This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON POWER SYSTEMS 1

Determination of Transient Stability Constrained Interface Real Power Flow Limit Using Trajectory Sensitivity Approach
Guanji Hou, Student Member, IEEE, and Vijay Vittal, Fellow, IEEE
AbstractFast determination of the transient stability constrained interface power transfer limit is a critical problem in power system real time operation. Conventional methods normally require repeated time domain simulations to determine this interface transfer limit; therefore, they are computationally burdensome and time consuming. This paper presents an application of the trajectory sensitivity method to determine the transient stability constrained interface transfer limit. For a specied contingency, the trajectory sensitivity method is used to determine the stability constrained maximum output of the key generators that affect the power ow of the interface. Then using the power transfer distribution factors (PTDFs) of those key generators on the tie-lines constituting the interface, the maximum interface transfer increment can be determined. The advantage of this method is that it requires much less computation compared to the conventional methods while having the same level of accuracy. An application of the proposed method to the WECC test system is presented in detail. The results obtained demonstrate the efciency and accuracy of the proposed method. Index TermsInterface transfer limit, power transfer distribution factor (PTDF), total transfer capability (TTC), trajectory sensitivity.

I. INTRODUCTION

PERATION of a power system requires that all security constraints established by the North American Electric Reliability Corporation (NERC) be met for all possible operating conditions. Generally these constraints include static constraints such as thermal limits, and dynamic constraints such as voltage limits, transient stability limits [1] and small signal stability limits [2]. A critical component of successful electric power market operation is the determination of the associated total transfer capability (TTC) representing the transmission capacity available for trading. NERC rules require that the TTC values be posted and updated on a public domain website at regular intervals,
Manuscript received October 19, 2011; revised January 18, 2012, May 23, 2012, and September 01, 2012; accepted September 29, 2012. This work was supported by the National Science Foundation under the Grant NSF EEC-9908690 at the Power System Engineering Research Center. Paper no. TPWRS-00985-2011. The authors are with the Department of Electrical, Computer, and Energy Engineering, Arizona State University, AZ, USA (e-mail: guanji.hou@asu.edu; vijay.vittal@asu.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRS.2012.2222942

e.g., the ISO-NE updates and publishes the TTC for every interface in its system every hour during the day on their website [3]. Hence, the underlying method for this TTC calculation should be fast with acceptable accuracy [1][5]. TTC is generally determined by the minimum value of certain limits, including thermal limits, voltage limits, small signal and transient stability limits [1]. Among these limits, the determination and calculation of the transient stability limit is the most complicated and time consuming task. This paper aims at developing a fast and accurate method to determine the transient stability constrained interface real power ow limit. Conventional methods to calculate this transient stability constrained interface transfer limit can be summarized into two broad categories: deterministic methods and probabilistic methods. As for the deterministic methods, one widely used method is the transient stability constrained optimal power ow (TSCOPF) [6][11]. However, the calculation burden is very heavy for this method especially when dealing with a large system. Another widely used deterministic method is the energy function based methods. However, the drawback of these methods is the modeling restrictions [12][14]. The other deterministic method consists of running repeated time domain simulations while gradually increasing the interface power ow. As for the probabilistic methods, the most widely used ones are the Monte Carlo based probabilistic approaches [15], [16]. The drawback of these methods is that they are very time consuming and normally performed off-line. From the above literature survey it is known that these methods are unable to meet either the accuracy or the computational time requirements, especially when the system is large. Therefore, there is a need to develop a method that is computationally fast and accurate. It is observed that an interface ow limit is primarily determined by the generation at certain key generators in the exporting area. In other words, these key generators are the main sources that control the real power ow on the tie lines that constitute the specic interface. If the stability constrained maximum generation increment of these key generators can be determined, then with the use of power transfer distribution factors (PTDFs) of these key generators on the interface, the transfer limit of the interface can be accordingly determined [17]. The idea of using the sensitivity analysis in conjunction with PTDFs to calculate the interface ow limit was rst developed in [17] 20 years ago. In [17], the authors utilized the sensitivities of energy margin with respect to active generation changes to calculate the generation output limits of the key generators

0885-8950/$31.00 2012 IEEE

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE TRANSACTIONS ON POWER SYSTEMS

constituting the interface. Then generator distribution factors were used to calculate the interface ow increment/decrement to determine the interface ow limit. The method was developed based on the transient energy function (TEF) method. However, the limitations of the TEF method, including the difculties in the construction of a suitable Lyapunov function [18] and the modeling restrictions for various dynamic components [12][14] have limited the extensive applicability of the method. To account for the shortcomings of the TEF method, this paper applies the approach to time domain simulation using the trajectory sensitivity method to calculate the required sensitivity information. Research efforts have shown that the trajectory sensitivity approach can effectively complement time domain simulation and provide valuable insights in evaluating limits and account for changes in operating conditions and system parameters [19][26]. Reference [21] introduces a method using the trajectory sensitivity to reschedule power generation to ensure system stability for a set of pre-selected contingencies while satisfying economic constraints. Reference [22] adopts trajectory sensitivity to determine the best locations and also the amount of VAr compensation to avoid voltage collapse in a stressed large power system. Reference [23] introduces a trajectory sensitivity based model predictive control (MPC) method for real-time voltage stability emergency control. The most critical issue in this approach is to determine the stability constrained maximum generation increments or decrements of those key generators in an efcient manner. The trajectory sensitivity approach is proposed here for this purpose. This method can avoid repeated time domain simulations. The organization of the paper is as follows. The basis for the trajectory sensitivity procedure is introduced in Section II, the proposed method is presented in detail in Section III. Application of the proposed method to the WECC test system and the results are shown in Section IV and the conclusions are given in Section V. II. TRAJECTORY SENSITIVITY A. Analytical Basis The analytical basis for dynamic trajectory sensitivity is detailed in [27]. In a conventional time domain transient stability analysis package, a power system is represented by the following set of differential-algebraic equations (DAEs): (1) (2) where vector of state variables; vector of network variables; vector of parameters that are subject to change. includes machine (generator and induction motor) dynamic states such as rotor angle, velocity and appropriate ux linkages, as well as states associated with controllers of a machine, e.g., excitation control and governor. would include network

variables such as bus voltage magnitudes and angles [19]. is a variable characterizing parameter changes, e.g., load, generation, transmission line impedance or generator bus initial voltage set-point. These parameter changes will affect the operational security boundary imposed by dynamic constraints. For all simulation-based approaches, determination of the security boundary requires repeated time domain simulations for the changed operating condition, which is time consuming. The trajectory sensitivity approach is good alternative to this problem. Details will be introduced in the following paragraphs. Parameter sensitivities can be calculated numerically or analytically. The numerical method consists of making a nite change to each selected parameter and running a simulation for the changed condition. Details about this method are given in [22]. The advantage of the numerical method is that it requires no additional programming efforts. The disadvantage is that it requires an additional full time domain simulation besides the base case simulation to obtain the sensitivity with respect to each parameter. Compared to the numerical method, the analytical approach is preferable in terms of computational efciency. Therefore, this paper adopts the analytical method to calculate various types of parameter sensitivities. To account for these changes on the operational security boundary, (1) which is a set of nonlinear differential equations and (2) which is a set of nonlinear algebraic equations are augmented with the following set of linear equations (3) and (4): (3) (4) where vector of the partial derivatives of with respect to (w.r.t) ; vector of the partial derivatives of w.r.t

The set of equations (3) and (4) are called sensitivity equations. For every parameter which changes, the system equations (1) and (2) will have to be augmented with a set of sensitivity equations corresponding to (3) and (4), respectively. It is noted that when the parameter change is made to the pre-fault condition, it will also change the pre-fault stable equilibrium. The steps that are needed to account for this change and suitably set up (3) and (4) under this condition are introduced in [28]. Only sensitivities for smooth dynamics are considered in this paper. Methods to calculate sensitivities when dynamic behavior is inuenced by discrete events are detailed in [19]. When implicit integration methods such as the trapezoidal method are used to solve the system of DAEs (1) and (2), the Jacobian matrices that are needed to solve the trajectory sensitivity equations (3) and (4) are by-products of the process of solving (1) and (2) [19], [20], thus minimizing the computational effort required for solving these additional equations. Moreover, the independent nature of the sensitivity equations for different parameter changes facilitates a parallel computing approach to further enhance the computational efciency. A parallel computing cluster has been adapted for this purpose. This part of work was introduced by the authors in another paper [29].

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
HOU AND VITTAL: DETERMINATION OF TRANSIENT STABILITY CONSTRAINED INTERFACE REAL POWER FLOW LIMIT 3

B. Trajectory Approximation It is seen from (3)(4) that various parameter sensitivities are determined analytically along with the typical system differential algebraic equations (1)(2). However, the manipulation of the parameters sensitivities to approximate the system response for the changed operating condition is conducted numerically based on the rst order linearization. When a system parameter change occurs, the new values of the system variables can be linearly approximated based on the base case value by applying the following equation [20]: (5) where approximated variable trajectory; base case variable trajectory; evaluated sensitivity trajectory; perturbation size of the changed parameter. Equation (5) characterizes the approach to apply the trajectory sensitivity method to evaluate the effect of changes in system parameters as time evolves. III. PROCEDURES FOR THE PROPOSED METHOD Present methods that are widely used by various independent system operators (ISOs) for the determination of the transient stability constrained interface ow limit are based on the combination of the direct method and time domain simulation [30]. First, the direct method is used for contingency ranking and screening to identify the critical contingencies. Then for each of those critical contingencies, multiple time domain simulations are performed to determine the transfer limit. These methods basically monitor the system transient stability while increasing or decreasing the transfer across the interface which is simulated by adjusting the generation and load in the exporting and importing areas. The detailed mechanism for this adjustment varies. For example, one can adjust the generation and load according to a set of scaling factors. Though the contingency screening signicantly reduces the computational burden, multiple time domain simulations are still required to determine the transfer limit without violating stability constraints for each contingency considered. Another disadvantage of these methods is that the maximum interface MW ow limit is normally not optimal since the generation changes are based on a certain pre-determined scaling. It is observed that this interface limit can be even maximized when the generation changes are distributed based on their sensitivities. To tackle these problems, a new method based on trajectory sensitivity analysis is proposed and presented as follows. The feature of this proposed method is that it utilizes the trajectory sensitivity analysis method to ascertain that the generation increment at each key generator will not result in the system being unstable. Then a constrained optimization problem can be formulated to calculate the maximum real power ow through an interface of interest. The owchart of the proposed method is shown in Fig. 1 and the detailed procedure for this method is as follows:
Fig. 1. Flow chart of the proposed method.

1) Pre-determine an interface that is of interest and identify the key generators and transmission lines across this cut set. Methods to identify the key generators associated with the interface under consideration vary. They can be determined according to the operators experience, historical data or network topology analysis. Normally, they are identied by analyzing the power ow of the base case and identifying the sources that affect the ows on the interface under consideration. The source generators that impact the interface ow are the key generators. 2) Determine the stability limit of the system being studied. This proposed method requires the knowledge of the stability limit of the system in terms of the maximum relative rotor angle when the system is about to lose stability. In order to provide a more general guideline, in this study the critical case is obtained by gradually increasing the interface ow and running repeated time domain simulations for the changed conditions until the system has lost stability. The maximum relative rotor angle under this critical case is denoted as . This step can be skipped if the user is familiar with the system and knows this critical value based on experience. 3) Determine a suitable base case. The principle of determining a base case is that this case should be close to its stability limit so that the linear approximation (5) can be applied. The critical relative rotor angle is determined in step 2. For example, for an operating point, when the maximum relative rotor angle is close to , this case can be identied as the base case.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4 IEEE TRANSACTIONS ON POWER SYSTEMS

4) Calculate the PTDFs of those key generators on those transmission lines that constitute the interface. 5) Identify a set of load buses whose active loads are most likely to change. In the importing area, a set of load buses whose active loads are very likely to change and affect the interface ow needs to be pre-determined. This determination can be based on historical data or operational experience. Calculate the sensitivities of interest simultaneously using the computational cluster. For interface ow limit calculation, two types of sensitivities are needed: one is the sensitivities of the generator rotor angles with respect to active generation change at those key generators; the other one is the sensitivities of generator rotor angles with respect to active load change in the importing area. 6) Calculate the maximum interface real power ow. Solve the following constrained optimization equations to obtain the maximum increment or decrement of the real power ow through the interface.

base case. and are the upper and lower output limit of generator . In order to solve this optimization problem, in (7), it is assumed that the active loads of those load buses change proportionally based on some weighting factors as follows:

(8) is the load increase weighting factor for load bus , where and is the total load increment at all the load buses. The rst constraint equation in (7) ascertains that the total generation increment at all the key generators will not result in system instability. The set should not only contain the most advanced generator, but also include those advanced generators whose rotor angle is close to stability limit for the reason that the any generation increment might result in the second, third or fourth advanced generator becoming unstable. Therefore, this constraint ascertains all the advanced generators are stable when there are generation changes made in the system. The second constraint equation in (7) ascertains that the generation increment at each of the key generators will be within its output limit. The third constraint equation in (7) ascertains that the total generation increment is equal to the total load increment. Solution of (6)(7) yields and then based on (6) is obtained. However, the validity of this requires further examination since (7) is based on rst order linearization, which requires the perturbation size to be small. In this paper, the assessment of the validity of is performed using simulation. A simulation for the case with the changed applied to the base case is run and the plots of variables for the actual changed case are obtained and then compared with the approximated plots obtained by using the rst order linearization (5). If the comparison shows that the accuracy is acceptable, then this is valid and satisfactory. Otherwise, a limit should be put on the size of each changing parameter. For example (9) is a scaling factor to create a limit on the size of the where th parameter change. is the generation change of the th generator obtained from the previous solution of (6)(7). If the accuracy assessment shows that is too large to obtain an acceptable linearization accuracy, then (10) needs to be added as a constraint in the solution of (6)(7). This process is repeated until the solution is validated to be within a certain acceptable accuracy range. The choice of depends on the discrepancy between the actual perturbed value and the predicted value. If the mismatch is large, is recommended to be smaller, e.g., 50%, and if the mismatch is small, can be larger, e.g., 8090%. Then the stability constrained maximum interface real power ow can be determined as (10) where is the interface real power ow for the base case.

(6)

(7)

is the increment or decrement of the real power ow where through the interface. is the active generation change at generator . is the PTDF of generator on line . is the number of key generators in the interface. is the number of transmission lines constituting the interface. In calculating the PTDF the slack bus used in the base case power ow is retained as the slack bus. is the number of load buses whose active load changes are taken into account. The rotor angle difference obtained previously in step 2 is the criteria used to determine rotor angle stability/instability. is a set of the most advanced generators. is the time instant of interest. For this case, it is the time instant at which the most advanced generator reaches the maximum relative rotor angle. is determined by observing the angle plots of all the generators for the base case. The rst few, for example 5 or 10 generators depending on how close they are to the stability criterion , are selected in this set. is the generator rotor angle difference between the th generator in and least advanced generator at time instant . is the value of for the base case. and are the sensitivity of with respect to active generation change at generator and active load change at load bus at , respectively. is the active load change at load bus . is the active generation of generator for the

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
HOU AND VITTAL: DETERMINATION OF TRANSIENT STABILITY CONSTRAINED INTERFACE REAL POWER FLOW LIMIT 5

TABLE I SYSTEM CHARACTERISTICS OF THE WECC SYSTEM

It can be seen from the above description that the computations required in this method include some repeated time domain simulations to determine the stability limit for the system analyzed and some repeated time domain simulation for the determination of the base case. The advantage of the proposed method is that once this base case is found, the proposed method can be adopted to directly determine the interface ow limit instead of using a trial and error method which is the commonly used method in the industry for now. Another component that contributes to the computation burden of this proposed method is the multiple sensitivity calculations, depending on the number of key generators and load buses taken into consideration. Actually, this part of burden can be signicantly relieved by the parallel computing technique, which was introduced in [29]. Regarding the solution to the linear programming problem (6)(7), normally only the generators with the highest sensitivities will be participating in this calculation. Therefore, the dimension of this optimization problem is usually very low even in a large power network. Therefore, compared to the conventional methods, this proposed method is efcient in reducing the computational burden. The result obtained by solving (6)(7) under one specic fault consideration is only applicable for this particular case. For other contingencies, the same procedure should be run to obtain the ow limit accordingly. In an online setting contingency analysis is performed regularly. To obtain the safest interface MW ow limit, the ow limit under each analysis can be determined using the proposed method. The minimum value among these calculated ow limits can then selected as the maximum interface MW ow limit. By doing so, all the critical contingencies at this operating point will not result in system instability. A software package implementing the trajectory sensitivity calculation has been developed. The implementation is based on an open-source Matlab based power system analysis software package called power system analysis toolbox (PSAT) [31]. A computing cluster is used to perform multiple sensitivities evaluations in parallel. A detailed description of this cluster was presented by the authors in another paper [29]. IV. APPLICATIONS AND RESULTS The 2009 summer peak load case of the WECC system is used in the paper to test the performance of the proposed method. The system characteristics are shown in Table I. All the generators are equipped with a turbine governor, exciter and PSS. The total active generation is 165 129 MW. A. Base Case and Interface Descriptions An interface of interest on the WECC system is selected as shown in Fig. 2. There are two major power plants of interest: P1 and P2. Each of these plants contains several generators with high capacity. The congurations of these two plants are listed in Tables II and III, respectively. There are four transmission

Fig. 2. Description of the selected interface.

TABLE II CONFIGURATION OF THE POWER PLANT P1

TABLE III CONFIGURATION OF THE POWER PLANT P2

TABLE IV REAL POWER FLOWS THROUGH THE INTERFACE FOR THE BASE CASE

lines in this interface. The real power ows on these lines for the base case are listed in Table IV. The PTDF is dened as follows: (11) where is the PTDF of generator on line . is the real power ow change on line . is the active generation change at generator . A PTDFs calculation routine has been implemented by the authors in PSAT [31]. The results are listed as follows in Table V. B. Base Case Performance The critical contingency considered is a 3-phase fault on line #4 as depicted in Fig. 2. The fault occurs at s and is cleared after 0.06 s by removing this faulted line. After running the time domain simulation for the base case, it is found that Gens. #43, 44 and 45 are the most advanced generators. Therefore, these three generators are selected to constitute the set of advanced generators dened in (7). It is also identied that Gen. #320 is the least advanced generator. The relative rotor

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6 IEEE TRANSACTIONS ON POWER SYSTEMS

TABLE V PTDFS OF EACH KEY GENERATOR ON THE INTERFACE

TABLE VI SENSITIVITY TO ACTIVE GENERATION CHANGE AT

TABLE VII SENSITIVITY TO ACTIVE LOAD CHANGE AT

Fig. 3. Relative rotor angles with respect to Gen. #320.

angles between these key generators and Gen. #320 are plotted in Fig. 3. 17 load buses whose active loads are considered to change and affect the interface ow are determined. These 17 load buses are directly supplied by the four tie lines. By using the method stated in Section III step 2, after running 4 time domain simulations, the critical relative rotor angles is determined to be approximately 180 . The maximum relative rotor angle for this base case is 165.1 , which occurs at s. It can be seen that this system operating point is very close to its stability limit. The computational efciency of the time domain simulation is improved by using an early stopping criterion for either very stable cases or unstable cases developed in [32]. C. Calculation of Trajectory Sensitivities To solve the optimization problem (6)(7), two types of sensitivities are required. They are: Sensitivities of rotor angles of generators in the set with respect to active generation change at those key generators depicted in Tables II and III. Sensitivities of rotor angles of generators in the set with respect to active load change at the 17 load buses. These sensitivities are calculated using the parallel computing platform described in [29]. It can be seen from Fig. 3 that the maximum relative rotor angle occurs at time instant s, therefore the sensitivities at this time instant are of interest and they are listed in

Tables VI and VII. The sensitivities can also be computed numerically as follows: (12) The results of the sensitivity calculations using this numerical method are also listed in Tables VI and VII for comparison with the analytical method. In these two tables, Analy. denotes the analytical method and Numer. denotes the numerical method. From this comparison it can be seen that the results obtained from these two methods are almost identical. However, in order to evaluate each numerical sensitivity the required change is made and the simulation re-run to calculate the difference between the base case and the change case. D. Solution to the Optimization Problem In order to solve the linear programming problem (6)(7), the following assumptions are made: In (8), it is assumed that is positive. This indicates that all the key generators are increasing their output In (8), it is assumed that the load increments are distributed equally among the 17 load buses. The Matlab built-in solver for linear programming problems linprog is used to solve the linear programming problem (6)(7). Solution to (6)(7) yields MW. At the

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
HOU AND VITTAL: DETERMINATION OF TRANSIENT STABILITY CONSTRAINED INTERFACE REAL POWER FLOW LIMIT 7

TABLE VIII KEY GENERATOR OUTPUT INCREMENT

TABLE X COMPARISON OF THE REAL POWER FLOWS THROUGH THE INTERFACE

Fig. 4. Relative rotor angles with respect to Gen. #320 for the changed case. Fig. 5. Relative rotor angles with respect to Gen. #320with interface ow MW.
S

TABLE IX COMPARISON OF THE RELATIVE GENERATOR ROTOR ANGLES AT

solution, the generation increment of each key generator is listed in Table VIII. Then from (10), the interface real power ow limit can be calculated as MW. E. Result Verication To verify the results obtained above, another time domain simulation is conducted for the change case with the changes listed in Table VIII applied to the pre-fault power ow case. The relative rotor angle plots between the most advanced generators and Gen #320 are shown in Fig. 4. Also, the relative generator rotor angles between the three s are most advanced generator and Gen. #320 at veried and listed in Table IX. In the table, the results in the row From linear programming solution are the results obtained by solving the linear programming problem (6)(7). The results in the row From time domain simulation for changed case are the results obtained from running time domain simulation for the changed condition. Also, the real power ows in the four tie lines are veried and they are listed in Table X. It can be seen from Fig. 4 that the relative rotor angles at s are very close to , which is 180 for this study case. It is also seen from Tables IX and X that the results from the solution to (6)(7) are very close to the actual results from time domain simulation for the changed case.

Then the interface real power ow is increased slightly higher than the amount obtained from the sensitivity analysis in the previous section by 10 MW to 2176.97 MW, the relative angle plots are shown in Fig. 5. It is seen from the Fig. 5 that Gen. #43, Gen. #44 and Gen. #45 are swinging away from the rest of the generators, which indicates that the system is unstable. Thus the results verify the accuracy of the proposed method. F. Discussions of the Test Case In this test case, only the rst-swing instability problem is considered. However, the proposed method is applicable to the second-swing stability problem since the same procedure as introduced in the paper can be applied. During the time domain simulation for the base case, the maximum relative rotor angle will be monitored. As long as the relative rotor angle approaches a certain threshold, it can be concluded that the system is close to stability limit. The sensitivity calculation routine will then be activated to compute the sensitivity trajectories for this base case. The change in the time instant at which the maximum relative angle occurs will not affect the procedure since the method is dealing with the determination of the maximum pre-fault interface real power ow. V. CONCLUSION In this paper, a fast and accurate method based on time domain simulation for the determination of the transient stability constrained interface real power transfer limit is proposed. Compared to current methods, which require many off-line transient stability studies [30] for different operating conditions,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE TRANSACTIONS ON POWER SYSTEMS

this method avoids a large number of repeated time domain simulations to obtain the transfer limit without scarifying the accuracy. Therefore, this method can be used for on-line applications for fast determination of interface power transfer limit, which is of vital importance under a market environment. Besides, the linear programming problem formed to solve for the maximum interface ow increment is of a low dimension and thus can be solved easily with any existing linear programming solver. Therefore, this method has signicant advantage in terms of calculation efciency. The proposed method is also accurate as shown by comparing the results obtained from the proposed method with that from running time domain simulations. It can be seen from Table X that for the interface whose base ow is 1563 MW, the mismatch between the value obtained by the proposed method and the value from traditional methods is 11 MW, which is only 0.7% of the base ow. REFERENCES
[1] NERC, Available transfer capability denition and determination, North America Electric Reliability Council, Jun. 1996. [2] L. Wang and C. Y. Chung, Increasing power transfer limits at interface constrained by small-signal stability, in Proc. IEEE PES Winter Meeting, 2002, vol. 2, pp. 11841187. [3] ISO NE, TTC tables. [Online]. Available: http://www.iso-ne.com/ trans/ops/limits/ttc/new_england_ttc_tables.htm. [4] G. C. Ejebe, J. Tong, J. G. Waight, J. G. Frame, X. Wang, and W. F. Tinney, Available transfer capability calculations, IEEE Trans. Power Syst., vol. 13, no. 4, pp. 15211527, Nov. 1998. [5] X. Zhang, Y. H. Song, Q. Lu, and S. Mei, Dynamic available transfer capability (ATC) evaluation by dynamic constrained optimization, IEEE Trans. Power Syst., vol. 19, no. 2, pp. 12401242, May 2004. [6] M. Shaaban, Y. Ni, and F. F. Wu, Available transfer capability evaluation by decomposition, in Proc. Power Eng. Soc. Summer Meeting, 2001, vol. 2, pp. 11221126. [7] D. Gan, R. J. Thomas, and R. D. Zimmerman, Stability-constrained optimal power ow, IEEE Trans. Power System, vol. 15, no. 2, pp. 535540, May 2000. [8] A. Pizano-Martinez, C. R. Fuerte-Esquivel, and D. Ruiz-Vega, Global transient stability-constrained optimal power ow using an OMIB reference trajectory, IEEE Trans. Power Syst., vol. 25, no. 1, pp. 392403, Feb. 2010. [9] A. L. Bettiol, D. Ruiz-Vega, D. Ernst, L. Wehenkel, and M. Pavella, Transient stability-constrained optimal power ow, in Proc. IEEE Budapest Powertech., Hungary, Aug. 29Sep. 2, 1999. [10] M. Pavella, D. Ruiz-Vega, J. Giri, and R. Avila-Rosales, An integrated scheme for on-line static and transient stability constrained ATC calculations, in Proc. IEEE Power. Eng. Soc. Summer Meeting, 1999, vol. 1, pp. 273276. [11] D. Ruiz-Vega and M. Pavella, A comprehensive approach to transient stability control: Part Inear optimal preventive control, IEEE Trans. Power Syst., vol. 18, no. 4, pp. 14461453, Nov. 2003. [12] V. Vittal, E. Zhou, C. Hwang, and A. A. Fouad, Derivation of stability limits using analytical sensitivity of the transient energy margin, IEEE Trans. Power Syst., vol. 4, no. 4, pp. 13631372, Nov. 1989. [13] Y. Yamada, M. Nagata, and K. Tanaka, An energy function based contingency screening method for ATC assessment with transient stability constraints, in Proc. IEEE/PES Asia Pacic Transmission and Distribution Conf. Exhib. 2002, Oct. 610, 2002, vol. 2, pp. 886890. [14] K. Kim, S. Kim, S. Rhee, S. Lee, and K. Song, Assessment of total transfer capability subject to transient stability energy margin, in Proc. Asia and Pacic Transmission & Distribution Conf. Expo., 2009, pp. 14. [15] A. Berizzi, C. Bovo, M. Delfanti, M. Merlo, and M. S. Pasquadibisceglie, A Monte Carlo approach for TTC calculation, IEEE Trans. Power Syst., vol. 22, no. 2, pp. 735743, May 2007.

[16] W. Li, P. Wang, and Z. Guo, Determination of optimal total transfer capability using a probabilistic approach, IEEE Trans. Power Syst., vol. 21, no. 2, pp. 862868, May 2007. [17] A. A. Fouad and V. Vittal, Power System Transient Stability Analysis Using the Transient Energy Function Method. Englewood Cliffs, NJ: Prentice Hall, 1991, pp. 274282. [18] R. Z. Minano, T. V. Cutsem, F. Milano, and A. J. Conejo, Securing transient stability using time-domain simulations within an optimal power ow, IEEE Trans. Power Syst., vol. 25, no. 1, pp. 243253, Feb. 2010. [19] I. A. Hiskens and M. A. Pai, Trajectory sensitivity analysis of hybrid systems, IEEE Trans. Circuits Syst. I: Fundam. Theory Appl., vol. 47, no. 2, pp. 204220, Feb. 2000. [20] I. A. Hiskens and J. Alseddiqui, Sensitivity, approximation, and uncertainty in power system dynamic simulation, IEEE Trans. Power Syst., vol. 21, no. 4, pp. 18081820, Nov. 2006. [21] T. B. Nguyen and M. A. Pai, Dynamic security-constrained rescheduling of power system using trajectory sensitivities, IEEE Trans. Power Syst., vol. 18, no. 2, pp. 848855, May 2003. [22] B. Sapkota and V. Vittal, Dynamic VAr planning in a large power system using trajectory sensitivities, IEEE Trans. Power Syst., vol. 25, no. 1, pp. 461469, Feb. 2010. [23] M. Zima, P. Korba, and G. Anderson, Power system voltage emergency control approach using trajectory sensitivities, in Proc. IEEE Conf. Control Application, 2003, vol. 1, pp. 189194. [24] M. J. Laufenberg and M. A. Pai, A new approach to dynamic security assessment using trajectory sensitivities, IEEE Trans. Power Syst., vol. 13, no. 3, pp. 953958, Aug. 1998. [25] M. J. Laufenberg and M. A. Pai, Sensitivity theory in power systems: Application in dynamic security analysis, in Proc. IEEE Int. Conf. Control Applications, Dearborn, MI, Sep. 1518, 1996, pp. 738743. [26] C. Ferreira, J. Pinto, and F. Barbosa, Transient stability assessment of an electric power system using trajectory sensitivity analysis, in Proc. 39th University Power Engineering Conf., 2004, vol. 2, pp. 10911095. [27] R. Tomovic, Sensitivity Analysis of Dynamic Systems. New York: McGraw-Hill, 1963. [28] I. A. Hiskens, Iterative computation of marginally stable trajectories, Int. J. Nonlinear and Robust Control, vol. 14, pp. 911924, 2004. [29] G. Hou and V. Vittal, Cluster computing based trajectory-sensitivity analysis application to the WECC system, EEE Trans. Power Syst., vol. 27, no. 1, pp. 502509, Feb. 2012. [30] NYISO, NYISO Interface Stability Limit Analysis for all Lines I/S and Outage Conditions. [Online]. Available: http://www.nyiso.com/ public/webdocs/documents/studies_reports/operating_studies/interface_stability_limit_analysis.pdf. [31] F. Milano, An open source power system analysis toolbox, IEEE Trans. Power Syst., vol. 20, no. 3, pp. 11991206, Aug. 2005. [32] G. C. Ejebe, C. Jing, J. G. Waight, G. Pipper, and F. Jamshidian, Security monitor for on-line dynamic security assessment, in Proc. 4th Int. Conf. Power System Control and Management, 1996, pp. 5864. Guanji Hou (S11) received the B.E. and the M.S. degree in electrical engineering from South China University of Technology, Guangzhou, in 2006 and 2009, respectively. He is currently pursuing the Ph.D. degree in the Department of Electrical, Computer, and Energy Engineering, Arizona State University, Tempe, AZ. His research interests include the area of power system stability and control, applications of information and technology to power system and smart grid.

Vijay Vittal (S78F97) received the B.E. degree in electrical engineering from the B.M.S. College of Engineering, Bangalore, India, in 1977, the M.Tech. degree from the Indian Institute of Technology, Kanpur, India, in 1979, and the Ph.D. degree from Iowa State University, Ames, in 1982. He is the Ira A. Fulton Chair Professor in the Department of Electrical, Computer, and Energy Engineering at Arizona State University, Tempe, AZ. He currently is the Director of the Power System Engineering Research Center (PSERC). Dr. Vittal is a member of the National Academy of Engineering.