Sie sind auf Seite 1von 6

Integrated reliability evaluation of generation, transmission and distribution systems

A.M. Leite da Silva, A.M. Cassula, R. Billinton and L.A.F. Manso


Abstract: A new methodology evaluates the reliability of distribution systems considering the impact of failures from the generation and transmission (G&T) systems is presented. An integrated adequacy evaluation. including generation, transmission and distribution. is performed to provide detailed information on the interruptions experienced by consumers. G&T systems are represented by a fictitious equivalent network whose parameters are obtained by Monte Carlo nonsequential simulation. The equivalent G&T network is then connected with the distribution network and analysed using minimal cut-set theory. Traditional distribution indices (e.g. SAIFI, SAIDI, etc.) as well as the loss-of-load cost indices are disaggregated to measure the contribution of G&T and distribution systems on the overall system indices. The method is applied to a hybrid system created from two standard test systems. The results and their potential applications in the new power system competitive environment are discussed.

Introduction

Distribution systems [Id] have always received less attention than generation and transmission (G&T) systems with regard to reliability modelling and evaluation [7-121. An analysis of customer failure statistics of the most ut [I]. however, shows that the distribution system makes the greatest individual contribution to the unavailability of supply to a customer. In the last few years, distribution systems have begun to receive more attention mainly due to the restructuring and privatisation process of the power sector many countries in the world are undergoing. Therefore reliability worth will have to be accurately evaluated and inserted in the electricity tariff for compensating customers in case an intemption occurs. The reliability evaluation of a complete electric power including generation. transmission and distribution s (known as hierarchical level 3, or simply HL3) is an important objective in overall power system planning and operation [3]. The usual hierarchical levels are shown in Fig. I . Due to the dimension of the HL 3 problem, however, reliability studies are normally performed assuming that the G&T system supply points have unlimited capacity and 100% reliability. In this work, a new methodology is proposed to evaluate the impact of HL 2 failures on the distribution systems. This is achieved using nonsequential Monte Carlo simulation. Based on the load curtailment strategy at the distribution level, a fictitious network is created whose components simulate the HL 2 interruptions. This procedure generalises
<'

HLO HLI HLZ

I1 I'

HW

L -,-

h-i -diStrtbUtion
d

Fig. 1 Power .sJxlem Aiemnliicul level3

IEE. ZWZ

the concepts proposed in [3].Both fictitious and distribution networks are analysed using the minimal cut-set concepts [13]. In addition to the usual distribution indices for the system and load points, the loss-of/oud cost (LOLC) index or equivalent [2. 4, 9-12] is evaluated and disaggregated considering the hierarchical levels. The proposed methodology can therefore provide economic insight in terms of responsibilities for the customer damage costs.
2
Distribution system reliability

IEE Proccw1in;p online no. 2WZW81 Dol. IO. IWYiip-gld:?W2W8I Paper first received 6th Novemhr ?Ow and in revired form 23id May ?WI A.M. Lriic d Silva and A.M. Couua arc with the lnrtiriite of E1emic;il s
Enginefing. Federal Univenily

- EFEI. A?. BPS 1303. Pinhcinnho. 1wj;llabi.

MG, 37,500-903. Br-l R. Billinion is uith the Electrical Enginenng Depiinmeni. T h e Univenity of Saskatchewan. 5.~rha1oon. Canada L.A.F. Manso is with the Elclncal Enginrrfing Department. Fedenil IJnivmily. SCo JoCo Del Rei FUNREI. MC. Briiul
~

A distribution system is frequently represented as a network in which the system components are connected together either in series, parallel, meshed or a combination of these. There are different analytical techniques available to determine the solution and the evaluation of these networks [13]. If continuity of supply is the major concern, the
I

IEE Proc-Ge,iei: Trumr. Dlirrih.. CbL 14Y. No 1. Jmrur~v ZuO2

method bascd on minimal cut-set theory is the most useful, since the cut-sets are directly related to the system failure modes.

2.1 Evaluation techniques


The Markov technique and the frequency aiid duration approach form sound and precise modelling and evaluation methods for ii range of reliability applications [13]. For hrge and complex systems, such as distribution networks, approximate equations have been developed based on thFse methods. These approximate equations can be used' in conjunction with minimal cut-sct techniques to give- rapid and sufficiently accurate results for a wide range of practical distributions systems.

(consumer surveys). These studies describc different factors which may have some impact on the UC, and the duration of.the interruption is considered the most important one. The accuracy level established to evaluate this duration is an important element in the quality of the estimates of LOLC indices.

Impact of G&T failures on distribution system

2.2 Performance indices


It is essential for electric utilities to track distribution system reliability levels and define performance indices Ib assess their basic function of providing a cost-effective and reliable power supply to all sectors of society. This procedure is known by past performance or historical reliability assessment, and it is widely used by utilities. Future perlblmaucc or predictive reliability assessment is another valuable procedure which can be used to determine system reinforcements and to compare expansion alternatives. To measure the past or future performance of the supply adequacy at the customer load points, indices used are: failure rate i ,failure duration r, unavailability U, energyriol supplied ENS; and for the system indices SAIFI. SAID1 and CAlDI [I]. At hierarchical level 2, predictive assessment is the most common procedure uscd for system reliability planning. On the contrary, at HL3, predictive assessment is not as popular as historical assessment. Bearing in mind the new competitive environment, there is, however, a growing interest in economic optimisation approaches to distribution planning and expansion. In the new environment, all distribution companies will be looking to identify the best points in the network to receive the appropriate investments. Therefore predictive assessment will become an important tool in the distribution system decision making process. Reliability of any electric service, including distribution activities, should be bascd on balancing the costs to a utility and the value of the benefits received by its customers. A value-based reliability planning approach [4] must attempt to locate the minimum-cost solution, where the total cost includes the utility investment cost plus the operating cost plus customer interruption costs. Thereforc the LOLC, at both load point and system levels, will become the primary index for value-based reliability planning. Tlus index depends basically on the unit interruption cost (UC) of each consuming class, usually given in US$ per kWh. The UCs are obtained through specific economic studies

There are many benefits associated with the ability to perform overall reliability evaluation_ i.e. HL 3 reliability assessment. The overall indices provide reliability measures from a customer point of view, and can be used to rank functional zone contributions and therefore to optiinise investmeiits [6]. The assessment of overall indices is carried out in three major steps. First, an algorithm for evaluating the G&T system is used to generate a significant number of samples of power interruptions, at the high-voltage bus or buses, to which the dislrihution system is connected. In this work the HL2 assessmelit is done by a nonsequential Monte Carlo simulation. In the second step, parameters are extracted from the simulation process to assist in the definition of an equivalent network. which depends on the load curtailment strategy at the bus or buses involved. Once the fictitious G&T equivalent network is defined, it is then connected to the distribution network for the third and final step of the H L 3 analysis.

3.7 Parameters to characterise G&T failures To define, from the reliability point of view, the fictitious components which belong to the equivalent G&T network. it is necessary to know, their corresponding Failure rates i and their undvailabilities U. The methodology to determine such components is described as follows, considering a distribution network with N feeden. During the convergence of the H L 2 Monte Carlo simulation, the sequence of events which represent the failure states associated with the bus connected with the distribution network are stored. Each failure state is characterised by the following parameters [8,9]: incremental frequency,f;,,,,, the amount of load curtailed due to G&T and LCGT.The load curtailments are clustered within power intervals which correspond to the total amount of load connected to each feeder of the distribution network. This is illustrated in Table 1. Where NI is the number of the interval and P I , P2. P3 ... 4 ... P,,, the amounts of power associated with the are feeders RI, R2, R3 ... Rk .. . RN. The number of intervals is also the number of feeders connected to the high-voltage bus. This procedure can be easily extended to consider more than one bus connected to the same distribution network. The probability or unavailability U and the frequency associated with each power interval (wluch corresponds to

Table 1: Power intetval for fictitious components G&T


Feeder

NI
p,

Power intewa1

R1 R2
Rk

1 2

P2 Pk PN

P,
P1+Pz+-.. +P(k-,l P,+Pz+. . .+pim-

< LCGT <LCm


< LCGT
1)

5 P,
<P,+P* <P,+P,+...+P, ,,,+P,

RN

<LCGT

< P,+P,+.

. . +P(nN.l,+PRN

load curtailment or failure events) can be obtained using eqns. I . Observe that the Failure frequency is approximately the pseudofiiilure rate associated with the fictitious componcnt. Therefore the G&T fictitious components are entirely characterized by j. and U

where PA represents probability associated with the power interval k. U, is the unavailability of the power interval k , N , is the total number of failure states belonging to the power interval k. NT is the total number of simulations. j., is the failure rate associated with the power interval k.h.is the frequency of occurrence of the power interval k . rA is the average failure duration associated with the power interval k, and /&cl is the summation of the incremental frequency of states j belonging to the power interval k . The pirameter U is usually given in hours per year. i.e.

x;i,

20MW

15MW

40 MW

25 MW

I "
fictitious G&T component
{

t feeder to be disconnected

Fig. 2

G&T,znwor.k,/bur. Iod c u w i l ~ w po1ic:y UJ Tuble 2 ~l

3.2

Representation of load curtailment policies

Each distribution company adopts its own load curtailment policy or strategy. The adopted policy follows a set of criteria to reduce the effects caused by interruptions and also to mininiise the corresponding costs. The load curtailment policies can be represented or modelled by the way that the fictitious G&T components are arranged within the distribution system. To illustrate this point, consider a distribution system with four feeders: RI. R2. R3 and R4. The power capacity of each feeder and the disconnection policy are shown in Table 2.
Table 2 Load curtailment policy
Feeder capacity Power intewal Feeder disconnected

IMWI
R1 20
15

O<LCGi620 20<LCGTs35 35<LCGT675 75<LCGT <lo0

R1
R1. R2

R2 R3
R4

40

R1, R2, R3
R1. R2, R3.

25

R4

There is, however, the need to equate the voltage levels which is done by inserting a 138/11 kV transformer. The reactance of this transformer is 0.12585 pu (50% larger than the RTS transformers). Its failure rate is assumed to be 0.02 failures per year and its MTTR (or parameter r) to be 768 hours. The average and peak loads at bus 6 are 71.8 and 1 16.8MW. respectively. To calculate LOLC indices it is necessary to have the unit interruption costs. In this case the Ontario Hydro data for industrial, commercial and residential customer classes was used [I I . 121. I n the RBTS-bus2, the customer types small user and government/institutions have been considered as industrial and commercial customers. The load shedding policy for bus 6 of the G&T system considers that curtailments will be first applied to the distribution system. i.e. to bus 25. For the distribution system, the curtailment policy described in Table 2 and shown in Fig. 2 is initially used. The only diffei-ence is that the capacity or load associated with the feeders are those corresponding to the IEEE-RBTS-bus2 and described in [5]. The overall electric power system and the equivalent network for the H L 3 assessment are illustrated in Figs. 3 and 4.

The load curtailment (LC) strategy illustrated in Table 2 can be represented by connecting the fictitious G&T components according to the equivalent network shown in Fig. 2. For instance. feeder RI will be disconnected if the load. at the high-voltage bus. has to be curtailed from 0 to 20 MW. If the amount of power to be curtailed lies between 20 and 35MW. both the RI and R2 feeders will be disconnected, and so on. Note that the parameters i and U associated with these fictitious components have already been obtained from previous considerations. Different feeder capacities and load curtailment strategies at the distribution level will imply a different fictitious equivalent network.
4
Integrated reliability evaluation results

4.2

HL3 analysis at peak load level

4. I

Basic test system

The system to be analysed is composed of the distribution system RBTS-bus2 [5] connected to bus 6 of the RTS [14]. Overhead transmission lines are used instead of cables. Bus I of the RBTS is renumbered as bus 25. The average and peak loads at this bus are 12.29 and 20MW. respectively.
IEE Prloc-Gc.sl:
T,rsis,i. Diiiiih ,

This analysis was-carried out considering the peak load situation in which the G&T system is highly stressed and therefore more failures can be expected due to this condition. Table 3 shows the parameters obtained with the nonsequential Monte Carlo simulation for the fictitious G&T components; expressed in terms of i , I' and U associated with the power intervals. These intervals are defined according to the load curtailment policy and the power capacity of the main feeders of the distribution system, as shown in Table 4. Note that power interval I (from 0 to 5.934MW), which corresponds to feeder RI, has the highest failure rate i, while the removal of all four feeders (corresponding to the power interval from 14.491 to 20.0MW) bas the second highest failure rate. Table 5 shows the reliability indices for load points LP-I, LP-9. LP-I2 and LP-21, which belongs to feeders 1.2.3 and 4. respectively. The performance measures are the basic indices i (Failures per year), r (hours), U (hours per year) plus the indices ENS (MWh per year) and LOLC (US5 per year). All the H L 3 indices were obtained for the peak load
3

V 2 14Y. A?>. 1. Jomiiiri. 21x2

equivalent
G&Tsystem

I !

I t

.,

!
(24) (34)

itP

P.16

ILP

P-17
P-16

; ;
I

It
I t

P-10

; ;

P-20

jL

Another conclusion from the results shown in Tables 5 and 6 is that both systems, G&T and distribution, behave as two independent components connected in series. The reason is due to the assumption that all equipment in the distribution system (Le. lines, transformers, etc.) are able to support any loading situation. Due to this restriction it is not possible to measure the simultaneous effect of generation, transmission and distribution. This restriction, however, is due to the continuity criterion which is used in most distribution reliability assessment methodologies.

4.3 Effects of LC distribution policy To measure the effects of the load curtailment policy at the
distribution level, Table 7 provides a new strategy in which feeder R3 is the first one to be disconnected, followed by R4, RI and finally R2. The load curtailment policy at the G&T was kept the same, and therefore there was no need for a new H L 2 simulation. The only change is the way the power intervals are handled to define the new parameters of the fictitious G&T components. The structure of the equivalent G&T system is the same as shown in Fig. 4. Table 8 presents the system indices obtained with the new load curtailment policy. Comparing the results from Tables 6 and 8 it can be seen that the indices related with the distribution system have not changed. This could be expected because the inherent reliability of the distribution network depends only on the failure characteristics of its own components. On the other hand, the contribution of the G&T system indices changes, and so do the overall H L 3 indices. In fact, the system with this new policy becomes more costly from the interruption point of view. In the case where the LC distribution policy is such that it removes one feeder at a time as presented in Tables 4 and 7 (Le. RI +R1+ R2-.RI + R2+ R3-RI + R2+ R3+ R4, or R3+ R3 + R4+ R3+ R4+ R1 + R 3 + R4+ R1+ R2 etc.), the equivalent network is simply characterised by the failure of the fictitious components, as shown in Figs. 2 and 4. It is possible. however, that the LC policy involves the combination of feeders (e.g. R l + R 2 - . R l + W + R 3 - +
IEE Pm.-Gmer. Tranru~. Di.snib.. Yo/. 149. No. 1. Jmonury 2W2

condition and are disaggregated into G&T and distribution contributions. From the results in Table 5 one can clearly identify the G&T curtailment policy. Note that the failure rate for feeder RI (from where LPI is supplied) has a higher value (Le. 20.473 failures per year) than feeder R2 (Le. 10.863 failures per year), which, in turn. has a higher value than feeder R3 and so on, according to the adopted load curtailment strategy. The contribution of the failures originating from the G&T system is much larger than those originating from the distribution system itself. Note that the IEEE-RTS is not very reliable at the peak load condition mainly due to generation. Table 6 presents the results obtained for the system indices SAIFI (interruption per customer year), SAID1 (hours per customer year), CAIDl (hours per customer per interruption), ENS (MWh per year) and LOLC (US$ per year). All the H L 3 indices are disaggregated into G&T and distribution contributions. For instance, the SAIFI for the overall system (i.e. HL3) is equal to 11.4377 interruptions per customer, where 11.1845 interruptions come from the G&T system and only 0.2532 interruptions come from the distribution system.
4

Table 3 Parameters for G&T components, peak load

Table 7: Feeder peak load capacity, new LC policy

Parameters
i. (per year)

Power interval
1 9.609967 35.646540 342.562100 2 2.939667 39.366430 115.724200 3 3.159663 46.256100 146.153700 4 4.763195 28.360880 135.08840

Feeder
R1
R2
I

IMWI
5.934 3.500 5.057 5.509

LC due to G&T
~

Disconnected feeder
~~

O<LCGGrs5.057 5 . 0 5 7 ~s L, C 10.566 10.566<LCGis16.500 16.500~LCG,s20.000

R3 R3, R4 R3, R4. R1 R3, R4, R1. R2

rlh)

U lh Der

vr)

R3 R4

SI

Table 4 Feeder peak load capacity, LC policy

Table 8: System indices !or new LC policy, peak load

Feeder
R1 R2 R3 R4

IMWI

LC due to G&T

Disconnected feeder
R1 R1, R2 R1, R2. R3 R1. R2. R3. R4

System
G&T

SAlFl 13.0544 0.2532 13.3076

SAID1 463.3618 3.6239 466.9857

CAD 35.4947 14.3113 35.0916

ENS 7828792 61607 7890399

LOLC 37539720.00 304873.50 37841880.00

5.934
3.500

o< LCGi<5.934
5.934< L C ~ ~ s 9 . 4 3 4 9.434<LC~~414.491 14.491iLCcr~20.000

Distrib.
HL3

5.057 5.509

Table 5 Load point indices, peak load

separately evaluated and afterwards added to the distribution indicea.


ENS LOLC 128217.80 622.93 128840.80

System i
LP-1 G&T

20.4730 36.1259 739.6046 641089.20 0.2442 14.7114 3.5932 3114.63 20.7172 35.8734 743.1978 644203.90

Distrib.
HL3 LP-9 G&T

10.8630 36.5468 397.0081 743238.80 2378364.00 0.1448 3.5215 0.5098 954.36 4636.49 11.0078 36.1125 397.5178 744193.10 2381418.00 7.9230 35.4992 281.2603 205066.90 41013.37 534.07 41547.45

Distrib.
HW
LP-12 G&T

Distrib.
HL3 LP~21G&T

0.2605 14.0588 3.6625 2670.36 8.1835 34.8167 284.9228 207737.20

4.7630 28.3600 135.0787 123826.60 1077292.00 0.2573 13.9703 3.5943 3294.93 28665.87 5.0203 27.6225 138.6730 127121.60 1105958.00

Distrib.
HL3

Table 6 System indices, peak load


System
G&T SAlFl 11.1845 0.2532 11.4377 SAID1 390.3521 3.6239 393.9759 CAlDl 34.9011 14.3113 34.4453 ENS 7943191 61607 8004796 LOLC 36716580.00 304873.50 37018730.00

4.4 Final remarks The proposed technique can be used in distribution systems connecting more than one bus from the H L 2 system, or embedded generators in the distribution system. The assumption that failures at the HL 2 level are independent of distribution failures is used for the majority of large-scale networks. In cases where severe weather conditions affect specific network components at both levels (i.e. H L 2 and distribution). usually subtransmission components, they must be identified and represented in the system through common-cause failure rates. The proposed technique can then be applied. However; if there are some cascading effects to bs treated, only sequential or pseudochronological M o n k Carlo simulation [I21 should he considered. The example in this work has considered only the peak load condition. In fact the H L 2 simulation is carried out based on a load representation in which the peak condition last for 8760 hours of a year. There is no constraint to consider different load lcvcls and then combine the results appropriately. Another possibility for accounting timevarying loads, discussed in [12], is being considered for implementation in the integrated reliability evaluation of generation, transmission and distribution systems in the near future.
5

Conclusions

Distrib.
HL3

RI +R3- R2+ R 3 + R I + R2+ R3-RI R2+R3+ R4)_ or even specific customers (load points), which are disconnected according to some priority list. In these cases the fictitious components 3re obtained using the same procedure described in subsection 3. I; nevertheless the failure of G&T components must he associated with the opening of the breakers, which disconnect the feeders or load points according to the specified priority sequence. Moreover, as previously demonstrated, the impact of the equivalent G&T network on the H L 3 indices can be
IF Proc:Gm~r. Tronniz. DLwib., Yo/. 14Y. N a I. Jinzuur~ ZWZ

The distribution system is an important part of the total electricity supply, as it provides the final link between a utilitys bulk transmission system and its customers. In many countries the privatising process started with the distribution companies. The new utility framework creates significant changes in comparison with the previous one, in which many electrical utilities were state owned. Distribution companies will be, or already are, under tremendous pressure from the public and from regulatory commissions and agencies to provide high-quality service to their customers. In addition. standards, incentives, penalties, responsibilities. etc. will become an important discussion topic in the electric power sector throughout the world. Integrated reliability evaluation of generation. transmission and distribution system will play an important role in
5

the new competitive scenario as it can provide an overall view Of the system in terms Of both past and future performance. The present work makes a contribution to this HL 3 reliability area by providing indices, including to overall system perfomance, The methodology is based on a combination of nonsequential

distibution system data and mults', IEEE Trum Power Sysr., 1991. 6. (Z), pp. 813-8?0 test system for teaching overall power system reliability assessment', IEEE Trirn~ P o w r Sy.rr. 1996, I I . (4). pp. 1670-1676 7 LElTE DA SILVA. A.M.. MELO, A.C.G., and CUNHA. S.H.F.: 'Frequency and duration method for rcliability e v ~ l ~ a t i o n largeof scale h y d r o t h e m i generating s~stcmi'.IEE Pruc. C,Gener. Trun.

6 BILLINTON. R.. and JONNAVITHULA, S.: 'A

hierarchical levels. The proposed me tho do lo^ can provide technical and economic bases for discussing the standards and rcsponsibilities associated with customer damage.fl"'" ' ( i r r 6 References JP
BILLINTON. R.. and ALLAN. R.N.: 'Reliabilily evaluation of power systems', 2nd Ed". (Plcnum Press. NY , 1994) 2 ALLAN. R.N.. and DA SILVA, M.G.: 'Evaluarian of reliabiiity indices and outage costs in distibution systems'. IEEE Trilnr Pmwr sy.~t.. in, (1). pp. 4 1 ~ 9 1995. 3 BILLINTON. R., and SATISH, J.: ' I t tof rotational load shedding ETi on overall ~ o w system adesuacy indices', IEE Proc Gener. T r m m r ~ i ~ i h1996; i43;(2). pp. 181-is7 .. 4 CHOWDHURY. A.A.. and KOVAL. D.O.: 'Value-based distrihution system reliability planning', IEEE Tram. lml. Appl.. 1998, EA-.%, I

'.

Power Syrt., 1993, R. (3).'pp. lllE-Il2< GOEL, L,, and RILLINTON,R.: .UtiliDtion of energy assessment rates to evaluate reliability worth in eltitrial power systems', l E E t Truns. P w e r Sy.~i, 1993, 8. (i), 929-936 pp. ' 10 WENYUAN, LI. and RILLINTON. R.: 'A minimum cost assessmen1 method for composite generation and transmission system expansion planning', IEEE Trms. Poiver Sy.yt., 1991. 8, (2). pp. 628L l i

"2,

I1

5 ALLAN, R.N.. BILLINTON, R., SJARIEF.1.. GOEL, L.. and SO,


K.S.: 'A reliability test system for educational purposes-basic

,,,,pt,.
/I/

LJ-L/

7 1 30

MELLO, J.C.O.. LElTE DA SILVA. A.M., and PEREIRA, M.V.F.: 'Efficient loss-of-load cost eviilwation by combined pseudosequential and state transition simulation'. IEE Pwc. Gene?. Trmwi, Dimib.. 1997. 144. ( 2 ) pp. 147-154 I2 LElTE DA SILVA, A.M., MANSO, L.A.F., MELLO. J.C.O.. and BILLINTON, R.: 'Pscudwhionologjcal simulation for composite reliabitity analysis with time-raMng loads'. IEEE Tranr Poaw Swt. 2ooO. 15, (I). pp. 73-80 13 BILLINTON, R.. and ALLAN, R.N.: 'Rcliability evaluation of cngmering systems - concepts and techniques', 2nd E d n . (Plenum DIPi" h " l IOU,/ > .,., 14 'IEEE reliability test system'. I E t t Tuns. Punpr Appw Syrr.. 1979. 98, (6).pp. 2047-2054
~~

"",

Das könnte Ihnen auch gefallen