ADDITION of new IPP generation, uncertainty in

Transcription

1 1 Fast Fault Screening Approach to Assessing Transient Stability in Entergy s Power System V. Kolluri, Senior ember, IEEE, S. andal, ember, IEEE,. Y. Vaiman, ember, IEEE,.. Vaiman, ember, IEEE, S. Lee, Senior ember, IEEE, and P. Hirsch, Senior ember, IEEE Abstract This paper addresses a fast process for performing transient stability studies in a large transmission system. The paper describes how the most severe three-phase fault locations were identified in Entergy s power system network. The approach described in this paper offers a unique capability to automatically identify the most severe fault locations and perform ranking of these most severe faults in on-line and off-line environments. The study was performed using the Entergy loadflow and dynamic data to validate the proof of concept. It took approximately one minute to determine and rank 3 most severe fault locations in Entergy s power system. The Fast Fault Screening (FFS) capability described in this paper will allow the operators to assess transient (angular) stability in on-line and real-time environments. Index Terms-- Critical clearing time, dangerous fault locations, fast fault screening; transient stability; Ranking Index. I. INTRODUCTION ADDITION of new IPP generation, uncertainty in dispatch and increased power transfers can cause a power system to be more vulnerable to transient stability violations. Identifying weak/critical locations in the system is a very time consuming process that uses time-domain simulation approach, and can be easily overlooked. As a deregulated system is being operated closer to the system limits, system planners and operators need a fast screening tool that will assess the system s stability and identify the most severe fault locations in the system. Some of these locations might be already known to planners/operators while other new locations emerge as the system conditions dynamically change in an on-line operating environment. The standard utility practice is to run a pre-selected list of faults that have been historically known as critical using timedomain simulation technique. Thus, transient stability is usually not assessed in operations and real-time environments. Operators rely on the planning personnel to provide them with the results of transient stability analysis. In a planning environment, this can be a time-consuming and cumbersome task. V. Kolluri and S. andal are with Entergy Services Inc, New Orleans, LA USA ( vkollur@entergy.com; smandal@entergy.com). ).. Y. Vaiman and.. Vaiman are with V&R Energy Systems Research, Inc. Los Angeles, CA 90049, USA ( mvaiman@vrenergy.com; marvaiman@vrenergy.com ). S. Lee and P. Hirsch are with Electric Power Research Institute, Palo Alto, CA 94303, USA ( slee@epri.com, phirsch@epri.com). This paper addresses the development of a fast process for identifying the most severe fault locations. Also, the paper describes how this process was applied to Entergy s network and the most severe three-phase faults were identified. The program "Physical and Operational argins-transient Stability" (PO-TS) was used as the basis for all computations in this project [1]. PO-TS is a fast, user-friendly and comprehensive dynamic simulation tool. PO-TS is designed to determine transient stability limits after any disturbance is applied to a power system network of any practical dimensions. Any system quantities may be selected as an output and can be displayed graphically. PO-TS supports the standard IEEE library of dynamic models. It also allows for an easy inclusion of the userdefined models using PO scripting features without the need for external compilers. A library of standard scripts is provided with PO-TS. PO-TS supports the use of multiple integration methods. The application uses variable time step technique without loss in the accuracy of results. PO-TS was utilized during the present project as the basis for Fast Fault Screening functionality. II. FAST FAULT SCREENING (FFS) CAPABILITY Fast Fault Screening (FFS) capability of PO-TS was used to identify and rank the most severe three-phase faults, in a quick manner. During the study, fault severity was classified as follows (starting with the most severe situation): Post-fault regime does not exist (i.e., voltage instability) Loss of generation Situations corresponding to faults with small critical clearing time The FFS computation proceeds in two steps: Selection of fault locations At this step, locations of the most severe faults are identified. These are the weakest points in the network. Ranking of faults The most severe faults are ranked using the Ranking Index (RI). The FFS performs these steps within the same run. During the first step of analysis, the most severe faults are identified using the heuristic approach. These faults are ranked using the Ranking Index (RI). The index is based on the properties of the energy function method and overcomes the limitations of this method. The /07/$ IEEE.

2 index is an analytical tool with the coefficients derived using regression analysis. There are over 1300 buses in Entergy control area with base voltage of 100 kv and above, which are candidates for fault screening. III. ETHODOLOGY OF DETERINING THE OST SEVERE THREE-PHASE FAULT LOCATIONS To begin with, the most severe faults are identified using the heuristic approach. FFS allows the user to identify buses with large real power flowing through them from local generators. The following criteria are incorporated into FFS: Criterion 1 The difference between real power flow at a bus and generator real power in the vicinity of this bus. Criterion Real power leaving a bus The approach allows us to identify locations of the most severe faults. These are the weakest points in the network. A. Criterion 1: The Real Power Difference The first criterion utilized in the approach for selecting fault locations is the difference between real flow at a bus and generator real power in the vicinity of this bus. Power at a bus is considered as power entering a bus and power leaving a bus. Two rules are enforced for the real power difference: (a) the difference between real flow entering a bus and generator real power in the vicinity of this bus, and (b) the difference between real flow leaving a bus and generator real power in the vicinity of this bus. Criterion (a) considers that only real power output from local generation flows into a bus, which is considered as a severe fault location. This is a user-specified value. Various limits were tested, and, then, the limit of 5% was found to be acceptable and selected for obtaining the study results (i.e., 75% of the generator power flows to the bus). Criterion (b) excludes long-distance flows through a bus, and ensures that only local generation is considered. Power entering a bus should not exceed 10% of the real generator output in the vicinity of the bus. This is a user-specified value. The vicinity of the bus was defined as one or two buses away to account only for local generation. B. Criterion : Real Power Leaving a Bus The second criterion utilized in the approach accounts for the value of real power flow on the lines connected to the bus. This is the real flow that leaves the bus. Only flows on transmission lines are accounted for in this criterion; flows on transformers are not considered. Various limits were tested as shown in Table I. The number of identified fault locations changes with the change of this limit. TABLE I TESTING LIITS FOR CRITERION Power Leaving a Numbers of Fault Bus Locations 500 W 0 50 W W W 46 The limit of 50 W (~1% of generator real output in Entergy control area) was selected for obtaining the study results described in this paper. This is a user-specified value. When Criteria 1(a) and 1(b) are satisfied, and a limit of 50 W is used for Criterion, the number of identified severe fault locations in Entergy control area is 3. IV. ETHODOLOGY OF RANKING THE OST SEVERE THREE- PHASE FAULTS The approach described in Section III was used to identify the most severe fault locations in the Entergy System. FFS capability identified 3 fault locations based on the criteria discussed in Section II. These locations are considered to be the weakest points in the network. The next step was to rank the most severe faults. The Ranking Index (RI) was developed to rank the most severe faults that were selected using the criteria listed in Section III. The RI is a very fast estimation of the fault severity that does not require running time-consuming fault analysis for each fault. A. Using the Concept of the Energy Function as the Basis for the Ranking Index The Fast Fault Screening ranking approach is based on one of the features of the energy function. This feature is related to the change of the shape of the potential energy surface as system becomes more stressed. The potential energy surface of a non-stressed power system has a shape of a potential well (see Fig. 1, left plot). As the system becomes stressed, the surface changes its shape to a trough (see Fig. 1, right plot). The critical clearing time decreases as the surface becomes closer to the shape of the trough. Fig. 1. Potential Energy of a Non-Stressed System (Left) and a Stressed System (Right) This phenomenon was discovered by V&R Energy Systems Research, Inc. under the work supported by the National Science Foundation (NSF) award number III-

3 and described in the Report submitted to NSF Choice of Contingency Arming Schemes Actions Using Analytical Approaches, []. As follows from the energy function, characteristics that have the largest impact on the dynamic stability of a power system are: Generator kinetic energy Generator electrical torque Generator voltage Shape of the potential energy that is represented by the eigenvalue of the Jacobian matrix However, the energy function does not allow one to account for all dynamic models of a realistic power system, such as excitation system, governor models, etc. The proposed approach overcomes this limitation of the energy function by introducing a ranking index. The proposed ranking index allows us to perform the estimation of the shape of the potential energy very fast. The RI does not involve computation of the energy function. B. The Components of the RI The Ranking Index (RI) is based on the power system characteristics that have the largest impact on the dynamic stability. These characteristics are computed using timedomain simulation over a very short time period after a fault has been cleared. These characteristics are the components of the Ranking Index: RI = k 1 * KE + k * AX {( 0; (1 )} + + k 3 * IN {( 0; (1 )} + EigenValu + k4 *( VPREF Vgen ) + k5 ( 1) (1) EigenValu Where KE is the kinetic energy of a generator with the largest value of kinetic energy located in the vicinity of the fault is the electrical torque of the generator with the largest value of kinetic energy located in the vicinity of the fault at the moment the fault is cleared is the electrical torque of the generator with the largest value of kinetic energy located in the vicinity of the fault before the fault is applied EigenValue POSTF is the eigenvalue of the Jacobian matrix in the post-fault regime EigenValue PREF is the eigenvalue of the Jacobian matrix for the base case V PREF is voltage of the generator with the largest value of kinetic energy located in the vicinity of the fault before the fault is applied V gen is voltage of the generator with the largest value of kinetic energy located in the vicinity of the fault at the moment the fault is cleared The components of (1) have the following effect on the RI: Generator kinetic energy A bigger value of the kinetic energy of a generator rotor during a fault corresponds to a bigger chance of loss of synchronism. Generator electrical torque The ratio of the electrical torque at the moment the fault is cleared and before the fault is applied depends on the potential energy of the post-fault regime. These are the values of the electrical torque of the generator with the largest value of the kinetic energy located in the vicinity of the fault. Generator voltage The difference between generator voltage before the fault is applied and at the moment the fault is cleared. Voltage of the generator with the largest value of the kinetic energy located in the vicinity of the fault is considered. Voltage value depends on the parameters of the excitation system, low voltage corresponds to a more severe post-fault regime. Shape of the potential energy that is represented by the eigenvalue of the Jacobian matrix When the eigenvalue of the Jacobian matrix is close to zero, the Newton method diverges. Simulation was performed for each fault using PO-TS for only 0.1 sec in order to determine the values of generator kinetic energy, the electrical torque of the generator at the moment the fault is cleared, and generator voltage at the moment the fault is cleared. The value of RI for each of the most severe faults was compared against the critical clearing time. Critical clearing time for each fault was computed using PO-TS. The larger value of RI corresponds to a more severe fault. The smaller the critical time is, the more severe the fault is. C. The Coefficients of the RI Formula The rank correlation coefficients of the RI formula were derived using linear regression analysis. Computations under the project were made using the Spearman s formula: n 6 () rs = 1 ( xi i) n( n 1) i 1 The results were benchmarked against Kendall s formula 4 Q (3) t K = 1 n ( n 1) Both the Spearman s and Kendall s rank correlation coefficients [3] are symmetrically distributed on the range (-1, +1). Rank coefficients that are closer to +1 correspond to a more accurate ranking mechanism. At first, a criterion for ranking faults was selected. The criterion utilized in this approach is the critical clearing time (CCT). Thus, a fault with the smallest critical time is the most

4 4 severe fault, i.e., it has the highest ranking. Note, that the criterion also incorporates situations that lead to loss of generation and steady-state instability. Deriving rank correlation coefficients proceeded in the following steps: 1. The most severe 3 faults, that satisfy criteria specified in Section III, were identified.. Critical clearing time (CCT) was computed for each fault. 3. Faults were ranked using the value of the CCT; situations that led to loss of generation and steady-state instability were also considered. 4. Spearman s formula was adopted to FFS methodology as follows: n 6 (4) max rs ( k1,..., k5) = max (1 ( xi ( RI) i( CCT)) ) n( n 1) i 1 where k 1 + k + k3 + k4 + k5 = const or k5 = 1 RI is the Ranking Index CCT is the critical clearing time n = 3 5. A run, corresponding to severe fault conditions, was made. These conditions were approximated by outaging 85% of the power leaving a bus (i.e., Cutoff Factor = 85%). Spearman s and Kendall s rank correlation coefficients were determined. 6. The coefficients were tested for a less severe fault scenario, when only 5% of the power leaving a bus was outaged (i.e., Cutoff Factor = 5%). 7. Computation results, obtained in items (5) and (6) above, showed that rank correlation coefficients, used for ranking index formula, were selected very effectively. Based on steps (1) (7) above, the coefficients of the Ranking Index were derived as follows: RI = 1.75 * KE * AX {( 0; (1 )} * IN{(0; (1 )} + 0.5*( VPREF Vgen) + EigenValue POSTF + ( 1) EigenValue PREF where k 1 = 1.75, k = 4.5, k 3 = 0.5, k 4 = 0.5, k 5 = 1. V. UTILIZING PO SUITE OF APPLICATIONS FOR FAST FAULT SCREENING Physical and Operational argins suite of applications was used as the basis for computations. Physical and Operational argins (PO) and PO Transient Stability (PO-TS) applications were utilized in this study. FFS Tab was added to the interface to execute the (5) approach described in this paper (see Fig. ). Fig.. PO-Interface with the FFS Tab A. The FFS Options The FFS Tab contains FFS options that offer the control over the FFS analysis (see Fig. 3). Fig. 3. The FFS Tab There are six FFS options. FFS options are: Control Area Number Control area number, in which fast fault screening is performed. inimum Voltage Level, kv Voltage class of buses that are considered in FFS analysis. inimum Real Power, W The value of the real power flow on the lines connected to a bus. Only flows on transmission lines are accounted for; flows on transformers are not considered. This is the real power flow that flows from the bus (i.e., leaves the bus). The value, entered in this field, is utilized in Criterion. Power Difference, % The difference between real (W) flow at a bus (i.e., entering a bus) and generator real power in the vicinity of this bus. The value, entered in this field, is utilized in Criterion 1.

5 5 Cutoff Factor, % A fraction of the power leaving a bus, which is being outaged during a fault simulation. The factor is used to account for substation configuration and fault scenario. Reactance, p.u. Branch reactance limit. Only branches with the value of reactance equal or greater than the value entered in this field are considered in the fault selection criteria (see Section III). Branches with the value of reactance less than the value entered in this field are not considered. The value is used to account for circuit breakers. D. The FFS Output Upon completing computations, FFS outputs a file that contains information on ranking the most severe faults using the RI. Thus, detailed information about each bus identified by the FFS capability as a severe fault location, and its ranking using the RI is written to the file (see Fig. 5). B. FFS Options Used for the Study FFS options used for the study are: Control Area Number: 151 (Entergy) inimum Voltage Level: 100 kv inimum Real Power: 50 W Power Difference: 5 % Reactance: p.u. There were 3 fault locations identified by the FFS capability. These are the most severe fault locations in the Entergy system provided that the above options are utilized and the criteria described in Section III are satisfied. C. Executing FFS Capability To execute the FFS capability in PO, the user should enter the desired values in the FFS Tab fields, and click on the Execute button on the FFS Tab toolbar. The list of buses, identified by FFS as locations of the most severe faults, is displayed in the Output Tab of the INFORATION Pane (see Fig. 4). Detailed information about each bus and fault ranking using the RI is written to an output file. Clicking on the Show Script button on the FFS Tab enables the echo function and opens the PO script program that is utilized to execute the FFS capability. The user may then copy this script, modify it and reuse it. Fig. 4. Executing the FFS Capability Fig. 5. A Fragment of the FFS Output File The entries in the file are sorted by the value of RI. The entries are sorted in descending order. Thus, fault locations are shown in the order of severity with the most severe fault shown at the top of the file. E. Study Results There were 3 fault locations (i.e., buses) identified by FFS. These are the considered to be the most severe fault locations in the Entergy system based on the options given in Section V.B and the criteria described in Section III. Results of the fast approach were verified using full time domain simulation. Comparison of the value of the critical clearing time (CCT) against the proposed Ranking Index (RI) for the first seven buses that have the highest rank is given in Table II. Number TABLE II COPARING VALUES OF RI AND CCT Bus Number RI CCT, sec Table II shows that buses with the smallest critical clearing time have the largest value of RI, i.e., the highest rank. Two runs were made to simulate different fault scenarios. These runs were made for two different values of Cutoff Factors - 5 % and 85%.

6 6 Cutoff Factor offers the control over the fraction of the power leaving a bus, which is being outaged during a fault simulation. It corresponds to the number of lines being opened after a fault is cleared. The effect of the Cutoff Factor on the post-fault conditions was analyzed. Ranking of the most severe faults changed with the change of the Cutoff Factor. The increase in the Cutoff Factor leads to more severe post-fault conditions. VI. CONCLUSION The present paper describes a process for identifying severe three-phase faults in the Entergy power system. The objectives of this study are: Accurately determine the locations of the most severe faults that can lead to transient and voltage instability. Rank the most severe fault locations in order to identify the weakest locations in the power system network. The study showed that the technique, introduced in this paper, is a very fast approach for identifying and ranking the most severe three-phase fault locations. This method can be added to the existing dynamic simulation tools in order to identify the most severe fault locations in both on-line and the off-line environments. During the first step of analysis, locations of the most severe faults were identified using the heuristic approach. Then, these faults were ranked using the Ranking Index (RI). The index is based on the properties of the energy function method but overcomes the limitations of the method. The index is an analytical tool with the coefficients derived using regression analysis. The criterion used for ranking faults in the current project, is the value of the critical clearing time. A study was performed using the Entergy loadflow and dynamic data to validate the proof of concept. It took under one minute to determine and rank 3 most severe faults in the Entergy s system using the Fast Fault Screening (FFS) approach. The results of the study were verified using full time domain simulations. As this approach looks very promising, Entergy has decided to extend the project for identifying and ranking the unbalanced faults. VIII. BIOGRAPHIES Sharma Kolluri (S 86) received his BSEE degree from Vikram University, India in 1973, SEE from West Virginia University, organtown in 1978 and BA from University of Dayton in He worked for AEP Service Corporation in Columbus, Ohio from 1977 through 1984 in Bulk Transmission Planning Group. In 1984 he joined Entergy Services Inc, where he is currently a anager of Transmission Planning. Sharma has over 0 years of experience in Planning and Operations and his main areas of interests are Power System Planning, Operations, Stability, Reliability and Insulation Coordination. Sujit andal ( 99) received the B.Tech. degree in Electrical Engineering from the Indian Institute of Technology (IIT), Kanpur, India and the.s. degree in Electrical Engineering from Kansas State University, anhattan, KS in 1997 and 1999, respectively. He worked as a consultant at Power Technologies, Inc., Schenectady, NY, from 1999 to 000. Presently, he is with Technical System Planning, Entergy Services, Inc., New Orleans, LA.. ichael Vaiman ( 91) has over 40 years of power industry experience. He received his SEE degree from Kaunas Polytechnic University, Lithuania in 1961, Ph.D. degree from oscow University of Transportation Engineering, Russia in 1969, and D.Sc. degree from St. Petersburg Polytechnic University, Russia in He was a full professor, Department of Automatic Control and Telecommunications at oscow University of Transportation Engineering, Russia until Since 199 Dr. Vaiman is a President and Principal Engineer at V&R Energy Systems Research, Inc. His main areas of interest are power system stability and control, power flow and optimal power flow analysis, computer modeling of power system networks, selection of remedial actions for stability preservation; dynamic stability analysis. arianna Vaiman ( 97) received her BSEE and SEE degrees from oscow University of Transportation Engineering, Russia. She has 15 years of experience in power system studies. In 199 she joined V&R Energy Systems Research, Inc. (V&R), where she is currently Principal Engineer and Executive Vice President. She leads the work in the following areas at V&R: Software Development, Consulting Activities, Research & Development Activities. Steve Lee ( 69, S 75) is the Technical Executive, Power Delivery and arkets, at the Electric Power Research Institute. Dr. Lee has over 30 years of power industry experience. He received his S.B., S.. and Ph.D. degrees from.i.t. in Electrical Engineering, majoring in Power System Engineering. He worked for Stone & Webster Engineering in Boston, Systems Control, Inc. (now ABB) in California. He was Vice President of Consulting for Energy anagement Associates (EA). Before joining EPRI in 1998, Dr. Lee was an independent consultant in utility planning and operation. At EPRI, Stephen Lee is leading technical research programs for grid operations and planning and power markets. He is also active in the North American Electric Reliability Council (NERC) on issues related to interregional operation and planning. He has been actively developing new concepts and tools for power system operation and probabilistic transmission planning. Peter Hirsch (S 99) received his PhD from the University of Wisconsin. He is anager of Software Quality of the Power Delivery and arkets Division, at EPRI, in Palo Alto, California. Dr. Hirsch is also a project manager in Power System Assets, Planning and Operation and was responsible for facilitating the FERC standards and communication protocols for the OASIS project. Dr. Hirsch currently manages the Department of Homeland Security (DHS) Version 1 North America Electric Infrastructure SECurity (NESEC) System, His projects include T-FS, on-line DSA, VSA, and TRACE. Before joining EPRI in 199, Dr. Hirsch worked for IB Corporation from as a manager for advanced systems at the IB Scientific Centers. VII. REFERENCES [1] Physical and Operational argin (PO) Program anual; Physical and Operational argins Transient Stability (PO-TS) Program anual, V&R Energy Systems Research, Inc. Los Angeles, CA, 006. [] Choice of Contingency Arming Schemes Actions Using Analytical Approaches, NSF Grant No Prepared by V & R Co., Energy System Research, Principal Investigator.Y. Vaiman, [3].G. Kendall and A. Stuart, The Advanced Theory of Statistics: Volume, Inference and Relationship, London: Charles Griffin and Company Limited, 1961.