the load flow values in the distribution system, the proposed algorithm implements the backward/forward sweep method with DG [14,15], which is suitable for calculation of reverse flows from distributed generation. With load flow values, the energy losses for the distribution system are obtained. The calculation of reliability indicators ESAIFI and EENS is obtained from the equations of reliability for these indicators during the process of load flow [16]. 3.3.2. Weights of criteria The indicators of the OF are judged according to the decision making method AHP developed by Saaty [17]. The method is premised on the trial and the conversion of the criteria weights, so one can adjust the priority that each criterion has over the others. In [13] is presented the methodology in more details and calculations for setting the weights of the criteria and judgment matrices used in this work. The following results are assumed: w1 (ELosses) = 0.64; w2 (ESAIFI) = 0.26 and w3 (EENS) = 0.10. 3.3.3. Heuristic technique for selecting configurations The optimization technique employed in this work is based on the Branch Exchange adaptation proposed by [12]. This step is based on the analysis of the interconnections between the feeders, which are performed by normally open (NO) switches. For each interconnection, the best network configuration between two feeders is searched, by changing the status of a pair of switches. The method has the following advantages over other techniques from the literature: • it highly limits the search space of solutions, reducing calculation and computational processing time; • it does not require adjustments of optimization parameters due to changes in the network (such as grid expansion or installation of new equipment); • the radial network is guaranteed, without the need to verify if the combination of switches that comprise the topology tested is valid. The reconfiguration methodology was divided into two modules: (A) analysis of connection between feeders without DG
Fig. 7. Distribution network: (a) Original configuration. (b) Switching status changed.
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
63
Fig. 8. Switching status changed. Fig. 11. DG connected in feeder FD3.
Fig. 9. Switching status changed on the opposite direction.
connected to the network and (B) analysis of connection between feeders with DG. 3.3.3.1. Reconfiguration without DG on feeders. 1st Stage: Choose determined remotely controlled NO (normally opened) tie-switch at random, starting from the original configuration of the network (Fig. 7(a)). Go to the 2nd stage; 2nd Stage: Change the network configuration, closing an NO tie-switch and opening an NC (normally closed) switch of any of the two feeders involved. However, it should be the first switch upstream from the NO tie-switch (TS1), as shown in Fig. 7(b). Go to the 3rd stage; 3rd Stage: Perform the load flow and reliability calculation for this new configuration and check if there was a reduction of OF without violating the constraints. If OF is improved, proceed to the 4rd stage. Otherwise, go to the 5th stage. 4th Stage: Change the network configuration, closing the actual NO switch (S6) and opening the first NC switch upstream (S7), but in the same direction (feeder) of the previous iteration (Fig. 8). Go to the 6th stage. 5th Stage: Re-establish the initial configuration and change the network topology, closing the NO tie-switch and opening the NC switch of the other feeder upstream from the tie-switch (Fig. 9). Go to the 6th stage. 6th Stage: Change the network configuration in the same direction (feeder) while the OF is improved without violating the defined constraints. The process is finished when the OF does not improve or if any constraint violation occurs. In this case, the topology found in the previous iteration is the final solution. The same process must be repeated when another tie-switch is analyzed, always beginning from the initial configuration. The process is repeated until all remote controlled NO tie-switches without DG on feeders are analyzed.
Fig. 12. DG connected in feeder FD4.
3.3.3.2. Reconfiguration with DG on feeders. Assuming that distributed generation injects power into the grid, it is necessary to evaluate the gain of its connection to each feeder. Initially starts with the original configuration with DG, as shown in Fig. 10.
1st Stage: Test the gain of DG connection on the current feeder (FD3), closing the remotely controlled NO tie-switch (TS2) and opening the next NC switch (S14) downstream of tie-switch, as illustrated in Fig. 11. Go to the 2nd stage; 2nd Stage: Perform the load flow and reliability calculation for this new configuration and check if there was a reduction of OF without violating the constraints. If OF is improved, proceed to the 3rd stage. Otherwise, go to the 4th stage; 3rd Stage: Change the network configuration in the same direction (feeder) while the OF is improved without violating the defined constraints. Go to the 4th stage; 4th Stage: Test the connection of GD on the other feeder (FD4). From the original configuration with DG (Fig.), open the first NC switch (S13) upstream of DG and close the NO tie-switch (TS2), as illustrated in Fig. 12. Go to the 5th stage; 5th Stage: Perform the load flow and reliability calculation for this new configuration and check if there was a reduction of OF without violating the constraints. If the OF is improved, go to the 6th stage. Otherwise, the process of reconfiguration is finished, go to the 7th stage. 6th Stage: Change the network configuration in the same direction (feeder) while the OF is improved without violating the defined constraints (Fig. 13). The process of reconfiguration is finished when the OF does not improve or if any constraint violation occurs. Go to the 7th stage; 7th Stage: The configuration with the greatest reduction in the OF is chosen as best topology, comparing the results obtained in the connection of DG on feeder FD3 (3th Stage) and FD4 (6th Stage). The same process must be repeated when another source of DG is analyzed, always beginning from the initial configuration. The methodologies presented assume that only the remote controlled switches are analyzed and it consists in applying the branch exchange algorithm twice: first, to determine the individual result of each configuration test from the initial network topology and second, to determine the final sequence of configurations. In the
Fig. 10. Original configuration with distributed generation.
Fig. 13. Switching status changed.
64
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
Fig. 14. Network distribution with DG.
first case, the procedure is repeated for each normally open tieswitch starting from the initial network topology. At the end of this process it is applied the multicriteria decision method described in the previous section, to define the best sequence of switching. The procedure of optimization is then applied for the second time, but for each tie-switch tested, the analysis is made without returning to the initial configuration. 4. Results and discussion 4.1. Validation of the methodology An exhaustive search procedure was implemented in a simplified distribution network (Fig. 15), in order to assess the robustness of proposed method. The best solution for this network was previously obtained considering only the power losses minimization and compared with the proposed methodology, as shown in Table 1. Fig. 15 demonstrates a network with seven switches which leads to a total of 128 (27 ) possible combinations, and it is used to validate the proposed method. In actual networks is not feasible to apply the method of exhaustive search due to combinatorial explosion caused by the number of switching devices in the network. 4.2. Practical implementation of the methodology The proposed methodology was verified through several tests on the concession area of a power utility of Brazil. The real Table 1 Reconfiguration results for exhaustive search compared with the methodology. Parameter Initial solution loss (kW) Optimal solution loss (kW) Obtained loss (kW) Processing time (3.33 GHz)
Exhaustive search
34.72 32 s
Proposed method 37.74 34.72 34.72 8s
distribution network model presented in Fig. 14 is used as a case study. This network has 2 substations 69/13.8 kV, 5 feeders, 3 points of distributed generation, 15 tie-switches, 99 normally closed switches and over 21,000 consumers. The switches are remote controlled. The dotted lines represent the presence of normally open tie-switches (TS). The DG is composed of a SHP of 1 MW located at node 6, two wind turbines (E53) of 800 kW at node 33 and a photovoltaic park of 500 kW located at node 37. Data generation from these sources is updated for each load rate, in order to consider the power injection of DG on its current condition generation. Table 2 illustrates the test condition considered for each level of demand and each consumer class and generation scenario during the reconfiguration of distribution network. The reconfiguration algorithm is applied considering the individual analysis of each switch interconnection shown in Fig. 14. It evaluates the power injection of DG for each feeder involved based on the conditions of generation scenarios in Table 2. The individual analysis of the tie-switches leads to the results shown in Table 3 from original configuration of networks, comparing the results with and without DG for the period of 6:00 to 12:00. Only the cases where the objective functions analyzed presented positive evolution are shown. Note that in normal situations, the number of consumers of a feeder does not vary during the day, which suggests the same ESAIFI value in all the simulations. Each result in Table 3 is normalized by the maximum value found in the tests (Cn base ) as follows: Cn∗o pm =
Cno pm Cn base
(3)
After normalization, the individual results are multiplied by the weight vector and summed as in (1). The results are sorted by a global index (GI), from lowest to highest. This result indicates the sequence of switching for the best combination of results from
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
65
Fig. 15. Simplified distribution network with DG.
Table 2 Discretization curve of demand for consumer class and generation scenario of DG. Period – initial hour 0:00
6:00
12:00
13:00
17:00
21:00
Feeder class (demand)
Residential Industrial Commercial Rural
Low Low Low Low
Medium Heavy Heavy Medium
Low Medium Low Low
Medium Heavy Low Medium
Heavy Medium Heavy Heavy
Medium Low Low Medium
DG (generation)
Photovoltaic Windpower SHP
– Low Medium
Medium Medium Medium
Heavy Heavy Medium
Medium Heavy Medium
Low Medium Medium
– Low Medium
Table 3 Results for the individual analysis of tie-switches. Options: close/open
Rate 06:00–12:00 hour Initial configuration Option 1: TS-1/S3 Option 2: TS-2/S18 Option 3: TS-3/S12 Option 4: TS-5/S22 Option 5: TS-7/S59 Option 6: TS-8/S84 Option 7: TS-9/S63 Option 8: TS-10/S91
Configuration without DG
Configuration with DG
Losses (kWh)
EENS (MWh)
ESAIFI (fail./year)
Losses (kWh)
EENS (MWh)
ESAIFI (fail./year)
2263.32 2243.94 2265.89 2143.5 2266.98 2242.57 2037.01 2153.80 2011.90
603.20 590.90 596.30 554.40 603.50 577.70 569.80 581.20 567.10
11.78 11.30 11.46 9.93 11.74 11.66 11.75 11.62 11.67
2094.85 2079.20 2094.32 2010.91 2037.14 1983.50 1752.88 1928.56 1759.74
562.80 551.50 556.20 520.80 580.40 532.60 520.30 538.40 519.30
11.78 11.30 11.46 9.93 11.74 11.66 11.75 11.62 11.67
the individual switches interconnection analysis. Table 4 illustrates this procedure for the period of 6:00 to 12:00 with DG the system. The Branch Exchange algorithm is reapplied, following the sequence of operations from the GI ranking of Table 4. The
difference from the previous analysis is that the best configuration obtained with one tie-switch is maintained as the initial configuration to the following tie-switch to be tested. The final result of the optimization of the network to the rate demand analyzed is shown in Table 5.
Table 4 Normalized results, sequence and global index with AHP method. Options
Energy losses × 0.64
ESAIFI × 0.26
EENS × 0.10
Global index (GI)
Sequence
Rate 06:00–12:00 hour Option 1: TS-1/S3 Option 2: TS-2/S18 Option 3: TS-3/S12 Option 4: TS-5/S59 Option 5: TS-7/S59 Option 6: TS-8/S84 Option 7: TS-9/S63 Option 8: TS-10/S91
0.99 1.00 0.96 0.97 0.94 0.83 0.92 0.84
0.95 0.97 0.84 0.99 0.99 0.99 0.98 0.99
0.95 0.95 0.89 1.00 0.91 0.89 0.92 0.89
0.9796 0.9886 0.9232 0.9814 0.9551 0.8845 0.9383 0.8846
6 8 3 7 5 1 4 2
66
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
Table 5 Final results of reconfiguration analysis with DG. Sequence of test
Switching with best result
Energy losses (kWh)
ESAIFI (failures/year)
EENS (MWh)
Rate 06:00–12:00 hour Initial configuration Option 6 Option 8 Option 3 Final reduction (%)
– Close TS-8/Open S84 Close TS-10/Open S91 Close TS-3/Open S12
2094.85 1752.88 1756.54 1672.60 20.15%
11.78 10.75 11.67 9.90 15.95%
632.90 520.30 520.4 478.30 15.01%
All procedures of the main program have been successful and the final topology of the network determined by the program improved the indicators of the objective functions of the optimization method, with no constraints in relation to the necessary switching. One way to confirm if the solution presented is the optimal solution is using the final configuration obtained by proposed methodology as input to a new test sequence. If there is no change in the topology, there is a strong indication that the configuration obtained is the optimal solution or very close to the optimal solution. The tests were performed in about 3 min in a PC with Intel Core 3.33 GHz and 24 GB RAM. The scale of few minutes is acceptable for the proposed application, since the reconfiguration is analyzed at the beginning of each demand rate transition and the demand rate interval is in the range of one or few hours.
5. Conclusions This paper presented a new methodology for automatic reconfiguration of power distribution networks incorporating distributed generation. The main contributions of the proposed system are: real-time reconfiguration based on load rate analysis, modeling and generation profile of DG from different sources (wind, solar and SHP); application of the AHP method for multicriteria decision making and integration of computational analysis with a Supervisory Control and Data Acquisition of remote controlled switches to allow performing automatic reconfiguration in real-time. The switching are performed automatically, in the sequence determined by the program. The results stand out the contribution of DG for significant improvement on network performance indicators, reducing losses and increasing reliability in the process of reconfiguration of distribution systems in which they are connected. For a real evaluation of the system performance, case studies were performed with real data from a power utility in different operating scenarios. The results indicate the feasibility of the proposed methodology, which can be adapted to other real systems with DG. With the advent of the Smart Grid, the profile of load curves of distribution feeders will be subject to a different dynamic behavior than currently occurs. Some features, such as greater use of distributed generation, demand response and charging of electric vehicles will require a rapid response of the network to the new generation and load scenarios. The automatic reconfiguration in real-time will help to improve network performance and to promote more efficient use of its resources in this new context.
Acknowledgments The authors would like to thank the technical and financial support of AES Sul Distribuidora Gaúcha de Energia SA, Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundac¸ão de Amparo à Pesquisa do Estado do Rio Grande do Sul
(FAPERGS) and Coordenac¸ão de Aperfeic¸oamento de Pessoal de Nível Superior (CAPES). Appendix A. Nomenclature
AHP ELossesi ESAIFIi EENSi i Ii
Analytic Hierarchy Process Expected Loss of Energy in the primary network Expected System Average Interruption Frequency INDEX Expected Energy Not Supplied corresponds to the period of analysis: 1. . .6 current magnitude of each element must lie within its permissible limits (A) normally closed NC NO normally opened objective function OF PDGn active power delivered by the distributed generator n SCADA Supervisory Control and Data Acquisition w1 . . .w3 weights of the criteria in multicriteria method voltage magnitude of each node (kV) Vj References
[1] R.E. Brown, Impact of smart grid on distribution system design, IEEE Power and Energy Society General Meeting – Conversion and Delivery of Electrical Energy in the 21st Century (2008) 1–4. [2] T. Thakur, Jaswanti, Study and characterization of power distribution network reconfiguration, in: IEEE Transmission & Distribution Conference and Exposition, Latin America, 2006, pp. 1–6. [3] M.S. Thomas, S. Arora, V.K. Chandna, Distribution automation leading to a smarter grid, IEEE Innovative Smart Grid Technologies (2011) 211–216. [4] D. Haughton, G.T. Heydt, Smart distribution system design: automatic reconfiguration for improved reliability, IEEE Power and Energy Society General Meeting (2010) 1–8. [5] M. Merdan, W. Lepuschitz, T. Strasser, F. Andren, Multi-agent system for self-optimizing power distribution grids, IEEE International Conference on Automation, Robotics and Applications (2011) 312–317. [6] J.-C. Wang, H.-D. Chiang, G.R. Darling, An efficient algorithm for real-time network reconfiguration in large scale unbalanced distribution systems, IEEE Transactions on Power Systems 11 (1996) 511–517. [7] E. López, H. Opazo, L. Garcia, P. Bastard, Online reconfiguration considering variability demand: applications to real networks, IEEE Transactions on Power Systems 19 (2004) 549–553. [8] A. Vargas, M.E. Samper, Real-time monitoring and economic dispatch of smart distribution grids: high performance algorithms for DMS applications, IEEE Transactions on Smart Grid 3 (2) (2012) 866–877. [9] R.S. Rao, K. Ravindra, K. Satish, S.V.L. Narasimham, Power loss minimization in distribution system using network reconfiguration in the presence of distributed generation, IEEE Transactions on Power Systems 28 (1) (2013) 317–325. [10] V.F. Martins, C.L.T. Borges, Active distribution network integrated planning incorporating distributed generation and load response uncertainties, IEEE Transactions on Power Systems 26 (4) (2011) 2164–2172. [11] G. Celli, E. Ghiani, F. Pilo, G.G. Soma, Reliability assessment in smart distribution networks, Electric Power Systems Research 104 (2013) 164–175. [12] D.P. Bernardon, V.J. Garcia, A.S.Q. Ferreira, L.N. Canha, Multicriteria distribution network reconfiguration considering subtransmission analysis, IEEE Transactions on Power Delivery 25 (4) (2010) 2684–2691. [13] L.L. Pfitscher, D.P. Bernardon, L.N. Canha, V.F. Montagner, V.J. Garcia, A.R. Abaide, Intelligent system for automatic reconfiguration of distribution network in real time, Electric Power Systems Research 97 (2013) 84–92.
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67 [14] W.H. Kersting, D.L. Mendive, An application of ladder network theory to the solution of three-phase radial load-flow problems, IEEE Power Engineering Society Winter Meeting A76044-8 (1976) 1–6. [15] L. Zhang, W. Tang, H. Guan, L. Zhang, The back/forward sweep-based power flow method for distribution networks with DGs, IEEE International Conference on Power Electronics and Intelligent Transportation System (2009) 145–149.
67
[16] R.E. Brown, Electric Power Distribution Reliability, second ed., CRC Press, New York, 2009, ISBN 978-0-8493-7567-5. [17] T.L. Saaty, The Analytic Hierarchy Process: Planning, Priority Setting, Resource Allocation, McGraw-Hill, New York, 1980.
Contents lists available at ScienceDirect
Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr
Real-time reconfiguration of distribution network with distributed generation D.P. Bernardon ∗ , A.P.C. Mello, L.L. Pfitscher, L.N. Canha, A.R. Abaide, A.A.B. Ferreira UFSM – Federal University of Santa Maria, RS, Brazil
a r t i c l e
i n f o
Article history: Received 30 July 2013 Received in revised form 18 September 2013 Accepted 18 September 2013 Keywords: AHP Automatic reconfiguration Distribution network Distributed generation Remote controlled switches Smart grid
a b s t r a c t This paper presents a new methodology to perform the automatic reconfiguration of distribution networks incorporating distributed generation in normal operation. The power generation availability of wind turbines, solar photovoltaic panels and small hydropower are considered in the reconfiguration process. The real-time reconfiguration methodology is based on a heuristic method to determine the best settings. The method assumes that only remote controlled switches are considered in the analysis. The multicriteria analysis AHP (Analytic Hierarchy Process) method is employed to determine the best sequence of switching. The developed algorithms are integrated into a supervisory system, which allows real-time measurements and commands to the equipment. The proposed methodology is tested in a real network of a power utility and results are presented and discussed. To evaluate the performance and efficiency of the proposed method, different network reconfiguration scenarios with distributed generation were tested. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The need to improve the quality and reliability of power systems, as well as environmental needs with the quest for reducing greenhouse gas emissions and replacing fossil fuels in power generation, insert Distributed Generation (DG) using renewable energy sources in discussions about the future of distribution of electric power systems. New methodologies and analysis tools for distribution systems with distributed generators are necessary and have been receiving great attention from researches of Smart Grids and Microgrids. Smart grids are characterized by a number of technologies, methodologies and integrated procedures, including the ability for self-reconfiguration under changing operating conditions [1]. The reconfiguration of the network can be responsible for promoting more efficient use of distributed generators of renewable primary sources, such as solar radiation and wind energy, by analyzing the availability of each source for injecting energy into the system. As a result, it is possible to obtain a significant reduction of losses and to improve the reliability of power supply. Several researches are related to the reconfiguration of distribution network with distributed generators, based on mathematical methods, heuristics and artificial intelligence techniques [2].
∗ Corresponding author. Tel.: +55 55 32208344; fax: +55 55 32208030. E-mail addresses: [email protected], [email protected] (D.P. Bernardon). 0378-7796/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.epsr.2013.09.011
In recent years, new methodologies of reconfiguration of distribution network have been presented, exploring the greater capacity and speed of computer systems, the increased availability of information and the advancement of automation, particularly, the SCADA (Supervisory Control and Data Acquisition). With the increased use of SCADA and distribution automation through remote controlled equipment, the reconfiguration of distribution network became more viable as a tool for planning and control in real-time. Most of recent researches [3–5], however, do not take into account reconfiguration in real-time. The works of Wang et al. [6] and López et al. [7] present an online reconfiguration approach, the first with emphasis on the reduction of load flow calculation time and the last on the demand characterization. The work of Vargas and Samper [8] presents fast algorithms for a smart distribution management system, which include load estimation, load flow calculation and optimal feeder reconfiguration considering DG. Rao et al. [9] present a meta-heuristic algorithm to reduce losses of distribution networks and the allocation of DG units simultaneously. In these works, however, the optimization approach is for monocriterial analysis, and some feasibility aspects for real-time application are not taken into account, such as a proper definition of the frequency of reconfiguration and the coordination of protection devices. A multicriteria analysis is discussed in the work of Martins and Borges [10], using the reconfiguration of network as an alternative to system expansion. In general, the vast majority of works which involve DG, focuses on the planning of distribution systems and allocation of DG, and does not apply to the reconfiguration under normal operation and real-time.
60
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
The work of Celli et al. [11] considers the short-term variations in generation due to meteorological conditions. In real-time applications it is necessary consider this information because the generation profiles can change significantly. In this work, a methodology and a computational tool for automatic reconfiguration of distribution network considering DG in real-time are developed from the standpoint of Smart Grid. The methodology uses the information and functionality of remote controlled equipment installed in distribution systems, applying them in a computer system that allows the reconfiguration of the network in normal operation. The analysis consider the demand curves of feeders and generation curves based on Wind, Solar Photovoltaic and Small Hydropower energies, highlighting that these systems are interconnected to the distribution network. The developed methodology is an evolution of the works [12,13], but with significant contributions: (i) analysis of generation profiles related to distributed generators of different technologies; (ii) specific constraints for distributed generation; (iii) heuristic technique for selecting configurations incorporating DG; (iv) case studies considering real data of power utilities; (v) integration of the developed tool with SCADA system, allowing the automatic network reconfiguration. The proposed methodology is tested in a real distribution network, ensuring the practical applicability of the heuristic optimization algorithm employed. 2. Problem formulation The reconfiguration of distribution network can be characterized as an optimization problem. To improve network performance, one or more objectives (e.g. reduce losses and improve reliability) are established, and then one verifies which configuration produces
the best result, without violating constraints on proper and safe operation of the network. This configuration is defined as the optimal solution for the system. When more than one objective is set, the analysis should incorporate methods for multicriteria decision making, which may include expert opinion on the definition of a preference for one objective over another. The main problem is that optimization of real networks allows a number of configurations quite high due to the number of switching devices on the network. In general, it may be unfeasible to test all possible combinations and perform, for each one, the necessary calculations – such as load flow – in order to identify the configuration that results in best performance. To solve this problem, optimization methods that reduce the search space of the optimal solution are commonly used. In this work, the developed optimization algorithm is based on a heuristic method and on the AHP multicriteria decision making method to identify the best network configuration. In the sequence, some fundamental considerations are predefined to perform automatic reconfiguration in real-time. 2.1. Real-time characteristics and requirements Reconfiguration of distribution network in normal operation usually has as main objective the reduction of energy losses. When considering automatic reconfiguration in real-time, some aspects have to be included in the study: a) it is necessary to establish a cost-benefit relationship to determine the necessity and the effectiveness of the reconfiguration; b) the network must be flexible to allow the reconfiguration; c) the technical feasibility of the switching should be studied in real-time, considering the availability of current measurements of the network. The diagram in Fig. 1 shows the architecture employed in this work to meet these premises.
Fig. 1. Architecture of the proposed system.
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
900
0,6
800
5
0,5
700
4 3 2
P(kW)
600
0,4
Cp
500
0,3
400 300
0,2
200
1
0,1
100 0
0 Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Power Coefficient (Cp)
b
6
Power (kW)
Wind Speed (m/s)
a
61
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Dec
Wind Speed (m/s)
Month
Fig. 2. (a) Wind resource (b) power curve for wind turbine Enercon E53.
The real-time information access is needed to implement the methodology, such as the data and communication status of the devices controlled remotely, which is obtained from SCADA. The reconfiguration software evaluates the load rates and generation scenarios of DG, and it performs the optimization method considering the heuristic search and multicriteria analysis based on AHP. The proposed system also sends commands to open and close the switches by SCADA. The steps of the presented architecture are discussed in Section 3.
3.2.2. Photovoltaic power systems The profile of photovoltaic generation was defined based on indices of sunshine and solar radiation incidence in the South region of Brazil. Fig. 4 illustrates the typical profile throughout the year. It was considered a set of PV modules with a total power of 0.5 MW. Fig. 5 shows the photovoltaic generation curve for a typical day with clear sky normalized to maximum generation (0.5 MW).
3. Methodology for automatic reconfiguration in real-time 3.1. Demand rate evaluation
3.2. Profile of distributed generation The DG incorporation considered in this paper refers to a moderate insertion of midrange DG with installed power of a few MW from renewable sources. The DG is assumed to be connected to the medium voltage of the distribution feeders. During the reconfiguration occurs the simultaneous analysis of the demand and generation curves. The program verifies the availability of DG power for injection into the system.
3.2.1. Wind power (W) In this work, it was considered a small wind farm and a typical generation scenario from historical data of average wind speed obtained by database of a weather station located in the South region of Brazil, with hourly measurements. Fig. 2(a) shows the monthly incidence of wind during the study period. The wind turbine model used was the Enercon E53, with rated power of 800 kW, as shown in the power curve in Fig. 2(b). From the meteorological parameters and the power curve of the turbine, it is possible to obtain the daily curve of the potential for wind power generation. Fig. 3 illustrates the wind power curve for a typical day with good wind incidence, normalized to the maximum generation value (800 kW).
Fig. 3. Wind generation curve.
8
Daily Radiation (kWh/m²/d)
The first step on the proposed methodology is to determine when a change in network configuration is actually needed. This avoids frequent switching which may cause equipment degradation. The strategy employed in this work is based on demand rates and load forecasting using typical curves for residential, industrial, commercial and rural consumers. In order to obtain a good discretization of the demand curve and to avoid frequent reconfigurations in the network, it is employed a set of six demand rates, as presented in work [13].
7 6 5 4 3 2 1 0 Jan
Feb
Mar
Apr
May
Jun Jul Aug Daily Radiation
Sep
Oct
Nov
Fig. 4. Monthly radiation profile throughout the year.
Fig. 5. Photovoltaic generation for a typical day with clear sky.
Dec
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
Subject to
1
|Ii | ≤ Ii max Vj min ≤ Vj ≤ Vj max
0,8
(2)
Pmin ≤ PDGn ≤ Pmax
0,6 0,4 0,2 0
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00
Small Hydroeletric Power (p.u)
62
Period (hour) Fig. 6. SHP curve generation.
3.2.3. Small hydroelectric power plant (SHP) The power plant operation depends on the river flow and on the reservoirs capacity, so there is also considerable uncertainty regarding this type of generation and the share that can be considered in the studies. The flow data were extracted from the database of an SHP, in the South region of Brazil. From the knowledge of the generation characteristics such as local water availability and SHP machines availability, it was possible to obtain the hydraulic power curve generated by the SHP. Fig. 6 illustrates the generation curve normalized to the maximum generation value (1 MW), for a typical day. Several researches that analyze DG interconnected with distribution networks consider that the DG unit is able to keep generating throughout of the peak demand period. However, this does not always happen because there are some uncertainty factors, such as changes in radiation and wind speed levels, hydrological risks and possibility of machines breakdowns due to the mechanical problems, causing the unavailability of generation. This work considers the behavior of generation sources, using the expected values for each period of analysis. In the worst-case scenarios with low rate of DGs generation availability (for example, solar energy in the period of light load) these ones do not participate in the reconfiguration process.
3.3. Heuristic reconfiguration technique using the potential generation of DG 3.3.1. Objective functions and constraints The objective function (OF) used in this work includes the minimization of three distinct functions: (i) Normalized Expected Loss of Energy in the primary network (ELossesi *) (ii) Normalized Expected System Average Interruption Frequency Index in the system (ESAIFIi *) and (iii) Normalized Expected value of Energy Not Supplied (EENSi *). The total gain obtained in each configuration combines the three functions and it is presented in (1): OF = min(ELossesi ∗ · w1 + ESAIFIi ∗ · w2 + EENSi ∗ · w3 )
(1)
where i corresponds to the period of analysis: 1. . .6, w1 . . .w3 are weights of the criteria in multicriteria method, Ii is current magnitude of each element must lie within its permissible limits (A), Vj is voltage magnitude of each node must lie within its permissible ranges (kV) and PDGn is active power delivered by the distributed generator n. The values of Pmin and Pmax depend on the DG technology considered (kW). In addition to these constraints the system should always keep radiality after reconfiguration of feeders and should not allow islanded operation of DG units. To calculate the load flow values in the distribution system, the proposed algorithm implements the backward/forward sweep method with DG [14,15], which is suitable for calculation of reverse flows from distributed generation. With load flow values, the energy losses for the distribution system are obtained. The calculation of reliability indicators ESAIFI and EENS is obtained from the equations of reliability for these indicators during the process of load flow [16]. 3.3.2. Weights of criteria The indicators of the OF are judged according to the decision making method AHP developed by Saaty [17]. The method is premised on the trial and the conversion of the criteria weights, so one can adjust the priority that each criterion has over the others. In [13] is presented the methodology in more details and calculations for setting the weights of the criteria and judgment matrices used in this work. The following results are assumed: w1 (ELosses) = 0.64; w2 (ESAIFI) = 0.26 and w3 (EENS) = 0.10. 3.3.3. Heuristic technique for selecting configurations The optimization technique employed in this work is based on the Branch Exchange adaptation proposed by [12]. This step is based on the analysis of the interconnections between the feeders, which are performed by normally open (NO) switches. For each interconnection, the best network configuration between two feeders is searched, by changing the status of a pair of switches. The method has the following advantages over other techniques from the literature: • it highly limits the search space of solutions, reducing calculation and computational processing time; • it does not require adjustments of optimization parameters due to changes in the network (such as grid expansion or installation of new equipment); • the radial network is guaranteed, without the need to verify if the combination of switches that comprise the topology tested is valid. The reconfiguration methodology was divided into two modules: (A) analysis of connection between feeders without DG
Fig. 7. Distribution network: (a) Original configuration. (b) Switching status changed.
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
63
Fig. 8. Switching status changed. Fig. 11. DG connected in feeder FD3.
Fig. 9. Switching status changed on the opposite direction.
connected to the network and (B) analysis of connection between feeders with DG. 3.3.3.1. Reconfiguration without DG on feeders. 1st Stage: Choose determined remotely controlled NO (normally opened) tie-switch at random, starting from the original configuration of the network (Fig. 7(a)). Go to the 2nd stage; 2nd Stage: Change the network configuration, closing an NO tie-switch and opening an NC (normally closed) switch of any of the two feeders involved. However, it should be the first switch upstream from the NO tie-switch (TS1), as shown in Fig. 7(b). Go to the 3rd stage; 3rd Stage: Perform the load flow and reliability calculation for this new configuration and check if there was a reduction of OF without violating the constraints. If OF is improved, proceed to the 4rd stage. Otherwise, go to the 5th stage. 4th Stage: Change the network configuration, closing the actual NO switch (S6) and opening the first NC switch upstream (S7), but in the same direction (feeder) of the previous iteration (Fig. 8). Go to the 6th stage. 5th Stage: Re-establish the initial configuration and change the network topology, closing the NO tie-switch and opening the NC switch of the other feeder upstream from the tie-switch (Fig. 9). Go to the 6th stage. 6th Stage: Change the network configuration in the same direction (feeder) while the OF is improved without violating the defined constraints. The process is finished when the OF does not improve or if any constraint violation occurs. In this case, the topology found in the previous iteration is the final solution. The same process must be repeated when another tie-switch is analyzed, always beginning from the initial configuration. The process is repeated until all remote controlled NO tie-switches without DG on feeders are analyzed.
Fig. 12. DG connected in feeder FD4.
3.3.3.2. Reconfiguration with DG on feeders. Assuming that distributed generation injects power into the grid, it is necessary to evaluate the gain of its connection to each feeder. Initially starts with the original configuration with DG, as shown in Fig. 10.
1st Stage: Test the gain of DG connection on the current feeder (FD3), closing the remotely controlled NO tie-switch (TS2) and opening the next NC switch (S14) downstream of tie-switch, as illustrated in Fig. 11. Go to the 2nd stage; 2nd Stage: Perform the load flow and reliability calculation for this new configuration and check if there was a reduction of OF without violating the constraints. If OF is improved, proceed to the 3rd stage. Otherwise, go to the 4th stage; 3rd Stage: Change the network configuration in the same direction (feeder) while the OF is improved without violating the defined constraints. Go to the 4th stage; 4th Stage: Test the connection of GD on the other feeder (FD4). From the original configuration with DG (Fig.), open the first NC switch (S13) upstream of DG and close the NO tie-switch (TS2), as illustrated in Fig. 12. Go to the 5th stage; 5th Stage: Perform the load flow and reliability calculation for this new configuration and check if there was a reduction of OF without violating the constraints. If the OF is improved, go to the 6th stage. Otherwise, the process of reconfiguration is finished, go to the 7th stage. 6th Stage: Change the network configuration in the same direction (feeder) while the OF is improved without violating the defined constraints (Fig. 13). The process of reconfiguration is finished when the OF does not improve or if any constraint violation occurs. Go to the 7th stage; 7th Stage: The configuration with the greatest reduction in the OF is chosen as best topology, comparing the results obtained in the connection of DG on feeder FD3 (3th Stage) and FD4 (6th Stage). The same process must be repeated when another source of DG is analyzed, always beginning from the initial configuration. The methodologies presented assume that only the remote controlled switches are analyzed and it consists in applying the branch exchange algorithm twice: first, to determine the individual result of each configuration test from the initial network topology and second, to determine the final sequence of configurations. In the
Fig. 10. Original configuration with distributed generation.
Fig. 13. Switching status changed.
64
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
Fig. 14. Network distribution with DG.
first case, the procedure is repeated for each normally open tieswitch starting from the initial network topology. At the end of this process it is applied the multicriteria decision method described in the previous section, to define the best sequence of switching. The procedure of optimization is then applied for the second time, but for each tie-switch tested, the analysis is made without returning to the initial configuration. 4. Results and discussion 4.1. Validation of the methodology An exhaustive search procedure was implemented in a simplified distribution network (Fig. 15), in order to assess the robustness of proposed method. The best solution for this network was previously obtained considering only the power losses minimization and compared with the proposed methodology, as shown in Table 1. Fig. 15 demonstrates a network with seven switches which leads to a total of 128 (27 ) possible combinations, and it is used to validate the proposed method. In actual networks is not feasible to apply the method of exhaustive search due to combinatorial explosion caused by the number of switching devices in the network. 4.2. Practical implementation of the methodology The proposed methodology was verified through several tests on the concession area of a power utility of Brazil. The real Table 1 Reconfiguration results for exhaustive search compared with the methodology. Parameter Initial solution loss (kW) Optimal solution loss (kW) Obtained loss (kW) Processing time (3.33 GHz)
Exhaustive search
34.72 32 s
Proposed method 37.74 34.72 34.72 8s
distribution network model presented in Fig. 14 is used as a case study. This network has 2 substations 69/13.8 kV, 5 feeders, 3 points of distributed generation, 15 tie-switches, 99 normally closed switches and over 21,000 consumers. The switches are remote controlled. The dotted lines represent the presence of normally open tie-switches (TS). The DG is composed of a SHP of 1 MW located at node 6, two wind turbines (E53) of 800 kW at node 33 and a photovoltaic park of 500 kW located at node 37. Data generation from these sources is updated for each load rate, in order to consider the power injection of DG on its current condition generation. Table 2 illustrates the test condition considered for each level of demand and each consumer class and generation scenario during the reconfiguration of distribution network. The reconfiguration algorithm is applied considering the individual analysis of each switch interconnection shown in Fig. 14. It evaluates the power injection of DG for each feeder involved based on the conditions of generation scenarios in Table 2. The individual analysis of the tie-switches leads to the results shown in Table 3 from original configuration of networks, comparing the results with and without DG for the period of 6:00 to 12:00. Only the cases where the objective functions analyzed presented positive evolution are shown. Note that in normal situations, the number of consumers of a feeder does not vary during the day, which suggests the same ESAIFI value in all the simulations. Each result in Table 3 is normalized by the maximum value found in the tests (Cn base ) as follows: Cn∗o pm =
Cno pm Cn base
(3)
After normalization, the individual results are multiplied by the weight vector and summed as in (1). The results are sorted by a global index (GI), from lowest to highest. This result indicates the sequence of switching for the best combination of results from
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
65
Fig. 15. Simplified distribution network with DG.
Table 2 Discretization curve of demand for consumer class and generation scenario of DG. Period – initial hour 0:00
6:00
12:00
13:00
17:00
21:00
Feeder class (demand)
Residential Industrial Commercial Rural
Low Low Low Low
Medium Heavy Heavy Medium
Low Medium Low Low
Medium Heavy Low Medium
Heavy Medium Heavy Heavy
Medium Low Low Medium
DG (generation)
Photovoltaic Windpower SHP
– Low Medium
Medium Medium Medium
Heavy Heavy Medium
Medium Heavy Medium
Low Medium Medium
– Low Medium
Table 3 Results for the individual analysis of tie-switches. Options: close/open
Rate 06:00–12:00 hour Initial configuration Option 1: TS-1/S3 Option 2: TS-2/S18 Option 3: TS-3/S12 Option 4: TS-5/S22 Option 5: TS-7/S59 Option 6: TS-8/S84 Option 7: TS-9/S63 Option 8: TS-10/S91
Configuration without DG
Configuration with DG
Losses (kWh)
EENS (MWh)
ESAIFI (fail./year)
Losses (kWh)
EENS (MWh)
ESAIFI (fail./year)
2263.32 2243.94 2265.89 2143.5 2266.98 2242.57 2037.01 2153.80 2011.90
603.20 590.90 596.30 554.40 603.50 577.70 569.80 581.20 567.10
11.78 11.30 11.46 9.93 11.74 11.66 11.75 11.62 11.67
2094.85 2079.20 2094.32 2010.91 2037.14 1983.50 1752.88 1928.56 1759.74
562.80 551.50 556.20 520.80 580.40 532.60 520.30 538.40 519.30
11.78 11.30 11.46 9.93 11.74 11.66 11.75 11.62 11.67
the individual switches interconnection analysis. Table 4 illustrates this procedure for the period of 6:00 to 12:00 with DG the system. The Branch Exchange algorithm is reapplied, following the sequence of operations from the GI ranking of Table 4. The
difference from the previous analysis is that the best configuration obtained with one tie-switch is maintained as the initial configuration to the following tie-switch to be tested. The final result of the optimization of the network to the rate demand analyzed is shown in Table 5.
Table 4 Normalized results, sequence and global index with AHP method. Options
Energy losses × 0.64
ESAIFI × 0.26
EENS × 0.10
Global index (GI)
Sequence
Rate 06:00–12:00 hour Option 1: TS-1/S3 Option 2: TS-2/S18 Option 3: TS-3/S12 Option 4: TS-5/S59 Option 5: TS-7/S59 Option 6: TS-8/S84 Option 7: TS-9/S63 Option 8: TS-10/S91
0.99 1.00 0.96 0.97 0.94 0.83 0.92 0.84
0.95 0.97 0.84 0.99 0.99 0.99 0.98 0.99
0.95 0.95 0.89 1.00 0.91 0.89 0.92 0.89
0.9796 0.9886 0.9232 0.9814 0.9551 0.8845 0.9383 0.8846
6 8 3 7 5 1 4 2
66
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67
Table 5 Final results of reconfiguration analysis with DG. Sequence of test
Switching with best result
Energy losses (kWh)
ESAIFI (failures/year)
EENS (MWh)
Rate 06:00–12:00 hour Initial configuration Option 6 Option 8 Option 3 Final reduction (%)
– Close TS-8/Open S84 Close TS-10/Open S91 Close TS-3/Open S12
2094.85 1752.88 1756.54 1672.60 20.15%
11.78 10.75 11.67 9.90 15.95%
632.90 520.30 520.4 478.30 15.01%
All procedures of the main program have been successful and the final topology of the network determined by the program improved the indicators of the objective functions of the optimization method, with no constraints in relation to the necessary switching. One way to confirm if the solution presented is the optimal solution is using the final configuration obtained by proposed methodology as input to a new test sequence. If there is no change in the topology, there is a strong indication that the configuration obtained is the optimal solution or very close to the optimal solution. The tests were performed in about 3 min in a PC with Intel Core 3.33 GHz and 24 GB RAM. The scale of few minutes is acceptable for the proposed application, since the reconfiguration is analyzed at the beginning of each demand rate transition and the demand rate interval is in the range of one or few hours.
5. Conclusions This paper presented a new methodology for automatic reconfiguration of power distribution networks incorporating distributed generation. The main contributions of the proposed system are: real-time reconfiguration based on load rate analysis, modeling and generation profile of DG from different sources (wind, solar and SHP); application of the AHP method for multicriteria decision making and integration of computational analysis with a Supervisory Control and Data Acquisition of remote controlled switches to allow performing automatic reconfiguration in real-time. The switching are performed automatically, in the sequence determined by the program. The results stand out the contribution of DG for significant improvement on network performance indicators, reducing losses and increasing reliability in the process of reconfiguration of distribution systems in which they are connected. For a real evaluation of the system performance, case studies were performed with real data from a power utility in different operating scenarios. The results indicate the feasibility of the proposed methodology, which can be adapted to other real systems with DG. With the advent of the Smart Grid, the profile of load curves of distribution feeders will be subject to a different dynamic behavior than currently occurs. Some features, such as greater use of distributed generation, demand response and charging of electric vehicles will require a rapid response of the network to the new generation and load scenarios. The automatic reconfiguration in real-time will help to improve network performance and to promote more efficient use of its resources in this new context.
Acknowledgments The authors would like to thank the technical and financial support of AES Sul Distribuidora Gaúcha de Energia SA, Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundac¸ão de Amparo à Pesquisa do Estado do Rio Grande do Sul
(FAPERGS) and Coordenac¸ão de Aperfeic¸oamento de Pessoal de Nível Superior (CAPES). Appendix A. Nomenclature
AHP ELossesi ESAIFIi EENSi i Ii
Analytic Hierarchy Process Expected Loss of Energy in the primary network Expected System Average Interruption Frequency INDEX Expected Energy Not Supplied corresponds to the period of analysis: 1. . .6 current magnitude of each element must lie within its permissible limits (A) normally closed NC NO normally opened objective function OF PDGn active power delivered by the distributed generator n SCADA Supervisory Control and Data Acquisition w1 . . .w3 weights of the criteria in multicriteria method voltage magnitude of each node (kV) Vj References
[1] R.E. Brown, Impact of smart grid on distribution system design, IEEE Power and Energy Society General Meeting – Conversion and Delivery of Electrical Energy in the 21st Century (2008) 1–4. [2] T. Thakur, Jaswanti, Study and characterization of power distribution network reconfiguration, in: IEEE Transmission & Distribution Conference and Exposition, Latin America, 2006, pp. 1–6. [3] M.S. Thomas, S. Arora, V.K. Chandna, Distribution automation leading to a smarter grid, IEEE Innovative Smart Grid Technologies (2011) 211–216. [4] D. Haughton, G.T. Heydt, Smart distribution system design: automatic reconfiguration for improved reliability, IEEE Power and Energy Society General Meeting (2010) 1–8. [5] M. Merdan, W. Lepuschitz, T. Strasser, F. Andren, Multi-agent system for self-optimizing power distribution grids, IEEE International Conference on Automation, Robotics and Applications (2011) 312–317. [6] J.-C. Wang, H.-D. Chiang, G.R. Darling, An efficient algorithm for real-time network reconfiguration in large scale unbalanced distribution systems, IEEE Transactions on Power Systems 11 (1996) 511–517. [7] E. López, H. Opazo, L. Garcia, P. Bastard, Online reconfiguration considering variability demand: applications to real networks, IEEE Transactions on Power Systems 19 (2004) 549–553. [8] A. Vargas, M.E. Samper, Real-time monitoring and economic dispatch of smart distribution grids: high performance algorithms for DMS applications, IEEE Transactions on Smart Grid 3 (2) (2012) 866–877. [9] R.S. Rao, K. Ravindra, K. Satish, S.V.L. Narasimham, Power loss minimization in distribution system using network reconfiguration in the presence of distributed generation, IEEE Transactions on Power Systems 28 (1) (2013) 317–325. [10] V.F. Martins, C.L.T. Borges, Active distribution network integrated planning incorporating distributed generation and load response uncertainties, IEEE Transactions on Power Systems 26 (4) (2011) 2164–2172. [11] G. Celli, E. Ghiani, F. Pilo, G.G. Soma, Reliability assessment in smart distribution networks, Electric Power Systems Research 104 (2013) 164–175. [12] D.P. Bernardon, V.J. Garcia, A.S.Q. Ferreira, L.N. Canha, Multicriteria distribution network reconfiguration considering subtransmission analysis, IEEE Transactions on Power Delivery 25 (4) (2010) 2684–2691. [13] L.L. Pfitscher, D.P. Bernardon, L.N. Canha, V.F. Montagner, V.J. Garcia, A.R. Abaide, Intelligent system for automatic reconfiguration of distribution network in real time, Electric Power Systems Research 97 (2013) 84–92.
D.P. Bernardon et al. / Electric Power Systems Research 107 (2014) 59–67 [14] W.H. Kersting, D.L. Mendive, An application of ladder network theory to the solution of three-phase radial load-flow problems, IEEE Power Engineering Society Winter Meeting A76044-8 (1976) 1–6. [15] L. Zhang, W. Tang, H. Guan, L. Zhang, The back/forward sweep-based power flow method for distribution networks with DGs, IEEE International Conference on Power Electronics and Intelligent Transportation System (2009) 145–149.
67
[16] R.E. Brown, Electric Power Distribution Reliability, second ed., CRC Press, New York, 2009, ISBN 978-0-8493-7567-5. [17] T.L. Saaty, The Analytic Hierarchy Process: Planning, Priority Setting, Resource Allocation, McGraw-Hill, New York, 1980.