- Email: [email protected]
Microgrids Reliability Evaluation with Renewable Distributed Generation and Storage Systems
Proceedings of the 18th World Congress The International Federation of Automatic Control Milano (Italy) August 28 - September 2, 2011
Microgrids Reliability Evaluation with Renewable Distributed Generation and Storage Systems Carmen L. T. Borges. Eduardo Cantarino.
Federal University of Rio de Janeiro, Electrical Engineering Program, Rio de Janeiro, Brazil (Tel: +55-21-25628027; e-mail: [email protected], [email protected]).
Abstract: This paper presents a model for reliability evaluation of Microgrids with Distributed Generation based on renewable energy resources. In this approach, DG participates in the network operation coordination as an active agent and not as a decentralized generation under which the utility has no control. Stochastic models are considered for representing generation availability related to wind energy and solar energy. The use of storage is explored as a means of reducing supply intermittency by a specific model for representing the charge state. The influence of microgrids on distribution network reliability is evaluated considering also the probability of success of the islanding operation and the impacts of the storage systems on the power availability. Keywords: Reliability, Microgrids, Smart Grids, Distributed Generation, Renewable Energy, Storage.
1. INTRODUCTION Distributed Generation (DG) is an important alternative considered for generation expansion planning. DG is characterized by small generating units, usually based on renewable resources, connected to distribution and subtransmission systems (Jenkins, N., Ekanayake, J. and Strbac, G., 2006). However, with the use of renewable energy based DG, the uncertainties involved in system planning and operation become larger. Uncertainties related to renewable DG are due to two main aspects: the intermittent nature of the primary energy source and the possible unavailability of the unit when it is required to generate. The combination of these aspects may lead to generation deficit, which can heavily compromise the security, reliability and quality of power supply. Recently, microgrids (Chowdhury, S. P., Crossley, P. and Chowdhury, S., 2009) are being considered as a way to maintain the supply of part of the network in islanded operation from the main system, in order to improve reliability. Microgrids can be understood as active distribution networks that have modern communication, protection and control systems that allow islanded operation, but still attending operative constraints. Thus, microgrids are strongly related to the concept of Smart Grids (Gellings, C. W., 2009). Microgrids can significantly improve customer supply reliability. However, they can provoke several impacts on system operation, network control and protection equipments, especially when associated with intermittent energy resources. The reliability study of microgrids is a very complex subject because there are several issues to be taken into account, each with its own peculiarity. For example, modelling all forms of generation available for use, protection, modes of operation 978-3-902661-93-7/11/$20.00 © 2011 IFAC
(islanded or connected to the grid), operation in parallel with the grid, synchronization, islanding, etc., are issues that need to be included in the study. Recently, some works related to the reliability of microgrids were published. In (Bollen, M., et al., 2009) the islanded operation of microgrids is discussed and new indicators are suggested, especially the ones related to reliability. In (Bagen, B. and Billinton, R., 2005) a simulation technique that uses deterministic and probabilistic criteria for reliability studies of isolated systems with wind power and solar generation in parallel with energy storage systems is showed. In (Leite, A. P., Borges, C. L. T. and Falcão, D. M., 2006) a probabilistic computational model for a Markovian representation of wind generation is presented, using time series of wind speed and turbine models. In (Park, J., et al., 2009) details about the reliability evaluation in systems with solar generation are presented. In (Costa, P. M. and Matos, M. A., 2005) the reliability of the system proposed in (Billinton, R. and Jonnavithula, S., 1996) with addiction of microgrids of constant generation is evaluated. In (Manwell, J. F., et al., 2006) models for many equipments for operational studies, like batteries, solar and wind generators can be found. Finally, in (Chowdhury, S. P., Crossley, P. and Chowdhury, S., 2009) concepts related to microgrids are presented and economic issues, reliability, deployment and energy quality are discussed. This paper presents a model for reliability evaluation of microgrids with distributed generation based on renewable energy resources. In this approach, DG participates in the network operation coordination as an active agent and not as a decentralized generation under which the utility has no control. Stochastic models are considered for representing generation availability related to wind energy and solar energy, among others. The use of storage is explored as a
11695
10.3182/20110828-6-IT-1002.01090
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
means of reducing supply intermittency of both the wind and solar energy generation. One of the biggest challenges in microgrids reliability evaluation is modelling the switching between the interconnected and islanded operation modes. The way islanding is treated in this paper aims to incorporate its effects on the Microgrid reliability. The success or failure of the microgrid islanding is incorporated through probabilities based on the switching process quality indicators. The reliability evaluation model is characterized by a hybrid model that combines the analytical method for network reliability assessment with Monte Carlo simulation for representing the uncertainties inherent to the study. Several changes in the network may be necessary in order to make microgrids operation feasible, such as rewiring due to current increase next to generators, changes in protection scheme due to short-circuit current increase, installation of intelligent electronic devices that ensure proper operation using smart grids technologies, etc. These changes imply in a large cost for the distribution utility and they only are justified if there is enough benefit to make them economically attractive. In this context, reliability studies become crucial in examining the feasibility of microgrids, since they can demonstrate part of or even the greatest benefits derived from their implantation. 2. MICROGRIDS Distribution networks that have DG connected to them, that are integrated to the functions of distribution control centers, are called Active Distribution Networks (ADN). In ADN, the DG becomes part of network operation coordination as an active agent and not as a decentralized generation under which the utility has no control. Microgrids can be understood, in broad lines, as ADN with intelligent communication, protection and control systems that allow its secure islanded operation, respecting all technical operative constraints. Thus, microgrids are strongly related to the smart grids concept. Microgrids can bring a significant improvement in the customer supply reliability, besides a lot of other benefits, such as voltage regulation, reactive power control, etc. However, nowadays, microgrids are still topic of studies or, in maximum, object of small pilot projects. Microgrids are not implanted in large scale yet, because of the lack of specific technical and regulatory standards for controlling its operation. This is necessary due to the strong impacts on the system operation coordination and on the control and protection equipments, especially when they are associated with DG based on intermittent sources of energy. Microgrids may be attractive both from the customer and the utility points of view, since they provide an increase in the energy supply reliability, resulting in an increase in the service continuity and, consequently, an increase in energy sales. However, their implantation can run into several technical obstacles and lots of challenges need to be overcome to make them technically and economically feasible. The reliability analysis of distribution network with microgrids is extremely complex due to many variables that
must be considered, such as DG based on different energy sources (wind, solar, fossil fuels, etc.), protection coordination, switching between operation modes (interconnected and islanded), etc. 3. DISTRIBUTED ENERGY RESOURCES MODELS The use of DG as an alternative to expand the supply capacity of the electrical system is being considered around the world. The use of small generating units, connected in a dispersed way at distribution/sub-transmission systems, can bring benefits to consumers and utilities, especially in places where there is deficiency of the electric energy transport system. However, the uncertainties involved in the planning and operation of the system become larger with the presence of DG. The impact of DG on reliability of electric systems depends mainly on two aspects:
the operational model of DG and the purpose of its connection to the system;
and the energy source in which the DG unit is based.
The source of energy in which the DG is based has fundamental influence on reliability. Units based on nonintermittent and storable energy sources, such as oil and biomass, can be more easily represented, since energy can always be considered available in reliability studies. The only issue considered in the unavailability of generation is the not scheduled failure of the generating unit. This kind of DG tends to be more reliable. On the other hand, the units based on intermittent and not storable energy sources, such as wind and solar power, require a more complex model in reliability studies, where the energy availability also needs to be represented. The unavailability of generation of DG can be caused by unavailability of energy, failure of unit generator or insufficient level of available energy. The availability model of energy normally requires an analysis of time series of measurements on the energy input (wind speed, sunlight, etc.). Then, the generation availability of DG must be modelled by the combination of the availability model of energy and the availability model of generators. 3.1 Wind Generation Model Wind energy can be considered as the renewable energy source with the most successful exploration in the world today. However, wind generation has disadvantages as a regular source of energy, and therefore, it is considered less reliable than conventional sources. The daily amount of energy available can vary widely and its use is limited to places of high and relatively constant winds. The connection of a growing number of wind farms to the electrical systems implies in the need to study their effects. The operating characteristics of the wind farm, heavily dependent on the local regime of wind, imply that the conventional power plant stochastic model is inappropriate to be applied to it.
11696
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
In (Leite, A. P., Borges, C. L. T. and Falcão, D. M., 2006) a reliability model that considers all factors that influence the generation of a wind farm was developed, modeling the stochastic behaviour of wind velocity, the operating behaviour of turbines and the characteristic of power generated by wind turbines. The model provides the estimated annual energy produced and calculates the wind farm reliability indices. In this paper, the availability of wind power is modelled through its cumulative distribution function (CDF) as in the non-sequential Monte Carlo simulation. 3.2 Solar Generation Model Solar energy is also presenting a quick growth and success usage. This is mostly due to being a clean energy. However, like the wind, its availability may be lower due to the intermittency inherent to sunlight. For example, in many regions, half the day there is little or even no generation due to variation of solar radiation. The solar radiation varies randomly depending on various aspects, like the weather being cloudy or not, the environmental conditions of the region, etc. In order to reduce this randomness, storage systems must be used to provide power at night, for example. Therefore, the conventional power plant stochastic models cannot be used either. In (Park, J., et al., 2009) a method for evaluating the reliability of a power system with solar generators is presented. The solar generation is modeled by a multi state model based on data obtained in Korea. In this paper, as for the wind generation, a CDF of power availability is used to represent the behavior of solar generation plants. 3.3 Storage System Model One way to reduce the intermittence in the supply of renewable energy sources is to use storage system (Bagen, B. and Billinton, R., 2005). Batteries are devices that convert chemical energy into electricity (DC) when they are in the discharge process and electricity into chemistry when they are charging. The types of batteries that are used in conjunction with solar and wind generators are: OPzS (ventilated), OPzV (sealed) and PVV (VRLA – Valve Regulated Lead Acid battery). All these batteries support long and deep discharges (discharges up to 80%). Generally, these batteries are 2V and they have a variety of capacities (2V – 1500Ah, for example). Batteries representation in reliability studies requires a particular model due to their operational characteristics: their behavior is not Markovian and as their state of charge depends on the system operation, it is difficult to set a model for their use in stochastic simulation. Furthermore, its lifetime, that depends on charge/discharge cycles that the battery is submitted (as it can be seen in Figure 1), should be taken into consideration. Besides, a possible failure not related to the lifetime, as a failure caused by a battery defect, must also be considered in the model.
Figure 1. Charge/discharge cycle versus depth of discharge In an analytical methods or independent simulation, it is hard to represent all battery features, because there is no time dependency. Thus, representing its state of charge or its lifetime, for example, loses the sense, since it depends on its previous state. But, if needed, this situation can be looked from another point of view. It is possible to calculate how much of lifetime is lost in each fault or how much discharge was given in the battery, using the battery energy used during its operation time. In this paper, it was chosen not to take these peculiarities into account. It is just verified the impact of putting batteries in parallel with solar and wind generators in order to supply the energy that generators cannot supply for any reason. For such, the battery availability CDF is used in order to quantify the influence in the reliability indices, if the battery were not available 100% of the time. In addition, it is assumed that the batteries are correctly sized for supplying the load that they are responsible for during all the network failure duration. 4. MICROGRIDS RELIABILITY EVALUATION The reliability evaluation is performed in this paper using a hybrid method that combines an analytical method with random sampling techniques. This allows to consider the impact of intermittent energy sources (solar and wind power) in the system reliability indices. The indices calculated were:
SAIDI: System Average Interruption Duration Index (h/yr-customer);
SAIFI: System Average Interruption Frequency Index (int./yr-customer);
ASAI: Average Service Availability Index;
ENS: Energy Not Supplied (kWh/yr);
AENS: Average Energy Not customer).
Supplied
(kWh/yr-
Since the basis is the analytical method, the simulation is performed as follow:
11697
1.
Sample the generation availability based on the CDF of each energy source (wind, solar, etc.);
2.
With the generation sampled values, calculate the network reliability indices using the analytical method;
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
3.
This process is repeated until the coefficient of variation (α) is less than 5% for all indices or the maximum number of iterations is reached;
4.
At the end, the average values are calculated for each reliability index.
To avoid a premature convergence caused by a sequence of favorable samples, a minimum number of iterations are disregarded before starting to calculate the coefficient of variation. The number used in the tests is 10 iterations. For example, in any iteration, all generation values are sampled. With these values, failure rates and unavailability are calculated for each load point. When all points were calculated, a value for SAIDI, SAIFI, ASAI, ENS and AENS are calculated for the system at that iteration. It is verified if the coefficient of variation is less than 5% for all indices. If so, it is calculated the mean of the values obtained for each index. If not, the generations are sampled again and the process is repeated.
Figure 2. RBTS-Bus5 system with microgrids.
The islanding of part of the distribution network (microgrid formation) and the microgrid reconnection to the main system can be considered as one of the greatest current challenges in microgrids reliability studies. In general, the reliability studies evaluate the impact of network components or DG failures in the microgrid operation without exploring the process of islanding and reconnection to the main system. The way islanding is addressed in this paper is through the success or failure probability of the microgrid islanding process. This approach is adopted given the high degree of complexity and stochastic data requirements to incorporate dynamic analysis in reliability studies. Moreover, this approach allows highlighting the improvements that must be made to the network electronic devices in order to enable successful islands formation. Currently, solar systems, for example, have a low probability of forming or maintaining a microgrid (Verhoeven, B., 2002). The dynamics analysis of switching between modes of operation can be performed using transient and/or voltage stability simulators to complement the reliability study.
Table 1. Load point data. Load point
Load level (kW)
Number of customers
1,2,20,21
762.5
210
3,5,8,17,23
1110
1
4,6,15,25
745
240
7,14,18,22,24
740
15
9,10,11,13,26
574
195
12,16,19
616.7
1
Table 2. MV network stochastic data. MV Branch 1,6,9,13,14,18,21,25,27,31,35,36,39,42 4,7,8,12,15,16,19,22,26,28,30,33,37,40 2,3,5,10,11,17,20,23,24,29,32,34,38,41,43
Failure rate Mean time to (f/yr) repair (h) 0.020 0.026 0.032
30 30 30
The duration for fault isolation and closing of switches S4 and S12 are considered to be 3.5h and 3h, respectively.
5. RESULTS 5.1 Test System and Studies Conditions The test system used is the medium voltage (MV) distribution network based on RBTS-Bus5 (Billinton, R. and Jonnavithula, S., 1996) containing microgrids with DG based on renewable energy sources, as shown in Figure 2. The load point data are shown in Table 1 and the network elements failure rate and mean time to repair are shown in Table 2. The DG units are connected at the low voltage (LV) distribution network. The normally opened (NO) switches (S4 and S12) are closed in case of failure in order to transfer load only if the feeder that is taking the failed feeder loads has sufficient capacity. The capacity of each feeder is considered to be 1.3 times the sum of their own loads (with switches S4 and S12 opened).
The wind and solar power installed capacity at each microgrid are 1,500 kW and 2,000 kW, respectively. These installed capacities have been adopted in order to provide a considerable probability of supplying the entire microgrid load, since these sources have low capacity factors. The installed capacity is considered to be dispersed along the LV network. The capacity of the storage system is 300 kW. The total load of the system is 20,000 kW. The generation availability CDF for wind and solar power plants were obtained from (Leite, A. P., Borges, C. L. T. and Falcão, D. M., 2006) and (Park, J., et al., 2009), respectively. In the solar energy case, the original generation values were divided by 5 for adjusting to the system load demand. This means that less photovoltaic cells are being used for electric
11698
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
energy generation while the probability of solar radiation remains the same. Initially, the microgrid islanding success probability is considered to be 100% and four different assessments are made considering simple contingencies:
1. Case 1: Without having microgrids; 2. Case 2: With microgrids having constant generation equal to its load;
3. Case 3: With microgrids having random generation available from wind and solar power sources;
4. Case 4: With microgrids having random generation
large part of the load. This resulted in approximating the indices of Case 4 to the ones of Case 2. By changing the battery power and/or its probability of being able to supply the required load (here it was used a probability of 95% of success), the indices can approximate the ones of Case 1, if these values are reduced, or they can approximate even more the indices of Case 2, if they are increased. In Case 3, the randomness of the generation sources values caused difficulties for obtaining indices with the desired coefficient of variation. As Figure 3 shows, there is a stabilization of the coefficient of variation around these values, independently of the increase in the number of samples. The final standard deviation of each index of Case 3 is shown in Table 5.
available from wind and solar power sources and using storage systems at both of them.
Afterwards, a sensitive analysis of the islanding success probability influence in the reliability indices is made, considering only Case 4. 5.2 Microgrid and Renewable Energy Influence on Reliability In the evaluations without microgrids and with microgrids of constant generation (Cases 1 and 2), the analytical method was used for calculating the reliability indices. In the evaluation of microgrids based on wind and solar power generation (Cases 3 e 4), the hybrid method combining the analytical method with random sampling of the generation availability was used. For the cases where random sampling was used, the maximum number of samples was 80,000. Table 3 presents the indices calculated and Table 4 presents the coefficient of variation for each index for Cases 3 and 4. Table 3. System indices for each case. SAIDI
SAIFI
ENS
AENS
ASAI
Case 1
4.8411
0.2847
101,890
35.65
0.9994474
Case 2
2.6143
0.1946
65,598
22.95
0.9997016
Case 3
3.5825
0.2336
86,871
30.40
0.9995910
Case 4
2.7276
0.1949
74,785
26.17
0.9996886
a - SAIDI
a - SAIFI
a -ENS
a - AENS
a - ASAI
0.1206
0.0796
0.0661
0.0661
0.0000049
Case 4
0.0321
0.0029
0.0453
0.0453
0.0000010
Table 5. Standard deviation of each index for Case 3. Standard deviation
SAIDI
SAIFI
ENS
AENS
ASAI
0.4318
0.0186
5,743
2.0095
0.0000493
5.3 Islanding Success Influence on Reliability
Table 4. Coefficient of variation for each system index. Case 3
Figure 3. Evolution of the coefficients of variation of SAIDI, SAIFI and ENS for Case 3.
Comparing the other cases with Case 1, the benefit of having microgrids is clearly shown.
In order to assess the islanding influence on the microgrid reliability, Case 4 was re-evaluated by varying the success probability of the island formation from the previous value of 100%. Table 6 presents the indices calculated for different values of islanding success probability and Table 7 shows the coefficient of variation for each index. It is worthwhile to highlight that the case with 100% probability corresponds to Case 4 while the case with 0% corresponds to Case 1.
Cases 1 and 2 represent the upper and lower bounds, respectively, for the indices that are expected for the other cases, since in Case 2 the capacity of the DG is considered as always available. However, Cases 3 and 4 provide more realistic indices since they consider the variation of energy availability of the wind and solar plants. Comparing the indices of Cases 3 and 4, it can be noticed that the battery provides a considerable gain in reliability. For the system under study, the value considered for battery power allows the microgrid to supply, most of the time, the entire load or a 11699
Table 6. System indices for different islanding success probability. SAIDI
SAIFI
ENS
AENS
ASAI
100 % success
2.7276
0.1949
74,785
26.17
0.9996886
75% success
3.3053
0.2188
83,739
29.30
0.9996227
50% success
3.8187
0.2408
89,916
31.46
0.9995641
25% success
4.3288
0.2627
95,960
33.58
0.9995058
0% success
4.8411
0.2847
101,890
35.65
0.9994474
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
behaviour of the generation, extracting a higher benefit from microgrids implementation.
Table 7. Coefficient of variation for different islanding success probabilities. a - SAIDI
a - SAIFI
a -ENS
a - AENS
a - ASAI
100 % success
0.0321
0.0029
0.0453
0.0453
0.0000010
75% success
0.1324
0.0872
0.0655
0.0655
0.0000050
50% success
0.1286
0.0896
0.0621
0.0621
0.0000561
25% success
0.0976
0.0708
0.0474
0.0474
0.0000048
The increase in the islanding success probability implies in an improvement of the reliability indices. For example, varying the islanding success probability from 25% to 75% implies in a gain of 24% in SAIDI, 17% in SAIFI and 13% in ENS. These improvements are explained by the fact that the microgrid can supply the island load more times, thus guaranteeing the service continuity for longer. However, to enhance the islanding success, efforts must be concentrated on best control and protection devices of the microgrids, what, on the other hand, implies in higher costs. As occurred for Case 3, the desired coefficient of variation wasn’t reached for the cases with lower islanding success probability. The reason is also the increase in the randomness of the simulation when the uncertainty about the islanding success is included. As an example, Figure 4 shows the evolution of the coefficient of variation for the case of 50% islanding success probability, where the stabilization of the coefficient of variation around these values is clearly shown.
Figure 4. Evolution of the coefficients of variation for 50% islanding success probability. 6. CONCLUSIONS The hybrid method for reliability assessment of microgrid containing renewable energy sources and storage systems was effective for a rapid analysis, incorporating the intermittency of the renewable energies and the islanding process success on the system reliability indices.
Another important conclusion is that ensuring that the islanding process is successful is what can really bring a large impact on the microgrid reliability. Currently, generations such as wind and solar power have a small probability of maintaining an island. However, with the application of smart grid technologies, which will be a reality in the near future, the microgrid formation success may be guaranteed, as well as the success of the reconnection to the main grid. REFERENCES Bagen, B. and Billinton, R. (2005), Incorporating Well-Being Considerations in Generating Systems Using Energy Storage, IEEE Transactions on Energy Conversion, v. 20, n. 1. Billinton, R. and Jonnavithula, S. (1996), A Test System for Teaching Overall Power System Reliability Assessment, IEEE Transactions on Power Systems, v. 11, n. 4. Bollen, M., et al. (2009), Performance indicators for microgrids during grid-connected and island operation, IEEE Power Tech Conference. Chowdhury, S. P., Crossley, P. and Chowdhury, S. (2009), Microgrids and Active Distribution Networks, The IET, London. Costa, P. M. and Matos, M. A. (2005), Reliability of Distribution Networks with Microgrids, IEEE Power Tech Conference. Gellings, C. W. (2009), The Smart Grid: Enabling Energy Efficiency and Demand Response, CRC Press. Jenkins, N., Ekanayake, J. and Strbac, G. (2006), Distributed Generation, The IET, London. Leite, A. P., Borges, C. L. T. and Falcão, D. M.(2006), Probabilistic Wind Farms Generation Model for Reliability Studies applied to Brazilian Sites, IEEE Transactions on Power Systems, v. 21, n. 4, p. 14931501, 2006. Manwell, J. F., et al. (2006), Hybrid2 – A Hybrid System Simulation Model – Theory Manual, National Renewable Energy Laboratory. Park, J., et al. (2009), A Probabilistic Reliability Evaluation of A Power System Including Solar/Photovoltaic Cell Generator, IEEE Transactions on Power Systems, v. 24, n. 2. Verhoeven, B. (2002), Probability of Islanding in utility networks due to grid connected photovoltaic power systems, Report IEA-PVPS T5-07: 2002.
Microgrids implementation at distribution systems brings significant improvements when the concerns are about reliability. However, if they are maintained by generators that depend on random energy, as the case of solar and wind power, the benefits on the reliability indices are reduced. Nevertheless, the installation of storage systems together with intermittent sources can reduce drastically the random
11700
Microgrids Reliability Evaluation with Renewable Distributed Generation and Storage Systems Carmen L. T. Borges. Eduardo Cantarino.
Federal University of Rio de Janeiro, Electrical Engineering Program, Rio de Janeiro, Brazil (Tel: +55-21-25628027; e-mail: [email protected], [email protected]).
Abstract: This paper presents a model for reliability evaluation of Microgrids with Distributed Generation based on renewable energy resources. In this approach, DG participates in the network operation coordination as an active agent and not as a decentralized generation under which the utility has no control. Stochastic models are considered for representing generation availability related to wind energy and solar energy. The use of storage is explored as a means of reducing supply intermittency by a specific model for representing the charge state. The influence of microgrids on distribution network reliability is evaluated considering also the probability of success of the islanding operation and the impacts of the storage systems on the power availability. Keywords: Reliability, Microgrids, Smart Grids, Distributed Generation, Renewable Energy, Storage.
1. INTRODUCTION Distributed Generation (DG) is an important alternative considered for generation expansion planning. DG is characterized by small generating units, usually based on renewable resources, connected to distribution and subtransmission systems (Jenkins, N., Ekanayake, J. and Strbac, G., 2006). However, with the use of renewable energy based DG, the uncertainties involved in system planning and operation become larger. Uncertainties related to renewable DG are due to two main aspects: the intermittent nature of the primary energy source and the possible unavailability of the unit when it is required to generate. The combination of these aspects may lead to generation deficit, which can heavily compromise the security, reliability and quality of power supply. Recently, microgrids (Chowdhury, S. P., Crossley, P. and Chowdhury, S., 2009) are being considered as a way to maintain the supply of part of the network in islanded operation from the main system, in order to improve reliability. Microgrids can be understood as active distribution networks that have modern communication, protection and control systems that allow islanded operation, but still attending operative constraints. Thus, microgrids are strongly related to the concept of Smart Grids (Gellings, C. W., 2009). Microgrids can significantly improve customer supply reliability. However, they can provoke several impacts on system operation, network control and protection equipments, especially when associated with intermittent energy resources. The reliability study of microgrids is a very complex subject because there are several issues to be taken into account, each with its own peculiarity. For example, modelling all forms of generation available for use, protection, modes of operation 978-3-902661-93-7/11/$20.00 © 2011 IFAC
(islanded or connected to the grid), operation in parallel with the grid, synchronization, islanding, etc., are issues that need to be included in the study. Recently, some works related to the reliability of microgrids were published. In (Bollen, M., et al., 2009) the islanded operation of microgrids is discussed and new indicators are suggested, especially the ones related to reliability. In (Bagen, B. and Billinton, R., 2005) a simulation technique that uses deterministic and probabilistic criteria for reliability studies of isolated systems with wind power and solar generation in parallel with energy storage systems is showed. In (Leite, A. P., Borges, C. L. T. and Falcão, D. M., 2006) a probabilistic computational model for a Markovian representation of wind generation is presented, using time series of wind speed and turbine models. In (Park, J., et al., 2009) details about the reliability evaluation in systems with solar generation are presented. In (Costa, P. M. and Matos, M. A., 2005) the reliability of the system proposed in (Billinton, R. and Jonnavithula, S., 1996) with addiction of microgrids of constant generation is evaluated. In (Manwell, J. F., et al., 2006) models for many equipments for operational studies, like batteries, solar and wind generators can be found. Finally, in (Chowdhury, S. P., Crossley, P. and Chowdhury, S., 2009) concepts related to microgrids are presented and economic issues, reliability, deployment and energy quality are discussed. This paper presents a model for reliability evaluation of microgrids with distributed generation based on renewable energy resources. In this approach, DG participates in the network operation coordination as an active agent and not as a decentralized generation under which the utility has no control. Stochastic models are considered for representing generation availability related to wind energy and solar energy, among others. The use of storage is explored as a
11695
10.3182/20110828-6-IT-1002.01090
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
means of reducing supply intermittency of both the wind and solar energy generation. One of the biggest challenges in microgrids reliability evaluation is modelling the switching between the interconnected and islanded operation modes. The way islanding is treated in this paper aims to incorporate its effects on the Microgrid reliability. The success or failure of the microgrid islanding is incorporated through probabilities based on the switching process quality indicators. The reliability evaluation model is characterized by a hybrid model that combines the analytical method for network reliability assessment with Monte Carlo simulation for representing the uncertainties inherent to the study. Several changes in the network may be necessary in order to make microgrids operation feasible, such as rewiring due to current increase next to generators, changes in protection scheme due to short-circuit current increase, installation of intelligent electronic devices that ensure proper operation using smart grids technologies, etc. These changes imply in a large cost for the distribution utility and they only are justified if there is enough benefit to make them economically attractive. In this context, reliability studies become crucial in examining the feasibility of microgrids, since they can demonstrate part of or even the greatest benefits derived from their implantation. 2. MICROGRIDS Distribution networks that have DG connected to them, that are integrated to the functions of distribution control centers, are called Active Distribution Networks (ADN). In ADN, the DG becomes part of network operation coordination as an active agent and not as a decentralized generation under which the utility has no control. Microgrids can be understood, in broad lines, as ADN with intelligent communication, protection and control systems that allow its secure islanded operation, respecting all technical operative constraints. Thus, microgrids are strongly related to the smart grids concept. Microgrids can bring a significant improvement in the customer supply reliability, besides a lot of other benefits, such as voltage regulation, reactive power control, etc. However, nowadays, microgrids are still topic of studies or, in maximum, object of small pilot projects. Microgrids are not implanted in large scale yet, because of the lack of specific technical and regulatory standards for controlling its operation. This is necessary due to the strong impacts on the system operation coordination and on the control and protection equipments, especially when they are associated with DG based on intermittent sources of energy. Microgrids may be attractive both from the customer and the utility points of view, since they provide an increase in the energy supply reliability, resulting in an increase in the service continuity and, consequently, an increase in energy sales. However, their implantation can run into several technical obstacles and lots of challenges need to be overcome to make them technically and economically feasible. The reliability analysis of distribution network with microgrids is extremely complex due to many variables that
must be considered, such as DG based on different energy sources (wind, solar, fossil fuels, etc.), protection coordination, switching between operation modes (interconnected and islanded), etc. 3. DISTRIBUTED ENERGY RESOURCES MODELS The use of DG as an alternative to expand the supply capacity of the electrical system is being considered around the world. The use of small generating units, connected in a dispersed way at distribution/sub-transmission systems, can bring benefits to consumers and utilities, especially in places where there is deficiency of the electric energy transport system. However, the uncertainties involved in the planning and operation of the system become larger with the presence of DG. The impact of DG on reliability of electric systems depends mainly on two aspects:
the operational model of DG and the purpose of its connection to the system;
and the energy source in which the DG unit is based.
The source of energy in which the DG is based has fundamental influence on reliability. Units based on nonintermittent and storable energy sources, such as oil and biomass, can be more easily represented, since energy can always be considered available in reliability studies. The only issue considered in the unavailability of generation is the not scheduled failure of the generating unit. This kind of DG tends to be more reliable. On the other hand, the units based on intermittent and not storable energy sources, such as wind and solar power, require a more complex model in reliability studies, where the energy availability also needs to be represented. The unavailability of generation of DG can be caused by unavailability of energy, failure of unit generator or insufficient level of available energy. The availability model of energy normally requires an analysis of time series of measurements on the energy input (wind speed, sunlight, etc.). Then, the generation availability of DG must be modelled by the combination of the availability model of energy and the availability model of generators. 3.1 Wind Generation Model Wind energy can be considered as the renewable energy source with the most successful exploration in the world today. However, wind generation has disadvantages as a regular source of energy, and therefore, it is considered less reliable than conventional sources. The daily amount of energy available can vary widely and its use is limited to places of high and relatively constant winds. The connection of a growing number of wind farms to the electrical systems implies in the need to study their effects. The operating characteristics of the wind farm, heavily dependent on the local regime of wind, imply that the conventional power plant stochastic model is inappropriate to be applied to it.
11696
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
In (Leite, A. P., Borges, C. L. T. and Falcão, D. M., 2006) a reliability model that considers all factors that influence the generation of a wind farm was developed, modeling the stochastic behaviour of wind velocity, the operating behaviour of turbines and the characteristic of power generated by wind turbines. The model provides the estimated annual energy produced and calculates the wind farm reliability indices. In this paper, the availability of wind power is modelled through its cumulative distribution function (CDF) as in the non-sequential Monte Carlo simulation. 3.2 Solar Generation Model Solar energy is also presenting a quick growth and success usage. This is mostly due to being a clean energy. However, like the wind, its availability may be lower due to the intermittency inherent to sunlight. For example, in many regions, half the day there is little or even no generation due to variation of solar radiation. The solar radiation varies randomly depending on various aspects, like the weather being cloudy or not, the environmental conditions of the region, etc. In order to reduce this randomness, storage systems must be used to provide power at night, for example. Therefore, the conventional power plant stochastic models cannot be used either. In (Park, J., et al., 2009) a method for evaluating the reliability of a power system with solar generators is presented. The solar generation is modeled by a multi state model based on data obtained in Korea. In this paper, as for the wind generation, a CDF of power availability is used to represent the behavior of solar generation plants. 3.3 Storage System Model One way to reduce the intermittence in the supply of renewable energy sources is to use storage system (Bagen, B. and Billinton, R., 2005). Batteries are devices that convert chemical energy into electricity (DC) when they are in the discharge process and electricity into chemistry when they are charging. The types of batteries that are used in conjunction with solar and wind generators are: OPzS (ventilated), OPzV (sealed) and PVV (VRLA – Valve Regulated Lead Acid battery). All these batteries support long and deep discharges (discharges up to 80%). Generally, these batteries are 2V and they have a variety of capacities (2V – 1500Ah, for example). Batteries representation in reliability studies requires a particular model due to their operational characteristics: their behavior is not Markovian and as their state of charge depends on the system operation, it is difficult to set a model for their use in stochastic simulation. Furthermore, its lifetime, that depends on charge/discharge cycles that the battery is submitted (as it can be seen in Figure 1), should be taken into consideration. Besides, a possible failure not related to the lifetime, as a failure caused by a battery defect, must also be considered in the model.
Figure 1. Charge/discharge cycle versus depth of discharge In an analytical methods or independent simulation, it is hard to represent all battery features, because there is no time dependency. Thus, representing its state of charge or its lifetime, for example, loses the sense, since it depends on its previous state. But, if needed, this situation can be looked from another point of view. It is possible to calculate how much of lifetime is lost in each fault or how much discharge was given in the battery, using the battery energy used during its operation time. In this paper, it was chosen not to take these peculiarities into account. It is just verified the impact of putting batteries in parallel with solar and wind generators in order to supply the energy that generators cannot supply for any reason. For such, the battery availability CDF is used in order to quantify the influence in the reliability indices, if the battery were not available 100% of the time. In addition, it is assumed that the batteries are correctly sized for supplying the load that they are responsible for during all the network failure duration. 4. MICROGRIDS RELIABILITY EVALUATION The reliability evaluation is performed in this paper using a hybrid method that combines an analytical method with random sampling techniques. This allows to consider the impact of intermittent energy sources (solar and wind power) in the system reliability indices. The indices calculated were:
SAIDI: System Average Interruption Duration Index (h/yr-customer);
SAIFI: System Average Interruption Frequency Index (int./yr-customer);
ASAI: Average Service Availability Index;
ENS: Energy Not Supplied (kWh/yr);
AENS: Average Energy Not customer).
Supplied
(kWh/yr-
Since the basis is the analytical method, the simulation is performed as follow:
11697
1.
Sample the generation availability based on the CDF of each energy source (wind, solar, etc.);
2.
With the generation sampled values, calculate the network reliability indices using the analytical method;
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
3.
This process is repeated until the coefficient of variation (α) is less than 5% for all indices or the maximum number of iterations is reached;
4.
At the end, the average values are calculated for each reliability index.
To avoid a premature convergence caused by a sequence of favorable samples, a minimum number of iterations are disregarded before starting to calculate the coefficient of variation. The number used in the tests is 10 iterations. For example, in any iteration, all generation values are sampled. With these values, failure rates and unavailability are calculated for each load point. When all points were calculated, a value for SAIDI, SAIFI, ASAI, ENS and AENS are calculated for the system at that iteration. It is verified if the coefficient of variation is less than 5% for all indices. If so, it is calculated the mean of the values obtained for each index. If not, the generations are sampled again and the process is repeated.
Figure 2. RBTS-Bus5 system with microgrids.
The islanding of part of the distribution network (microgrid formation) and the microgrid reconnection to the main system can be considered as one of the greatest current challenges in microgrids reliability studies. In general, the reliability studies evaluate the impact of network components or DG failures in the microgrid operation without exploring the process of islanding and reconnection to the main system. The way islanding is addressed in this paper is through the success or failure probability of the microgrid islanding process. This approach is adopted given the high degree of complexity and stochastic data requirements to incorporate dynamic analysis in reliability studies. Moreover, this approach allows highlighting the improvements that must be made to the network electronic devices in order to enable successful islands formation. Currently, solar systems, for example, have a low probability of forming or maintaining a microgrid (Verhoeven, B., 2002). The dynamics analysis of switching between modes of operation can be performed using transient and/or voltage stability simulators to complement the reliability study.
Table 1. Load point data. Load point
Load level (kW)
Number of customers
1,2,20,21
762.5
210
3,5,8,17,23
1110
1
4,6,15,25
745
240
7,14,18,22,24
740
15
9,10,11,13,26
574
195
12,16,19
616.7
1
Table 2. MV network stochastic data. MV Branch 1,6,9,13,14,18,21,25,27,31,35,36,39,42 4,7,8,12,15,16,19,22,26,28,30,33,37,40 2,3,5,10,11,17,20,23,24,29,32,34,38,41,43
Failure rate Mean time to (f/yr) repair (h) 0.020 0.026 0.032
30 30 30
The duration for fault isolation and closing of switches S4 and S12 are considered to be 3.5h and 3h, respectively.
5. RESULTS 5.1 Test System and Studies Conditions The test system used is the medium voltage (MV) distribution network based on RBTS-Bus5 (Billinton, R. and Jonnavithula, S., 1996) containing microgrids with DG based on renewable energy sources, as shown in Figure 2. The load point data are shown in Table 1 and the network elements failure rate and mean time to repair are shown in Table 2. The DG units are connected at the low voltage (LV) distribution network. The normally opened (NO) switches (S4 and S12) are closed in case of failure in order to transfer load only if the feeder that is taking the failed feeder loads has sufficient capacity. The capacity of each feeder is considered to be 1.3 times the sum of their own loads (with switches S4 and S12 opened).
The wind and solar power installed capacity at each microgrid are 1,500 kW and 2,000 kW, respectively. These installed capacities have been adopted in order to provide a considerable probability of supplying the entire microgrid load, since these sources have low capacity factors. The installed capacity is considered to be dispersed along the LV network. The capacity of the storage system is 300 kW. The total load of the system is 20,000 kW. The generation availability CDF for wind and solar power plants were obtained from (Leite, A. P., Borges, C. L. T. and Falcão, D. M., 2006) and (Park, J., et al., 2009), respectively. In the solar energy case, the original generation values were divided by 5 for adjusting to the system load demand. This means that less photovoltaic cells are being used for electric
11698
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
energy generation while the probability of solar radiation remains the same. Initially, the microgrid islanding success probability is considered to be 100% and four different assessments are made considering simple contingencies:
1. Case 1: Without having microgrids; 2. Case 2: With microgrids having constant generation equal to its load;
3. Case 3: With microgrids having random generation available from wind and solar power sources;
4. Case 4: With microgrids having random generation
large part of the load. This resulted in approximating the indices of Case 4 to the ones of Case 2. By changing the battery power and/or its probability of being able to supply the required load (here it was used a probability of 95% of success), the indices can approximate the ones of Case 1, if these values are reduced, or they can approximate even more the indices of Case 2, if they are increased. In Case 3, the randomness of the generation sources values caused difficulties for obtaining indices with the desired coefficient of variation. As Figure 3 shows, there is a stabilization of the coefficient of variation around these values, independently of the increase in the number of samples. The final standard deviation of each index of Case 3 is shown in Table 5.
available from wind and solar power sources and using storage systems at both of them.
Afterwards, a sensitive analysis of the islanding success probability influence in the reliability indices is made, considering only Case 4. 5.2 Microgrid and Renewable Energy Influence on Reliability In the evaluations without microgrids and with microgrids of constant generation (Cases 1 and 2), the analytical method was used for calculating the reliability indices. In the evaluation of microgrids based on wind and solar power generation (Cases 3 e 4), the hybrid method combining the analytical method with random sampling of the generation availability was used. For the cases where random sampling was used, the maximum number of samples was 80,000. Table 3 presents the indices calculated and Table 4 presents the coefficient of variation for each index for Cases 3 and 4. Table 3. System indices for each case. SAIDI
SAIFI
ENS
AENS
ASAI
Case 1
4.8411
0.2847
101,890
35.65
0.9994474
Case 2
2.6143
0.1946
65,598
22.95
0.9997016
Case 3
3.5825
0.2336
86,871
30.40
0.9995910
Case 4
2.7276
0.1949
74,785
26.17
0.9996886
a - SAIDI
a - SAIFI
a -ENS
a - AENS
a - ASAI
0.1206
0.0796
0.0661
0.0661
0.0000049
Case 4
0.0321
0.0029
0.0453
0.0453
0.0000010
Table 5. Standard deviation of each index for Case 3. Standard deviation
SAIDI
SAIFI
ENS
AENS
ASAI
0.4318
0.0186
5,743
2.0095
0.0000493
5.3 Islanding Success Influence on Reliability
Table 4. Coefficient of variation for each system index. Case 3
Figure 3. Evolution of the coefficients of variation of SAIDI, SAIFI and ENS for Case 3.
Comparing the other cases with Case 1, the benefit of having microgrids is clearly shown.
In order to assess the islanding influence on the microgrid reliability, Case 4 was re-evaluated by varying the success probability of the island formation from the previous value of 100%. Table 6 presents the indices calculated for different values of islanding success probability and Table 7 shows the coefficient of variation for each index. It is worthwhile to highlight that the case with 100% probability corresponds to Case 4 while the case with 0% corresponds to Case 1.
Cases 1 and 2 represent the upper and lower bounds, respectively, for the indices that are expected for the other cases, since in Case 2 the capacity of the DG is considered as always available. However, Cases 3 and 4 provide more realistic indices since they consider the variation of energy availability of the wind and solar plants. Comparing the indices of Cases 3 and 4, it can be noticed that the battery provides a considerable gain in reliability. For the system under study, the value considered for battery power allows the microgrid to supply, most of the time, the entire load or a 11699
Table 6. System indices for different islanding success probability. SAIDI
SAIFI
ENS
AENS
ASAI
100 % success
2.7276
0.1949
74,785
26.17
0.9996886
75% success
3.3053
0.2188
83,739
29.30
0.9996227
50% success
3.8187
0.2408
89,916
31.46
0.9995641
25% success
4.3288
0.2627
95,960
33.58
0.9995058
0% success
4.8411
0.2847
101,890
35.65
0.9994474
18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011
behaviour of the generation, extracting a higher benefit from microgrids implementation.
Table 7. Coefficient of variation for different islanding success probabilities. a - SAIDI
a - SAIFI
a -ENS
a - AENS
a - ASAI
100 % success
0.0321
0.0029
0.0453
0.0453
0.0000010
75% success
0.1324
0.0872
0.0655
0.0655
0.0000050
50% success
0.1286
0.0896
0.0621
0.0621
0.0000561
25% success
0.0976
0.0708
0.0474
0.0474
0.0000048
The increase in the islanding success probability implies in an improvement of the reliability indices. For example, varying the islanding success probability from 25% to 75% implies in a gain of 24% in SAIDI, 17% in SAIFI and 13% in ENS. These improvements are explained by the fact that the microgrid can supply the island load more times, thus guaranteeing the service continuity for longer. However, to enhance the islanding success, efforts must be concentrated on best control and protection devices of the microgrids, what, on the other hand, implies in higher costs. As occurred for Case 3, the desired coefficient of variation wasn’t reached for the cases with lower islanding success probability. The reason is also the increase in the randomness of the simulation when the uncertainty about the islanding success is included. As an example, Figure 4 shows the evolution of the coefficient of variation for the case of 50% islanding success probability, where the stabilization of the coefficient of variation around these values is clearly shown.
Figure 4. Evolution of the coefficients of variation for 50% islanding success probability. 6. CONCLUSIONS The hybrid method for reliability assessment of microgrid containing renewable energy sources and storage systems was effective for a rapid analysis, incorporating the intermittency of the renewable energies and the islanding process success on the system reliability indices.
Another important conclusion is that ensuring that the islanding process is successful is what can really bring a large impact on the microgrid reliability. Currently, generations such as wind and solar power have a small probability of maintaining an island. However, with the application of smart grid technologies, which will be a reality in the near future, the microgrid formation success may be guaranteed, as well as the success of the reconnection to the main grid. REFERENCES Bagen, B. and Billinton, R. (2005), Incorporating Well-Being Considerations in Generating Systems Using Energy Storage, IEEE Transactions on Energy Conversion, v. 20, n. 1. Billinton, R. and Jonnavithula, S. (1996), A Test System for Teaching Overall Power System Reliability Assessment, IEEE Transactions on Power Systems, v. 11, n. 4. Bollen, M., et al. (2009), Performance indicators for microgrids during grid-connected and island operation, IEEE Power Tech Conference. Chowdhury, S. P., Crossley, P. and Chowdhury, S. (2009), Microgrids and Active Distribution Networks, The IET, London. Costa, P. M. and Matos, M. A. (2005), Reliability of Distribution Networks with Microgrids, IEEE Power Tech Conference. Gellings, C. W. (2009), The Smart Grid: Enabling Energy Efficiency and Demand Response, CRC Press. Jenkins, N., Ekanayake, J. and Strbac, G. (2006), Distributed Generation, The IET, London. Leite, A. P., Borges, C. L. T. and Falcão, D. M.(2006), Probabilistic Wind Farms Generation Model for Reliability Studies applied to Brazilian Sites, IEEE Transactions on Power Systems, v. 21, n. 4, p. 14931501, 2006. Manwell, J. F., et al. (2006), Hybrid2 – A Hybrid System Simulation Model – Theory Manual, National Renewable Energy Laboratory. Park, J., et al. (2009), A Probabilistic Reliability Evaluation of A Power System Including Solar/Photovoltaic Cell Generator, IEEE Transactions on Power Systems, v. 24, n. 2. Verhoeven, B. (2002), Probability of Islanding in utility networks due to grid connected photovoltaic power systems, Report IEA-PVPS T5-07: 2002.
Microgrids implementation at distribution systems brings significant improvements when the concerns are about reliability. However, if they are maintained by generators that depend on random energy, as the case of solar and wind power, the benefits on the reliability indices are reduced. Nevertheless, the installation of storage systems together with intermittent sources can reduce drastically the random
11700