DRPT2008 6-9 April 2008 Nanjing China

Risk Theory based On-line Power System

Security Assessment
Tianshu Bi, Member, IEEE, Xiaohui Qin, Renmu He and Qixun Yang

Abstract—Power system is approaching its limits due to the profit of the investment is a problem facing us.
deregulation of power industry and unavailability of new Some pioneers present the concept of risk to do the on-line
transmission corridor. However power system security security assessment [1-10]. The security risk is varied with the
assessment is carried out and the operating limits are set topology of the system, corresponding load level [1-2].
according to the given contingency list which might happen in However much attention is paid on the historical data based
very rare frequency. In other words, the frequency of fault security risk evaluation instead of setting the operating point
occurrence and the risk of blackout is varied when the operating according to the risk. Our idea is that the security risk of power
condition (load level, status of power equipment, and weather)
systems is varied with the operating condition, such as load
changes. How can we do the security assessment based on the
level, aging status of equipments, environment weather, etc.
operating condition to get the best equilibrium between the
security and the profit is the focus of this paper. The idea of risk
When the equipments are healthy and environment weather is
based on-line power system security assessment is suggested. A sunny, the risk might be low. The load level can be
precise calculation method of average failure rates of components deliberately set higher to get more profit. On the contrast,
is presented. Besides, operating condition based failure rates of when the transmission lines are ice-coated, the load level can
components are also analyzed systematically in the paper. The be set lower to make the system secure.
Monte-Carlo simulation demonstrates the viewpoint of the paper. In this paper, the idea of risk based on-line power system
security assessment is suggested. Moreover, a precise
Index Terms -- security assessment, risk, failure rate, false trip
calculation method of average failure rates of components is
of protection, power systems.
put forward, which is the key point of risk. Besides, operating
condition based failure rates of components are also analyzed
I. INTRODUCTION systematically in the paper. Finally, the Monte-Carlo
The operating environment of power system is becoming simulation is carried out to demonstrate the viewpoint of the
increasingly competitive. The introduction of new paper.
technologies (HVDC, FACT, etc) makes the dynamic
behavior of the system become much more complicated. The II. OVERALL SCHEME OF RISK BASED ON-LINE
deregulation of power industry makes the system approaching POWER SYSTEM SECURITY ASSESSMENT
its limit. Moreover, power grid interconnection makes the The mathematic model of risk based security assessment is
large-scale blackout occurs more frequently, which result in given below.
max(F )
incredible economic loss. All these factors present the need for
security assessment, which can clarify the on-line security
status of power system, give early warnings and thus improve u (1)
the security and reliability of system greatly. F = Bic − Pls ⋅ C ls
Currently power system security assessment is carried out Where, F denotes the objective function which should be
based on the given contingency list, where the most severe maximized, u represents the operating point, which mainly
fault scenarios are considered to get the operating mode and refer to system topology and load level. Bic is the profit
corresponding limit. Actually the most severe fault occurs in function corresponding to the given operating point. Pls
very rare frequency. That means the power system operating indicates the probability of losing stability, which is the
limit will be low to guarantee the security. In the environment function of system operating point u and component failure
of power market, each participant wants to maximize his profit, rate λ0. While the component failure rate λ0 depends on the
which conflicts with the principle of the existing security operating condition, such as aging status, environmental
assessment. Moreover the transmission corridors become weather, operation of relay and recloser, etc. Cls is the loss
harder to plan due to environmental concern. This is also the function of losing stability. The product Pls⋅Cls is called the
incentive to dig out the transmission capability of the existing security risk index of power system.
lines. How to solve the conflict between the security and the In this mathematical model, based on the operating
condition, the component failure rate λ0 can be calculated.
This work was supported by Program for New Century Excellent Talents Then the operating point u can be set to get the maximum
in University (NCET-05-0216), NSFC (50607005), the Key Project of profit.
Chinese Ministry of Education (NO.106052), “111 project” (B08013) and It should be pointed out that the probability of losing
Program for Changjiang Scholars and Innovative Research Team in
University (IRT0515). stability Pls can be obtained by Monte Carlo simulation once
T. S. Bi, X. H. Qin, R. M. He. and Q. X. Yang are all with Key λ0 and u known. In addition, the profit and loss functions (Bic
Laboratory of Power System Protection and Dynamic Security Monitoring and Cls) are both economical indices which will be discussed in
and Control under Ministry of Education, North China Electric Power another paper. Therefore, how to calculate component failure
University, Beijing 102206, P. R. China (e-mails: [email protected] and
[email protected]).
rate λ0 according to the operating condition is the key point the

978-7-900714-13-8/08/ ©2008DRPT 704

 DRPT2008 6-9 April 2008 Nanjing China

proposed model. µ p µu µs
Pup = (5)
µ p µ u µ s + λ p µ u µ s + λu µ p µ s + λ s µ p µ u
III. OPERATING CONDITION BASED COMPONMENT
FAILURE RATE λ p µu µ s
Ppo = (6)
A. The Traditional Fault Occurrence Model µ p µ u µ s + λ p µ u µ s + λu µ p µ s + λ s µ p µ u
Traditionally, the uncertainty of fault occurrence is λu µ p µ s
Pur = (7)
described using the following model [1]. µ p µ u µ s + λ p µ u µ s + λu µ p µ s + λ s µ p µ u
The probability of fault occurrence on a line can be modeled
using the Poisson distribution with a constant fault rate. From λs µ p µ u
Psr = (8)
the Poisson distribution formula, the probability of no fault in µ p µ u µ s + λ p µ u µ s + λu µ p µ s + λ s µ p µ u
the time period t is given by: where Pup, Ppo, Pur Psr are the state Probabilities of the up state,
e − λ0t (λ0 t ) 0 planned outage state, the unsuccessful reclosure state, and the
Pno = = e − λ 0t (2)
0! successful reclosure state. λp, λu, λs are the transition rates of the
where Pno is the probability of no fault occurrence, λo the planned outage state, the unsuccessful reclosure state and the
average fault rate, and t the duration considered. successful reclosure state. µp, µu, µs are the recovery rates from
So, the probability of a fault occurring on a line in t is the planned outage state, the unsuccessful reclosure state and
PO = 1 − e − λ0t (3) the successful reclosure state.
It is noticed that the recovery time from the successful
It is noticed that (3) is same as the fault probability following reclosure fault state to up state is extremely small, about 1
the exponential distribution. In fact, the Poisson and second. Therefore, µs is extremely large, so (5-8) can be
exponential distribution are essentially the same since both are simplified as follows.
based on the constant rate assumption.
µ p µu
If the condition of λot <<1 holds, (3) can be written as: Pup = (9)
PO = λ0 t (4) µ p µ u + λ p µ u + λu µ p
Usually, the average failure rate is approximately replaced λ p µu
by the frequency of fault occurrence, which could be obtained Ppo = (10)
µ p µ u + λ p µ u + λu µ p
from historical records, since the duration of a fault is always
extremely short. λu µ p
Pur = (11)
The fault occurrence probability is replaced by the forced µ p µ u + λ p µ u + λu µ p
outage rate (FOR) in some references [10, 12], that is also an
approximation. Psr ≈ 0 (12)
However, we cannot know whether the fault occurs from Equation (12) and extreme large µs also implies that the
the practical power system. The only information we can get mean duration residing the successful reclosure fault state is too
is whether the circuit breaker open or not, i.e. the recloser short.
works successfully or not. If the recloser closes the circuit f up = Pup (λ p + λu + λs )
breaker successfully, there is no outage. On the constrast, µ p µ u (λ p + λ u + λ s ) (13)
failure of recloser means the corresponding circuit breaker =
keeps open and there is outage. Aiming at this problem, a µ p µ u + λ p µ u + λu µ p
method to precisely calculate the average failure rate is f po = Pup λ p
presented.
λ p µ p µu (14)
B. The Calculation of Precise Average Failure Rate and =
µ p µ u + λ p µ u + λu µ p
Relative Monte Carlo Simulation
Based on the analysis in above subsection, the space f ur = Pup λu
diagram can drawn in Fig.1. λu µ p µ u (15)
=
µ p µ u + λ p µ u + λu µ p
f sr = Pup λs
λs µ p µu (16)
=
µ p µ u + λ p µ u + λu µ p
where fup, fpo, fur fsr are the frequencies of the corresponding
states(outages/year). In most data collection systems, the
outage rates (λp, λu, λs) are not directly colleted. The outage
frequencies (fpo, fur fsr) and the repair times (the reciprocals of
Fig.1 The state space diagram for line fault occurrence
mean values the repair times are µp, µu, respectively) of
Applying the Markov method to the state space diagram, we components are recorded. Therefore, we have to use (14-16) to
can obtain the following results. calculate λp, λu, λs. Then it is easy to get the total average fault
rate λo=λu+λs. The probability of the faults which result in the

705
 DRPT2008 6-9 April 2008 Nanjing China

unsuccessful reclosure in the total faults is P1=λu/λu+λs, and the Then, the total average failure rate for adverse weather
probability of the faults which result in the successful reclosure condition is λad,o=λad,u+λad,s, and for normal weather condition is
in the total faults is P2=λs/λu+λs. λno,o=λno,u+λno,s.
After the total average fault rate λo is obtained, the Monte Therefore, the corresponding failure rates for different kinds
Carlo simulation method can be applied in sampling the fault of weather condition can be used for on-line power system
occurrence. For a line, a uniformly distributed random number security assessment. The Monte Carlo simulation method also
R between [0, 1] is generated. If R＜PO, the fault occurs; can be applied for the different failure rates.
otherwise the fault does not occur. This is shown in Fig 2. Obviously, the method can be extended to the multi-type
weather conditions.
In the case of overhead line, it is possible that a line traverses
two regions where Region 1 is in the adverse weather and
Fig.2 Sampling fault occurrence Region 2 in the normal weather. The length exposed to the
If the fault occurs on a line, the Monte Carlo simulation adverse weather condition is x as shown in Fig 3, therefore the
method can also be used to sample the fault with a successful or total failure rate of the line λle is: λle = λ ad x + λ no (1 − x )
unsuccessful reclosure in a similar way. The fault location, the
fault types (single phase to ground, double phase to ground,
three-phase, phase-to-phase) and the fault clearing time can
also be sampled by Monte Carlo simulation method.
Fig.3 Schematic diagram of line traversing two regions
C. The Weather-dependant Failure Rate It is pointed out that the failure rate in region 1 is λad,
The failure rate of an overhead line varies with the weather therefore the fault location probability should be modified as
and other conditions. When the components are exposed to an follows.
adverse environment, their failure probabilities increase Assumed that the probability of fault location yields the
dramatically. If the historic data (such as the failure frequencies uniform distribution between [0, 1] in normal weather
and the repair times) are collected according to the classified condition. Then the probability of a fault occurring in Region 1
weather respectively, the various average failure rate of is modified as:
component in each type of weather can be calculated separately. xλ ad
But now, most data collection systems do not distinguish the Pl ,1 = (22)
xλ ad + (1 − x )λ no
weather condition but only respond to an average failure
The probability of a fault occurring in Region 2 is modified
frequency and an average repair time in past years.
as:
In terms of the calculation of precise failure rate in above
subsection, the weather-dependent failure rate described in (1 − x )λ no
Pl , 2 = (23)
reference [1] should be modified as below. xλ ad + (1 − x)λ no
If the weather is divided into two basic categories: normal Obviously, the method can be extended to the cases in which
and adverse. Assume that the frequencies of the planned outage an overhead line traverses more than two regions where
are not affected by the weather. We have the failure rates to the different weather conditions are assumed.
planned outage for both normal and adverse weather The impact of aging status of equipments on failure rate is
conditions: analyzed in [1] in detail.
λ ad , p = λ no, p = λ p (17)
D. The Impact of EMC Level and the Direct Current
where, the subscript ad and no denotes the adverse weather Magnetic Bias due to HVDC Single Pole Operation
and normal weather respectively. λp can be calculated by
(14-16). The EMI (electromagnetic interference) will result in the
Then, the failure rate of unsuccessful reclosure λad,u and the false or fail operation of relay or other IED (intelligent
failure rates of successful reclosure λad,s under adverse weather electronic devices). So, if the EMC (electromagnetic
condition can be obtained by solving (18, 19). compatibility) level of a substation is not sufficient, the
λ ad ,u µ ad , p µ ad ,u probability of the mal-operation or the refusal-operation of
f ad ,ur = (18) protection or other IED in this substation will increase. The
µ ad , p µ ad ,u + λ p µ ad ,u + λ ad ,u µ ad , p incremental degree can be described as a factor which is the
λad , s µ ad , p µ ad ,u multiple of the normal EMC level with respect to the current
f ad , sr = (19) EMC level.
µ ad , p µ ad ,u + λ p µ ad ,u + λad ,u µ ad , p The HVDC single pole operation will last half an hour at
The failure rate of unsuccessful reclosure λno,u and failure least. It results in the DC magnetic bias which adds an
rate of successful reclosure λno,s under normal weather additional DC voltage on the neutral point of the relative
condition can be obtained by solving (20, 21). transformer. Therefore, the over excitation of the transformer
λno,u µ no, p µ no,u will occur, and the shake and noise of the transformer is more
f no ,ur = (20) obvious. So the probability of the relative transformer failure
µ no , p µ no ,u + λ p µ no ,u + λno ,u µ no , p will increase in the period of HVDC single pole operation,
λno, s µ no , p µ no ,u which has to be taken into account in the on-line risk
f no , sr = (21) assessment.
µ no , p µ no ,u + λ p µ no ,u + λno,u µ no , p

706
 DRPT2008 6-9 April 2008 Nanjing China

E. The Impact of False-Trip of Protection on Failure Rate under this situation. It should be considered in security risk
The failure operation of protection and its impact on risk assessment.
assessment has been studied in reference [10-12]. The paper
only focused on the false-trip of protection and its impact.
The relays might falsely trip the circuit breakers. Due to the
randomicity of relays’ false-trips, it is hard to take this factor Fig.6 The schematic of false-trip of protection caused by “power direction
into consideration. But some kinds of false-trips have rules to inverse” in AC/DC hybrid systems
be followed and then such kind of false-trips can be When fault occurs on a line, the false trip of protection of
considered. We take the false trip caused by “power direction other line generally decreases the connection impedance
inverse” as an example, which is shown in Fig 4. between the subsystems and generators further, which
deteriorates the transient or dynamic performance of power
systems. The false trip of busbar protection caused by line fault
will yield more serious consequences.
Under the condition that no fault occurs on power systems,
the false trip of protection will cause a branch open which is a
Fig.4 The schematic of false trip of protection caused by small disturbance and will be corrected by the reclosure. But
“power direction inverse” for the cases with a fault, the false-trip of another line will
In double circuit lines or loop lines which are protected by worsen the system state and might cause blackout, which has
the pilot protection, when fault occurs on the side M of line I, been proven in recent occurred blackouts. Therefore, the
for line II, the fault power direction of side M is reverse, which false-trip of protection can not be ignored in security risk
sends the blocking signal to both sides to block trip. But after assessment of power systems. Particularly, the mal-operation
fault clearing in line I, the load power direction of side M in line should be emphasized when it is incorporated with generator
II is inversed to forward instantly, which stop sending the rejection scheme (GRS) and other remedial actions.
blocking signal to both sides. At the same time, the load power The probability of mal-operation of protection without faults
direction of side N in line II is tuned to reverse instantly, which on power system can be obtained from the historical data of
begin to send the blocking signal to both sides. But for the pilot protection management system. The probability of
protection of side M of line II, the blocking signal sent by side mal-operation of protection with fault in nearby region also can
M disappeared first, and the blocking signal sent by side N be obtained from the historical data of protection management
arrives later due to the channel delay shown in Fig 5. system. When fault occurs on double circuit lines, especially
nearby the rectifier or inverter, the probability should be larger
than that under other situations. Monte Carlo simulation
method also can be applied to sample the occurrence of
mal-operation of protection once the probabilities of
Fig.5 The blocking signal sequences in false-trip of protection mal-operation are obtained. The mal-operation of protection is
caused by “power direction inverse” treated as an event in the disturbance sequences in time domain
Therefore, in the period ∆t, the protection of side M of line II simulation of security risk assessment.
are not blocked by any signal and then mal-operates to trip the
breaker. IV. TEST RESULTS AND ANALYSIS
Similarly, if the fault clearing time of side M in line I is The relative small WSCC 9-bus test system shown in Fig. 7
shorter than that of side N, the “power direction inverse” is studied to illustrate the idea in this paper.
phenomenon will also appear and cause the false trip of
protection. In this case, because the inverse power is the fault
power, the phenomenon is more obvious.
The false trip of protection caused by the “power direction
inverse” in AC systems can be avoided using the time delay
coordination. But in AC/DC hybrid systems shown in Fig 6, the
false trip of protection caused by the “power direction inverse”
due to the failure of phase change in rectifier or inverter and the
subsequent DC block is hard to be avoided.
Fig.7 WSCC 9-bus system
If the fault occurs on the line nearby the AC bus of the Assume that the 40% length of line Bus2-BusC and 50%
rectifier or inverter substation, the possibility of phase change length of line Bus2-BusA are exposed to the adverse weather
failure in rectifier or converter will increase dramatically. The condition. The EMC level in substation C is not sufficient and
consequence of the phase change failure is DC block. At the the modification factor is assumed as 2.0, which denotes the
instant that DC is blocked, the fault power direction of side M probabilities of mal-operation and refuse-operation of
in line II is inversed to forward, and the relative protection will protections in substation C is multiplied by 2.0. The
falsely trip line II as described above. The time from the fault probability of mal-operation of protection without faults on
occurring to DC block is too short to be distinguished. The power system is assumed as 0.001 for other lines and buses.
waveforms distortion in one time window caused by phase The probability of mal-operation of protection with fault in
change failure and DC block will also confuse the direction nearby region (here, is the neighboring line) is assumed as
component. Therefore it is hard to prevent false-trip of relay

707
 DRPT2008 6-9 April 2008 Nanjing China

0.002. The probabilities of refusal-operation of all kinds of forward. Then, the operating condition based failure rates are
protections are all assumed as 0.01. The three-phase trips do not analyzed systematically. The Monte-Carlo simulation
result in the reclosure. For all lines, fpo, fur fsr are assumed as 1, demonstrates the viewpoint of the paper.
0.5, 1 respectively, and µp, µu are assumed as 365, 400
respectively. The probability of fault types is the same as that in VI. REFRENCES
reference [1] and the probability of fault location is assumed to [1] Li Wenyuan. Risk Assessment of Power Systems Models-Methods, and
yield the uniform distribution. For all buses, the failure rates Applications[M]. IEEE Press and Wiley & SONS, USA and Canada, 2005.
are assumed as 0.002. For all transformers and generators the [2] Li Wenyuan, Lu Jiping. Monte Carlo Method for Probabilistic transient
stability assessment[J]. Proceedings of the CSEE,2005,25(10):18-23.
failure rates are assumed as 0.004. The fault clearing time is [3] CIGRE Task Force 38.0312. Power System Security Assessment, a position
assumed to yield the normal distribution with 0.12s mean and paper[J]. Electra, 1997,(175):49-77.
the standard deviation 2%. The time of backup protection is [4] McCally J D, Fouad A A, Vittal V et al. A Risk-Based security index for
determining operating limits in stability-limited electric power systems[J].
assumed to yield the normal distribution with 0.62s mean and IEEE Transactions on Power Systems, 1997,12(3):1210-1219.
the standard deviation 2%. The time of reclosure is assumed to [5] McCally J D, Vittal V, Abi-Samra N. An Overview of Risk Based Security
yield the normal distribution with 1.0s mean and the standard Assessment. IEEE power engineering society summer meeting[C].1999,
1:173-178.
deviation 2%. The insulation aging failures are only considered [6] Ming Ni, James D McCally, Vijay Vittal et al. On-line Risk-based Security
for transformer Gen3–Bus3, and the survived time past normal Assessment [J]. IEEE Transactions on Power Systems, 2003,18(1):258-265.
life is 1 year. [7] Ming Ni, James D McCally, Vijay Vittal et al. Software Implementation of
Online Risk-based Security Assessment[J]. IEEE Transactions on Power
10000 sample cases are generated using nonsequential Systems, 2003,18(3):1165-1172.
Monte Carlo method. The probability of instability Pis is [8] Ding Ming, Huang Kai, Li Shenghu. Probability Stability Assessment for
defined as the proportion of the instability cases in the total Hybrid AC/DC Power System[J]. Proceedings of the CSEE, 2002,
22(8):11-16.
sample cases. It can be used as the risk index. The time period [9] Ding Ming, Li Shenghu, Wu Hongbin. Integrated Evaluation of Power
considered in on-line assessment is 15 minutes and the whole System Adequacy and Stability[J]. Proceedings of the CSEE, 2003,
simulation spends about 13 minutes and 30 seconds. 23(3):20-25.
[10] Feng Yongqing, Wu Wenchuan, Sun Hongbin. A Preliminary Investigation
on Power Systems Operation Risk Evaluation in the Modern Energy
Table 1 The comparison of risk indices in different situation Control Center[J]. Proceedings of the CSEE, 2005, 25(13):73-78.
situation Pis [11] Hong Mei. Element modeling of bulk power system reliability
Basic power flow, the method in this paper is not adopted 0.0011 evaluation[D]. Thesis .Hefei: Hefei University of Technology, 1997.
Basic power flow, the method in this paper is adopted 0.0013 [12] Li Shenghu, Ding Ming, Wang Min, et al. Search of Power System Chained
The injection power of Generator 2 increases 40MW, 0.0021 Failure Mode Considering Uncertainty of Element Fault and Performance
while that of other generator decreases 20MW, and the of Protective Relaying[J]. Power System Technology, 2004,28(13):27-31.
method in the paper is adopted
From Table 1, we can draw some conclusions as follows:
compared with the traditional method, the risk index obtained VII. BIOGRAPHIES
by the new method in the paper is increased because the failure Tianshu Bi received Ph.D degree from The University of Hong Kong in 2002. She
rate is more precise and more information is taken into account. is currently a professor in North China Electric Power University and vice director
of the Key lab of MOE. Her research interests include power system protection and
The incremental power of generate 2 also increase the risk control, the application of PMU and the fault diagnose.
index, which will attract the attention of dispatcher.
Xiaohui Qin was born in china in 1979. He is currently a graduate student for Ph.D
degree in North China Electric Power University. His research interests focus on
V. CONCLUSIONS
the application of PMU and power system risk assessment.
In this paper, the idea of risk based on-line power system
Qixun Yang was born in china in 1937. He is currently a Chinese engineer
security assessment is suggested to achieve the best
academician and a professor in North China Electric Power University. His
equilibrium between the security and the profit. The research interests include power system protection and control, and substation
calculation method of precise average failure rates is put automation system.

708