Simulating the Smart Electric Power Grid of the 21st Century – Bridging the Gap between Protection and Planning Ashok Gopalakrishnan Quanta Technology, USA I. Sandro G. Aquiles-Pérez Donald M. MacGregor Daryl B. Coleman Paul F. McGuire Kevin W. Jones Electrocon International, Inc., USA Xcel Energy, USA Jay Senthil James W. Feltes Glenn Pietrow Anjan Bose Siemens-PTI, USA Washington State University, USA Introduction Recent blackouts in the United States such as the September 2011 and August 2003 disturbances were exacerbated by protective relays that unnecessarily tripped electrical facilities. The North American Electric Reliability Corporation (NERC) and the Federal Energy Regulatory Commission (FERC) concluded that overly conservative relay settings, combined with the fact that relay systems operated too quickly, led to large scale outages with millions of people losing power. With respect to protective relay operations during these events, the following factors were identified as contributing to the blackout [1, 2]: • • • • Overly conservative relay settings Cascading relay trips due to high line and transformer loading Relay protection operating too quickly, and not allowing enough time for operators to take mitigating action Coordination of remedial action scheme (RAS) not adequately tested (2011 event) For the August 2003 event, NERC also concluded that “the relay protection settings for the transmission lines (zone 3 impedance), generator and under-frequency load-shedding in the Northeast may not be entirely appropriate and are certainly not coordinated and integrated to reduce the likelihood and consequence of a cascade – nor were they intended to do so.” What is needed, therefore, is a method to simulate the bulk electric power system in a way that models the effect of protective relay operations on the dynamic behavior of the system. The reports issued by NERC/FERC for the disturbances recommend that protection systems including remedial action schemes (RAS) be adequately modeled in planning studies so that the impact of their behavior on system stability can be properly studied. This paper describes an Integrated Protection-Planning Simulation (IPPS) environment that allows engineers to do precisely that. Its main features are: • Uses the protection system model with its thousands of relays (distance, overcurrent, out-of-step, frequency, voltage, V/Hz, etc.) that relay engineers have developed settings for and performed coordination studies on. Georgia Tech Protective Relaying Conference 2014 1 • Uses the transient stability model that planning engineers utilize to study system dynamics and stability. • Simulates the planning and protection models together so that the interdependence between system dynamics and relay actions can be captured and cascading failures can be studied. Relay operation is actually simulated and not assumed. • Allows the study of different contingencies and scenarios, some of which may lead to cascading outages, while others may not. For the ones that do, the engineers will be able to tune the relay settings to confine their effect to the appropriate area. • Provides the ability to study relay behavior for unbalanced faults and the consequent impact on stability. In typical transient stability studies, relay operation is assumed and the associated switching action simulated. Actual relay performance for balanced or unbalanced faults are not actually simulated systematically. • Provides a platform for developing and testing special protection schemes (remedial action schemes) and their associated wide-area protection/control algorithms. Typical applications of the IPPS platform include: • Conventional planning studies, with full consideration of protective relay behavior. Relay settings used are the ones calculated and verified by the protection engineers. • Post-mortem analysis of events where protective relay operations played a part in the blackout. • Operations planning studies – for example, if transmission facilities (line, generator or transformer) are out of service for maintenance or other reason, and if a fault occurs, are the protective relay settings still able to maintain dependability and security? Furthermore, do relay operations create stability problems for the system? • Development of relay settings adjustments to help prevent stability problems while maintaining the ability to operate quickly and securely for faults. • Checking, documenting and maintaining compliance with NERC protection and planning standards. Integrating protection and planning studies in a single environment brings other benefits as well: • The planning and protection network models will be better aligned with each other, allowing for easier exchange of data between the two departments. While the time horizon for the two functions is different, the basic network being modeled is the same. • Regulatory bodies should appreciate this convergence. Georgia Tech Protective Relaying Conference 2014 2 In the rest of this paper, we provide more details about the integrated environment, including information about how unbalanced faults are handled and the issues/problems one would face in the initial preparatory work that needs to be performed. A case study is also presented in some detail in the paper. In the summer of 2008, a lightning strike temporarily knocked out more than 800MW of generation in the service area of a utility in the southwestern United States. Other transmission paths had previously been removed from service for maintenance or other reasons. This caused the remaining tie lines to neighboring systems to be tripped on overload. An under-frequency condition then developed in the utility’s service area resulting in the shedding of more than 600MW of load. Portions of this event were recreated using the integrated simulation platform. II. The Protection and Planning Environments Planning and protection engineers typically work with their own models of the network data. Detailed protection information is not included in the planning study. Dynamic behavior of the electrical system (machine and load dynamics, frequency effects) are not modeled in a protection coordination study. The simulation tool of choice for protection engineers is usually a protection simulation program with the following features: • • • • • • • Short-circuit fault capability. Generators modeled as ideal voltage sources behind a fixed impedance. Accurate zero-sequence representation of network components (lines, transformers, machines, mutual coupling). Accurate protective relay models with internal supervision logic and comparators, resembling as closely as possible the actual relay. Teleprotection (pilot) schemes. Single-pole tripping, relay reset characteristics, and automatic reclosing. Buses modeled in detail to allow simulation of bus splitting schemes and breaker failure relaying. For planning engineers, the simulation environment is an electromechanical transient stability / load flow program with the following features: • • • • • Detailed models of machine behavior. Detailed models of excitation, governing and stabilizing systems. Non-linear load-voltage and load-frequency characteristics. Full-featured load flow tools. Basic models of protective relay functions. It should also be noted that transient stability programs operate in the positive-sequence domain – the domain in which system dynamics occur. Furthermore, transient stability simulations are performed in the time domain whereas protection simulations are fundamentally phasor-based calculations (steady state), including the quasi-time-domain stepped-event simulation method  used by many utilities today. Georgia Tech Protective Relaying Conference 2014 3 II-A. Including Protection Information in Planning Studies Transient stability programs do provide several models of protective devices including overcurrent and distance relays, but these models mimic relay functional behavior, rather than the “relay” as a manufacturer-specific device as would be done in a protection simulation environment. While relay models exist in the planning environment, they are traditionally under-utilized because of lack of relay setting data, which is maintained by the protection engineers. Manual transcribing of relay setting information for several thousand relays is not practical. Therefore an automated technique would be needed to transfer relay settings from the protection model to the planning model. There are several issues to consider: • • • Utilities do not normally have such automated procedures in place. Representation of relays in the protection environment is a lot more detailed than their representation in the planning environment. All that detail must be abstracted and converted to a generic representation of the relay that the transient stability environment needs. Since the transient stability environment is based on a positive-sequence network representation, the relay models that exist in that environment can respond only to positive-sequence (balanced three-phase) disturbances. Unbalanced faults must be simulated using a positive sequence equivalent of the unbalanced fault and the phase voltages and currents due to the unbalance are not calculated. Thus actual relay operation cannot be determined and we have to assume that they operate correctly. Given these issues, most utilities include relay behavior in the planning environment by maintaining a document that specifies actual times when lines, transformers or machines would be tripped open by relay action and incorporating those times into their stability studies. These times are obtained (in a manual fashion) from the protection department and entered into the planning case. In the wake of the September 8, 2011 blackout event in Arizona and Southern California, the Western Electricity Coordinating Council (WECC) was mandated by FERC to improve the modeling of protective relays and special protection schemes within the standard planning/dynamics studies. This representation must be shared among all entities responsible for the reliability of the system. To achieve this objective, WECC has set up a new task force (September 2012) called the Modeling Special Protection Schemes and Relays Ad-hoc Task Force (MSRATF). One of the goals of the MSRATF is to include relay behavior automatically in a stability study, and thereby eliminate the need to manually enter switching times into the stability runs to represent relay operations. The MSRATF will also develop specifications for relays that respond to unbalanced fault conditions. These tasks are ongoing and already progress has been made towards completing and approving a specification for overcurrent relays. It is important to note that the WECC approach will bring protective relay behavior into the planning environment by abstracting the detailed relay models that exist in the separate protection environment. Whenever there is a change made to relay settings, these changes must be communicated with appropriate conversion/adaptation to the planning domain to ensure the fidelity of planning studies. This is certainly a valid method but requires care to maintain compatibility. Georgia Tech Protective Relaying Conference 2014 4 In contrast, the type of protection-planning integration that we are reporting in this paper is a lot more ambitious in the sense that it does not require relay settings and relay model details to be abstracted and brought over into the planning environment. It directly combines the protection model (actively maintained by the protection engineers) with the planning model (actively maintained by the planning engineers), and allows for a unified study to take place. III. Integrated Protection Planning Simulation (IPPS) Platform The IPPS platform bridges the two separate planning and protection simulation programs by making them communicate with each other. In very general terms, the communication is carried out in a closed loop wherein: 1. The transient stability program computes the initial voltage profile at all buses in the electrical network, based upon a certain known loading condition in the network and topology. This steadystate initial condition is superimposed on the protection model. 2. Next, the transient stability program simulates the occurrence of a contingency (fault), evaluates network dynamics, and recalculates the positive-sequence voltage profile. 3. The protection simulation tool takes this positive-sequence voltage profile, calculates the negativeand zero-sequence voltage profiles as needed, and determines the operation of protective relays in the network. The operating times of the relays determine when the next circuit breaker is going to open and change the network topology. This time of operation is passed back to the transient stability calculation. 4. The transient stability program then modifies its network to account for this topological change (open or close breaker), simulates the dynamics of the generators and their controls, and recomputes a new positive-sequence voltage profile that it passes to the protection simulation program (to Step 3 above). The flow chart in Figure 1 summarizes this closed loop behavior: Georgia Tech Protective Relaying Conference 2014 5 PS: While present time < Total Simulation Time PS: -Read new voltage profile from TS -If unbalanced fault, convert positive-sequence voltage to unbalanced set -Calculate relay currents -Advance PS simulation by one time step -Evaluate relay operation and determine if any breakers are going to operate -Pass breaker operation information to TS TS: -Open (or close) breaker (line, transformer, machine, etc.) -Advance TS simulation by one time step -Calculate new positive-sequence voltage profile and return control to PS PS: Protection Simulation TS: Transient Stability Figure 1: Closed-Loop Interaction between Protection Simulation and Transient Stability. The closed loop can run in several modes: • • • Unattended with appropriate breaker operations, until the total simulation time elapses. Under user control with user asking the simulation to advance to the next time step. Under user control with user asking the simulation to advance to the next breaker operation. Not shown here are the initialization steps; these can be found in Appendix A at the end of the paper. IV. Modeling Unbalanced Faults The network model used in planning studies typically does not contain zero-sequence and negativesequence information. The dynamic behavior of the network is modeled in the positive-sequence only. The protection network model on the other hand is a full three-sequence model, with detailed zerosequence representation of transformers, mutual coupling, etc. The question then arises – how do we model unbalanced short-circuit conditions like single-line-toground or line-to-line faults? The response of the protection system, and therefore its effect on the dynamic behavior of the network, are definitely of interest under these conditions. Georgia Tech Protective Relaying Conference 2014 6 IV-A. Representing an Unbalanced Fault Condition as a Balanced Fault The four standard fault types that protection engineers work with are the single-line-to-ground (SLG), line-to-line (LTL), double-line-to-ground (DLG), and three-phase (TPH) faults. In short circuit analysis, a three phase system is decomposed into a set of decoupled sequence networks, namely, the positive-, negative-, and zero-sequence networks. Depending on the type of fault, one or more of the sequence networks will appear in the fault definition. For example, the SLG fault, when “looked at” from the faulted node, will consist of the positive-, negative-, and zero-sequence circuits connected in series through three-times the fault impedance ZF. Each of the sequence circuits consists of the Thévenin impedance in that circuit, denoted by Z1, Z2 and Z0 for the positive-, negative- and zero-sequence respectively. Further, the positive-sequence circuit includes a voltage source – the prefault voltage at the faulted node, denoted by VF. Then, for a SLG fault, the fault current IF may be computed from IF = VF Z1 + Z 2 + Z 0 + 3ZF (1) To represent this fault as a balanced fault in the transient stability simulation, we can simply add Z2, Z0 and 3ZF, and apply an equivalent positive-sequence shunt with this impedance at the faulted bus in the planning network (note that Z1 is already modeled in the network and hence is not included in the equivalent positive sequence shunt). As long as the positive-sequence Thévenin impedance Z1 is the same between the two networks, and the negative- and zero-sequence data in the protection model has been verified as accurate, the fault current IF produced by the balanced network will be the same as the fault current produced by the unbalanced network. Please note that the prefault voltage VF must be obtained from the planning network model. The other standard fault types are modeled in a similar manner, albeit with different combinations of the sequence impedance for the different fault types. Appendix B describes the connection of sequence circuits for the standard fault types. IV-B. Converting the Positive-Sequence Voltage Profile Obtained from the Dynamic Simulation into an Unbalanced Set of Three Phase Voltages At each time step, the voltage profile obtained from the dynamic simulation is impressed upon the protection simulation network. Then, the currents that protective relays “see” are computed by using the voltages of the end-buses of the relay branch and the impedance of the branch including any line charging. Georgia Tech Protective Relaying Conference 2014 7 This method works well when the applied fault is a three-phase, balanced fault. In that case, the positive-sequence voltage obtained at each time step from the dynamic simulation can be used directly to compute the relay currents. The individual phase voltage and current magnitudes will be equal, with angles 120° apart from each other. However, if an unbalanced fault is being simulated, we can no longer use the positive-sequence voltage alone. The following method is used to convert the positive-sequence voltage to an unbalanced set of three phase voltages: • • • The fault type and the fault location are known. Positive-sequence voltages at all buses neighboring the fault location are obtained from the dynamic simulation. These can be used directly. At the fault location, compute the positive-sequence current contribution from each of the neighbor buses, based on the voltages that were obtained from the dynamic simulation. Bus A I1A V1F V1A I1B Bus B V1B Figure 2: Recreating unbalanced voltages in the protection network from the balanced voltages obtained from the transient stability simulation. Let V1A, V1B and V1F be the positive-sequence voltages obtained from the dynamic simulation calculation in the present time step at buses A, B and the fault point F, respectively. The faulted bus must be modeled explicitly in the planning model so that its positive-sequence voltage can be obtained. V1A, V1B and V1F are used to compute the positive-sequence currents I1A and I1B flowing into the fault. The sum of these two currents is the total positive-sequence current in the unbalanced fault. An alternative method to calculate the positive-sequence fault current in the protection network is to use the voltage V1F (computed by the dynamic simulation), and divide it by the same positivesequence equivalent shunt impedance that was used to simulate the unbalanced fault in the planning network. As an example, for the SLG fault, the positive-sequence fault current can be calculated as I1F = V1F Z 2 + Z 0 + 3ZF (2) Then the total positive-sequence current and the arrangement of the sequence circuits for the given fault type are used to compute the negative- and zero-sequence currents that will flow into the fault. For the SLG fault example, I2F = I0F = I1F. Appendix B describes the connection of sequence circuits for the standard fault types. • These sequence currents are injected into the protection network model at the fault point. Georgia Tech Protective Relaying Conference 2014 8 • This enables calculation of the negative- and zero-sequence voltages at all buses in the protection network model by means of the following: YBUS,0 ⋅ ∆V0 = I0F YBUS,2 ⋅ ∆V2 = I2F • (3) where InF (n = 0, 2) is the appropriate sequence fault current into the network at the faulted node (zero at all non-faulted nodes), YBUS,n (n = 0, 2) is the nodal bus admittance matrix in the appropriate sequence, and ∆Vn (n = 0, 2) is the vector of change in sequence voltage from prefault to fault, at all buses in the network. The pre-fault negative- and zero-sequence voltages are zero, so ∆Vn (n = 0, 2) equals the desired post-fault solution. The positive-sequence voltage obtained from the dynamic simulation is combined with the negativeand zero-sequence voltages from equation (3), to derive the unbalanced set of three phase voltages. It is very important to note that the source of the positive-sequence voltage at each time step is the transient stability program. This voltage reflects the dynamic behavior of network components including machines and loads. The technique described above also assumes that the negative- and zero-sequence networks do not change over time (except related to changes in topology), in contrast to the positive-sequence network which is greatly affected by the time dynamics of the machines. This assumption allows the positivesequence voltage calculated by transient stability to be combined with the negative- and zero-sequence voltage calculation by the protection simulation program to generate an unbalanced set of phase voltages. It should also be pointed out here that transient stability programs generally include a short-circuit engine that is capable of calculating unbalanced fault conditions, provided the data is present. However, the unbalanced fault calculation engine is usually not incorporated into the time dynamics of the transient stability engine. Commercially available transient stability programs are expected to have this ability in the near future. IV-C. Non-Standard Faults For faults more complicated than the standard ones, it is impractical to determine the connection of sequence networks by hand. While one could potentially pre-compute some of the more common nonstandard multi-node faults like “one phase open” or “two phase open”, it is preferable to develop a general solution, which can then be applied for any type of condition, including single-pole tripping (opening exactly one pole of a circuit breaker, instead of all three). Sometimes, the only way to meet very short critical clearing time intervals is to trip the circuit breaker in just the faulted phase. Then the ability to model single-pole tripping in the simulation becomes important and a necessary capability. A new technique that will allow modeling of a general fault connection, without the need for special fault-type-specific algorithms has been developed. This technique makes use of the fact that a fault involving N electrical nodes, each node having 3 phases, can be reduced by Gaussian elimination to a positive-sequence, single-phase equivalent with N(N-1)/2 series branches, and N shunts [4, 5]. The Georgia Tech Protective Relaying Conference 2014 9 generators influence system dynamics only in the positive sequence, because the internal EMF has only a positive-sequence component. That is, if N = 1 (a standard shunt fault), we need just one shunt to represent the fault as a positivesequence equivalent (0 series branches). If N = 2 (one phase of a circuit breaker is open, the other two phases are closed), we need one series branch, and two shunts (one at each end of the series branch) to represent the open-phase condition as a positive-sequence equivalent. This is the well-known PI branch model. The positive-sequence equivalent so determined is submitted to the TS calculation as usual and applied as a disturbance. At the next time step, the positive-sequence voltage profile at all buses, including the ones involved in the fault, is obtained from TS and imported into the protection system calculation. This set of voltages and knowledge of the phase domain representation of the fault (converted to sequence coordinates) allows accurate reconstruction of the negative- and zero-sequence post-fault voltage at all buses in the network. Combined with the positive-sequence voltage obtained from transient stability, the unbalanced set of phase voltages can be calculated and the relay response evaluated. V. Tools for Prototyping Special Protection Schemes (Remedial Action Schemes) Most protective relays and associated functions monitor and take action based on local inputs. In the most common case, relays measure voltage, current, and frequency at the locations where they are employed. Another type of protection scheme that is used is called the special protection scheme, or SPS . Such special protection schemes have also been called remedial action schemes (RAS) or more recently, system integrity protection schemes (SIPS). Unlike conventional relaying, SPS may be designed to consider inputs from geographically dispersed locations. The inputs can be analog – voltage, current, frequency, real and reactive power; or digital – switch or circuit breaker status. Based on these inputs and pre-defined algorithms, the SPS might initiate a variety of control actions. If the output control actions include not just tripping of breakers but also controlling analog values like voltage or power flow (say, by the use of FACTS devices), then such a scheme is referred to as wide area control (WAC). Thus, the SPS/WAC is a dynamic security system, which controls the stability of the network from a point of view that is much larger than that of the local protective relaying. The utilization of such SPS/WAC systems to increase the efficiency (power transfer) and the reliability (avoid cascading outages) of the grid is one of the fundamental characteristics of the smart grid. The IPPS environment contains prototyping tools to help the engineer design and test such wide area protection and control schemes. With this prototyping tool, the engineer can • Select inputs to be monitored (both analog and digital). Georgia Tech Protective Relaying Conference 2014 10 • • Specify the action to be taken – typically open one or more circuit breakers to trip/outage a circuit element. Test the algorithm within the framework of the integrated protection and planning system (IPPS). Figures 3, 4 and 5 below show the implementation of the SIPS prototyping tool. Figure 3: SIPS Prototyping Tool – Definition of Logic Input: Avalon_Bio_MW_Flow. To the left of the form, we see a list of External Logic Inputs. These inputs can be provided a suitable name. The form shows four such user-defined names – Avalon_Bio_MW_Flow, Avalon_V_pu, Bio_V_pu, and Center_V_pu. The quantity Avalon_Bio_MW_Flow is defined as a Power System Quantity, measuring the MW flow on the transmission line from the Avalon 115kV station to the Bio 115kV station. The comparison type is “>”, which means that the logic input will assert TRUE if the MW flow from Avalon to Bio exceeds the constant value of 900MW, as shown above. The other logic inputs Avalon_V_pu, Bio_V_pu and Center_V_pu measure the positive-sequence voltage magnitude at three different buses. (Entries for Avalon_V_pu are shown below.) Georgia Tech Protective Relaying Conference 2014 11 Figure 4: Definition of Logic Input: Avalon_V_pu. The logic input will assert TRUE if the magnitude of the positive-sequence voltage at the Avalon 115kV bus drops below 0.8pu. The user can specify many such inputs and combine the outputs of these inputs via a Boolean expression to specify a condition under which the SIPS element will operate. See Figure 5. Figure 5: SIPS Prototyping Tool – Outputs. In the example above, the SIPS element will produce an output if one of the Avalon_Bio_MW_Flow, Avalon_V_pu, Bio_V_pu or Center_V_pu conditions becomes true. Typically, the output of the SIPS element will trip one or more circuit breakers to isolate a line, transformer or generator. Georgia Tech Protective Relaying Conference 2014 12 VI. Aligning the Protection and Planning Network Models The positive-sequence voltages calculated by the transient stability program have to be superimposed on the protection network model in each time step. Similarly, when the relays in the protection model operate to open/close a breaker, a corresponding topology change has to be made in the transient stability planning network model. Therefore, mapping between the buses and branches in the two network models must be established. This is crucial to making the integrated simulations work properly. It is also perhaps the hardest task to accomplish among the various tasks that need to be performed before the integrated simulations can take place. VI-A. Bus Mapping Issues Bus numbering in the planning model is dictated by the reliability coordinator for the region to which the utility belongs. Protection network models do not always follow the same numbering scheme, although we have seen several utility network models where major bus numbers in the protection and planning models are the same. This requires a conscious decision on the part of the two groups to make the numbers match. If no such correspondence exists, the mapping needs to be manually performed. Another common scenario is the presence of detailed bus models – ring, breaker-and-a-half, etc. – in the protection network model. The planning network model may contain just a single electrical node to represent that bus, as shown in Figure 6. A Electrical node in planning model B Detailed bus in protection model Figure 6: Mapping an Electrical Node to a Detailed Bus. To ensure that breaker openings in the protection model are accurately represented in the planning model, the bus in the planning model must also be represented with the same level of detail. This is the only way in which the two networks can be kept topologically aligned during a simulation. Georgia Tech Protective Relaying Conference 2014 13 This means creating additional nodes in the planning network, and will require interaction between the planning and protection departments. At least one transient stability program will support this bus detail in the near future. The bus map is stored in a text file, an example of which is shown below. Note that the bus numbers are different in the two models. The equivalence is defined in the file and is used by the IPPS platform. % Source Network = PLANNING MODEL % Target Network = PROTECTION MODEL % Source Network = 360001 8SHAWNEE FP 500.00 = 360002 7SHAWNEE FP 345.00 = 360003 5SHAWNEE FP1 161.00 = 360004 5SHAWNEE FP2 161.00 = 360005 5SHAWNEE G12 161.00 = 360006 5SHAWNEE G34 161.00 = … … … etc. VI-B. 1 2 3 4 5 6 Target Network Shawnee 8 500.00 Shawnee 7 345.00 Shawnee 5-1 161.00 Shawnee 5-2 161.00 Shawnee 5-A 161.00 Shawnee 5-B 161.00 Branch Mapping Issues At each time step in the integrated simulation, the protection model receives a new set of voltages from the stability calculation. Using the bus map information, the voltages are superimposed on the corresponding buses of the protection model. Then the relays are evaluated, which means branch currents have to be computed. A common problem that arises is shown in Figure 7. A B MU C Figure 7: Branch Mapping Issues. Let us assume that a map exists for both buses A and B. In the planning model, a branch between A and B is defined. But in the protection model, the branch between A and B is modeled with more detail. One node called MU is present to help with modeling partial mutual coupling to another line between A and C. If relay evaluation at A and B is to take place correctly, it is not sufficient to obtain voltages at A and B only. We must know what the voltage at the MU bus is and thus this voltage must also be provided by the stability calculation. Georgia Tech Protective Relaying Conference 2014 14 Therefore, the planning network model must be modified to insert a tap-bus on the branch from A to B, at the appropriate location. The line A-C must also be tapped at the appropriate location in the planning model. Like the bus map, the branch map is also stored in a text file; an example of which is shown here: % Source Network = PLANNING MODEL % Target Network = PROTECTION MODEL % Source Network 22553 05BROADF 360106 8SULLIVAN TN 22569 05NAGEL 360106 8SULLIVAN TN 22804 05HOLST1 360476 4BOONE HP 23072 05STINNE 360452 5PINEVILL KY 300035 8NEWMAD 338187 8DELL 5 300103 5NEWMAD 361261 5TIPTONVL TN … … … etc. VI-C. 1 1 1 1 1 1 = = = = = = = 106 106 476 452 6041 6079 Target Network Sullivan 8 6029 Broadford 8 Sullivan 8 6033 Nagel 8 Boone HY 4 6025 Holston 4 Pineville 5 6023 Stinnet AEP5 Dell, Ark 8 6083 New Madrid 8 New Madrid 5 1261 Tiptonvill 5 1 1 1 1 1 1 Additional Notes on Bus and Branch Mapping The process of developing the bus and branch mapping can seem daunting, but there are some mitigating factors that can help with the process: • In our experience, the protection model bus numbers usually match the planning model bus numbers, at least at all the major high voltage buses. This is because the planning model bus numbers must follow numbering rules laid down by the reliability authority the utility reports to. Further, the protection network model is often initially generated from the planning network model, which will ensure bus number matching to a large extent. There will be divergence when the protection network is enhanced with load taps, detailed buses etc., but there now is a starting point. • Sometimes, the protection model bus numbers are offset from the planning model bus numbers by a fixed integer. In this case, the process of developing the initial bus map can be automated, and then enhanced. • The bus and branch maps can be developed in stages. It is important to have accurate and reliable bus and branch maps only in the area where relays are to be simulated. So, one could start by building the maps for that area first, and then expand the area in stages as needed. • Network alignment is a one-time process. After the first map is developed, it only needs to be maintained as the networks actually change. In our opinion, the benefits accrued from the integrated protection-planning simulations easily outweigh the initial effort a utility would face in preparing for these integrated simulations. Georgia Tech Protective Relaying Conference 2014 15 VII. Relevant NERC Standards The integrated protection-planning simulation tool can be used to check and maintain compliance with several NERC standards. • PRC-019-1 Coordination of Generator Voltage Regulator Controls with Unit Capabilities and Protection: Study coordination of machine excitation protection (loss-of-field, V/Hz) with machine capability and with line protection relays. • PRC-023-2 Transmission Relay Loadability: Ensure that protective relay settings do not interfere with operators’ ability to take action to protect network reliability and at the same time reliably detect fault conditions and operate as needed. • PRC-025-1 Generator Relay Loadability: Ensure that backup distance protection applied on generators does not trip during non-critical system disturbances – those that do not pose a direct threat to the machine and associated equipment. (Standard under development.) • PRC-026-1 Stable Power Swing Relay Loadability: Ensure that distance relays do not operate on stable power swings to remove transmission and/or generation from service. (Standard under development.) • PRC-027-1 Protection System Coordination for Performance During Faults: Being developed in conjunction with the revision of PRC-001-1. Will address coordination of protection systems between generator and transmission owners. • Transmission Planning (TPL) standards to evaluate system performance under category A (no contingencies), B (loss of single element), C (loss of two or more elements) and D (cascading outages) events. VIII. Case Study – Verification of a System Islanding Event The following case study illustrates an application of the tool described above. In the summer of 2008, a lightning strike temporarily removed about 800MW of generation from service. A major 345kV tie line to a neighboring utility had been out of service at the time. The resulting load import of more than 800MW into the utility’s service area caused two other tie lines (115kV and 230kV) to trip almost simultaneously. An under-frequency condition then developed in the network resulting in more than 600MW of load being shed. The last remaining 345kV tie line also tripped, effectively islanding the utility’s network from the Eastern Interconnection. The lost generation came back online and stabilized the islanded system. Restoration commenced within one hour of the loss of generation and load. Portions of this event were recreated using the integrated simulation platform and salient information is presented in this section. Georgia Tech Protective Relaying Conference 2014 16 VIII-A. Summary of the Disturbance The following figures summarize the main events that occurred during the disturbance. • Network Prior to the Disturbance: The state of the network prior to the disturbance is shown below. The substation names have been changed. There was a net import into the utility’s service area of approximately 150MW. The TOTAL substation, site of the generation, is connected to the network, via paths not shown in Figure 8 below. Figure 8: State of the network prior to the disturbance. The 345kV line from TURBO to ORBIT was out of service. The arrows on the lines indicate relay locations of interest. • Disturbance: A lightning strike near the TOTAL station resulted in the turbine controllers detecting a power load unbalance (PLU). They closed the steam intercept valves to prevent overspeeding of the generators. Generation was reduced by more than 800MW (Figure 9). Figure 9: Generator control systems at TOTAL closed their respective intercept valves, resulting in loss of generation of more than 800MW. Georgia Tech Protective Relaying Conference 2014 17 • Two tie lines tripped at 2s: The “TROUBLE-STRIFE” 115kV, and “GINGKO-ELVIS” 230kV tie lines tripped approximately 2s after the reduction of generation due to PLU action. A zone 1 relay at STRIFE operated. Out-of-step protection was not enabled at STRIFE, but was enabled at TROUBLE. At GINGKO, another zone 1 relay operated. Out-of-step protection was not enabled at the GINGKO terminal either. See Figure 10. Figure 10: At 2s after the reduction in generation, the TROUBLE-STRIFE 115kV and GINGKO-ELVIS 230kV tie lines tripped. Zone 1 operation was seen at both STRIFE and GINGKO. Neither terminal had out-of-step protection enabled at that time. Out-of-step protection was present and enabled at TROUBLE. • Under-frequency load shedding between 2 to 3s: Several under-frequency relays operated between 2 to 3 seconds following the reduction in generation (Figure 11). Figure 11: Several under-frequency load shedding relays operate – 640MW of load is shed. The large red X marks indicate loads being dropped. • Last tie line tripped, generation recovered and system was islanded (4s): At around 4s after the reduction in generation, the last remaining 345kV tie line “FROST-HIGHTOP” tripped on out-of-step (Figure 12). During this period, the intercept valves on the generators reopened, and generation Georgia Tech Protective Relaying Conference 2014 18 recovered. But the system was now islanded from the Eastern Interconnection. Restoration of the load that was shed and reconnection of the tie lines commenced within the hour. Figure 12: The third tie line “FROST-HIGHTOP” 345kV tripped at around 4s, leaving the system operating islanded from the Eastern Interconnection. The steam generation intercept valves reopened and generation recovered. With this introduction, we can now consider the recreation of the disturbance using the integrated simulation platform. VIII-B. Recreating the Event – Preliminary Tasks Prior to running the integrated simulations, key preliminary tasks had to be performed. Since the event occurred in 2008, and we were trying to recreate it five years afterward, several issues had to be addressed. • Protection Model (Network): Obtain the protection network model, including all its relays from the utility. This model is a current model. With help from the utility engineers and historical information contained in the protection model database, we tried to put the protection network model in the state it was in at the time of the disturbance. The protection network model consists of 8,000 buses. • Planning Model: Obtain the planning network model consisting of load flow and dynamics data from the concerned reliability organization. The planning model consists of around 60,000 buses and represents the Eastern Interconnect system. This model is also a current model. While a load flow/dynamics model from 2008 was available, the presence of certain proprietary dynamics libraries made it unsuitable for use in the integrated simulation platform. Therefore, we had to work back from the present-day planning model to the 2008 status. It was critical to do this in the utility’s own service area, especially near the tie lines. Several generators had gone into service since 2008; load profiles had changed; lines had been reconfigured – tracking and reversing these changes had to be done. Georgia Tech Protective Relaying Conference 2014 19 Even then, this backtracking was not entirely complete and resulted in divergence between the actual event and its recreation, as reported later. After the event in 2008, the utility analyzed the event, including associated relay and SCADA records. This helped them to identify issues with their relay settings, and adjust them accordingly. However, the utility did not attempt to recreate the event using the simulation tools they had at their disposal. Therefore a planning model that could be used as the starting point for the event recreation reported here was not available. • Network alignment – Bus/Branch Mapping: Bus and branch mapping was aided by the fact that the utility uses the same bus numbering scheme in their protection and planning network models at 69, 115, 230 and 345kV. However, all major buses at 115kV and up were modeled in detail in the protection model. This required modifications to the planning model so that bus splitting due to relay action could be accurately represented in the planning model. A large number of branches required tapping in the planning model due to presence of mutual coupling nodes in the protection network. • Protection Model (Relay Settings): The utility traditionally does not include under-frequency relays in their protection model since they could not be simulated prior to the availability of this integrated platform. These relays were added so that load could be shed. An important part of this exercise was the need to restore relay settings to their values at the time of the disturbance. Historical relay setting information stored in the protection model database allowed us to perform this task efficiently. With these preliminary tasks out of the way, the event could now be studied. VIII-C. Recreating the Event Using the Integrated Protection Planning Simulation Platform i. Time Line of Major Events During the Simulation The table below gives a time line of the major events that occurred during the simulation, and will help with understanding the information presented in the sub-sections that follow. Event No. 1 Time in seconds measured from start of simulation 0.00 Time in seconds measured from loss of generation –1.00 2 1.00 0.00 3 2.00 1.00 4 2.30 1.30 Georgia Tech Protective Relaying Conference 2014 Highlights Simulation starts with initial load flow profile superimposed on the protection network. Intercept valves on both generators at TOTAL start to close; power output drops. Generator output power reduces significantly. Tie lines importing more than 800MW. Out-of-step blocking asserts at TROUBLE on the 20 5 2.65 1.65 6 3.00 2.00 7 4.00 3.00 8 4.25 3.25 9 4.69 3.69 115kV tie-line to STRIFE. Out-of-step blocking asserts at FROST on the 345kV line to HIGHTOP. Zone 1 trip at STRIFE on 115kV line to TROUBLE. Zone 1 trip at GINGKO on 230kV line to ELVIS. Under-frequency relays operate to shed more than 600MW of load. Generators at TOTAL recover, intercept valves start to open. FROST terminal on 345kV FROST-HIGHTOP line trips on out-of-step condition. System is islanded from the Eastern Interconnection. Table 1: Time Line of Major Events During the Simulation. ii. TOTAL Generating Station Prior to the Disturbance A one-line diagram of the TOTAL generating station with its two 540MW units is shown below. Prior to the disturbance, these units were producing 980MW. They were also absorbing some reactive power (operating in the underexcited region) at the time of the disturbance. Generation at 24kV is stepped-up to 230kV and then transmitted to the rest of the system. The diagram does not show all of the connections at the 230kV buses – only two are shown. Georgia Tech Protective Relaying Conference 2014 21 Figure 13: TOTAL generating station with two units producing about 980MW. Not all connections at the 230kV buses are shown. The 345kV “TURBO – ORBIT” line was out of service. This is shown below. Figure 14: TURBO-ORBIT 345kV line out of service. iii. Intercept Valves Closed 1s into the Simulation – Continue to 2s from Start The intercept valves on the TOTAL units were closed 1s from the start of the simulation. Generation starts to drop. The figures below show several quantities of interest. Georgia Tech Protective Relaying Conference 2014 22 • Generator Output as Measured on the GSU: The graph below shows the MW/MVAr outputs of one of the units at TOTAL, 1s after the intercept valve closed. Figure 15: Power output of unit 2 at TOTAL 2s into the simulation. The intercept valve closes at 1s into the simulation. Power is measured at the machine terminal, flowing into the GSU. • Load Import on the 115kV Tie Line “TROUBLE-STRIFE”: The TROUBLE-STRIFE line is importing close to 100MW following the reduction in generator power (Figure 16). Figure 16: Load Import on 115kV Tie Line “TROUBLE-STRIFE”. • Load Import on the 230kV Tie Line “GINGKO-ELVIS”: The GINGKO-ELVIS line is importing a little over 200MW following the reduction in generator power (Figure 17). Georgia Tech Protective Relaying Conference 2014 23 Figure 17: Load Import on 230kV Tie Line “GINGKO-ELVIS”. Georgia Tech Protective Relaying Conference 2014 24 • Load Import on the 345kV Tie Line “FROST-HIGHTOP”: This tie line import below is shown as export because both FROST and HIGHTOP stations belong to the utility. FROST is the boundary bus, and connects to two other neighboring utilities. It is convenient to monitor flows at FROST, flowing into HIGHTOP (Figure 18). Figure 18: Load import on 345kV Line “FROST-HIGHTOP”. Data is shown as load export is because both FROST and HIGHTOP belong to the utility. At FROST, there are two connections to neighboring utilities. The total import on all tie lines is more than 800MW, 1s after the disturbance. • Frequency at Peking 230kV Bus – a Major Interchange Station The frequency at Peking, a 230kV bus in the utility’s network was tracked by SCADA during the actual event. With a view to comparing the simulated event with the actual event, the plot of the frequency at this bus as reported by the simulation, is shown below in Figure 19. Comparison with the actual frequency plot is shown later on in this section. Georgia Tech Protective Relaying Conference 2014 25 Figure 19: Frequency at Peking 230kV Bus. iv. Continue to 2.65s from the Start – 1.65s after the Closing of the Intercept Valves at TOTAL At 1.65s after the closing of the intercept valves, the out-of-step blocking asserts at two stations – TROUBLE 115kV station on the line to STRIFE (2.30s from start), and FROST 345kV station on the line to HIGHTOP (2.65s from start). See Figure 20. Figure 20: Out-of-Step Blocking Asserts at TROUBLE and FROST. • Out-of-Step Blocking at TROUBLE: Figure 21 shows the assertion of out-of-step blocking at TROUBLE. t=2.24s t=2.44s OSBD = 3.62c = 0.06033s. Zone 1 OOS (INNER) OOS (OUTER) Figure 21: Out-of-step blocking at TROUBLE asserts at 2.30s from the start of the simulation (1.30s from closing of the intercept valves). The impedance trajectory enters the outer out-of-step Georgia Tech Protective Relaying Conference 2014 26 characteristic at 2.24s. The out-of-step block delay is 3.62 cycles (0.06s). The impedance enters the inner out-of-step characteristic at 2.44s. The block delay times out at 2.30s, 0.14s before the trajectory can enter the inner characteristic, and the relay declares an out-of-step block condition. • Out-of-Step Blocking at FROST: The impedance trajectory enters the outer out-of-step characteristic at 2.42s from the start (1.42s from closing of the intercept valves). The out-of-step block delay is 14 cycles (0.23s). This delay times out before the impedance can enter the inner characteristic and outof-step blocking is asserted at 2.42+0.23 = 2.65s from the start of the simulation. OSBD = 14c = 0.233s. t=2.42s t=2.42s OSBD = 14c = 0.233s. Zone 1 OOS (INNER) OOS (OUTER) Figure 22: Out-of-step (OOS) blocking is asserted at 2.65s from the start of simulation. The zoomed-in graph to the right shows that at 2.65s, the impedance has not quite entered the inner OOS characteristic. Therefore, the FROST terminal does not trip during the first slip cycle. It does trip during the second slip cycle as shown below. Georgia Tech Protective Relaying Conference 2014 27 v. Continue to 3s from the Start – 2s from the Closing of the Intercept Valves at TOTAL Zone 1 relays at GINGKO on the 230kV tie line to ELVIS and at STRIFE on the 115kV tie line to TROUBLE trip almost simultaneously (Figures 23 and 24). Note that STRIFE is the neighboring utility. Neither of these stations had out-of-step protection enabled. Out-of-step protection was present and enabled at TROUBLE. A block signal at TROUBLE had already been generated at 2.30s into the simulation. Figure 23: Zone 1 Tripping at STRIFE and GINGKO. Two of the tie lines have now tripped, leaving only the FROST-HIGHTOP 345kV connection in place. Zone 1 at GINGKO Zone 1 at STRIFE Figure 24: Zone 1 elements at STRIFE (2.963s) and GINGKO (2.988s) operate to open their respective circuit breakers. Measured apparent impedance exits the characteristic after the circuit breakers open. Circuit breaker time is considered – therefore, apparent impedance does not leave the characteristic until after breaker interrupting time elapses. Out-of-step protection is not enabled at either terminal. Georgia Tech Protective Relaying Conference 2014 28 vi. Continue to 4s from the Start – 3s from the Closing of the Intercept Valves at TOTAL Between 2 and 3s after the loss of generation at TOTAL (3-4s from start of simulation), several underfrequency load shedding relays operated to drop more than 600MW of load. The operation of these relays is shown in the text report in Figure 25. Under-frequency relay operations have been shaded. LZOP Summary Report ------------------Operating Times (s) at 2.746 seconds from start of simulation: ---------------------------------------------------------------------------Substation ID LZOP Name Type Trip Path LZOP Breaker Total ---------------------------------------------------------------------------STRIFE 1710 STRIFE-TROUBLE LINE 21P1 2.913 0.050 2.963 GINGKO 193 GINGKO-ELVIS MISC 21P1.S 2.938 0.050 2.988 CREAM #4 907 CREAM #4 UF MISC 81D1T 3.371 0.050 3.421 CREAM #3 904 CREAM #3 UF MISC 81D1T 3.371 0.050 3.421 CREAM #1 900 CREAM #1 UF MISC 81D1T 3.371 0.050 3.421 CREAM #2 902 CREAM #2 UF MISC 81D1T 3.371 0.050 3.421 EMERALD 1504 CB 5D70 UF MISC 81D1T 3.371 0.050 3.421 TROUBLE 2691 0612-0616 UF LINE 81D1T 3.379 0.050 3.421 HOPE 219 HOPE TO SHM LINE 81D1T 3.450 0.083 3.533 CUBA 682 Cuba UF MISC 81D1T 3.496 0.050 3.546 … … … and many more … … … SING BOWL GREAT 1595 VCB S345 1594 VCB S290 1111 S330 DISTR. MISC MISC MISC 81D1T 81D1T 81D1T 3.979 3.979 3.983 0.050 0.050 0.050 4.029 4.029 4.033 Figure 25: Under-frequency relay operations take place between 3 and 4s from the start (2 and 3s after closing of the intercept valves). Note that the last three operations shown indicate breaker operations just beyond 4s. More than 600MW of load is shed. LZOP = Local Zone of Protection (relay panel). vii. Continue to 4.69s from the Start – 3.69s from the Closing of the Intercept Valves at TOTAL At this time, the circuit breaker at FROST on the 345kV line to HIGHTOP opens on an out-of-step condition. The second slip cycle was faster than the first one, and the relay at FROST declared an out-ofstep trip condition at 4.29s, when the apparent impedance entered the inner out-of-step characteristic from the right. However, the out-of-step trip signal is issued only when the impedance trajectory leaves the inner OOS characteristic on the left. At this time, the angular separation between the equivalent voltage sources at the ends of the FROST-HIGHTOP line is reducing and this helps with keeping transient recovery voltages across circuit breaker contacts low, which reduces the chances of breaker failure. The term Trip On the Way Out (TOWO) is used to denote this type of out-of-step tripping. See Figure 26 below. Georgia Tech Protective Relaying Conference 2014 29 2nd slip cycle – OOS trip issued on the way out of the inner OOS characteristic at 4.64s. Breaker opens at 4.69s. 1st slip cycle 1st slip cycle Zone 1 OOS (INNER) OOS (OUTER) 2nd slip cycle - OOS OUTER 4.19s - OOS INNER 4.29s - OSBD = 0.23s. - Swing fast enough for OOS TRIP. Figure 26: On the second slip cycle, an out-of-step trip condition is asserted by the relay at FROST at 4.29s. The relay is configured to trip when the trajectory is on its way out of the inner OOS characteristic. Circuit breaker opens at 4.69s from the start of the simulation (3.69s from loss of generation). viii. Continue to 6s from the Start – 5s from the Closing of the Intercept Valves at TOTAL Figure 27 shows the power output from one of the generators at TOTAL. The intercept valves on the generators at TOTAL would have started to open at 4.25s (3.25s after they first closed). This allows generator output to recover. The simulation was stopped when 6s had elapsed. Georgia Tech Protective Relaying Conference 2014 30 Figure 27: Power output of unit 2 at TOTAL, 5s after the closing of its intercept valve (6s into simulation). Power is measured at the machine terminal, flowing into the GSU. VIII-D. Comparing the Simulation and the Actual Disturbance In general, the events that occurred during the simulation also occurred during the actual disturbance. Protective relay behavior was replicated in the simulations to a large extent. The table below compares relay behavior and the times at which they occurred. Note that all times are measured with respect to the closure of the intercept valves on the generators at TOTAL. No. 1 2 3 4 Relay Operation TROUBLE-STRIFE 115kV tie line opens at STRIFE GINGKO-ELVIS 230kV tie line opens at GINGKO Various under-frequency relays operate to shed load FROST-HIGHTOP 345kV line opens at FROST In the Actual Disturbance (s) 2.175 In the Simulation (s) 2.175 1.988s Between 2 and 3s. First UF at ~2.4s. 4.027s Between 2 and 3.2s. First UF at 2.421s 3.690s 1.963s Table 2: Comparing relay operation times between the actual disturbance and the simulation. Times shown are measured with respect to closure of the intercept valves at TOTAL. While the relays that operated in the disturbance also operated in the simulation, there was some discrepancy as to the time of operation of the FROST out-of-step trip. In the simulation, the 2nd swing through the FROST-HIGHTOP 345kV line was faster than the 2nd swing during the disturbance in 2008. This resulted in the relay at FROST operating around 0.3s faster in the simulation, compared to its operation time during the disturbance. The differences between the simulation and the actual disturbance can also be seen in the frequency recorded by SCADA at the PEKING 230kV bus, shown below in Figure 28. Breaker opening times at GINGKO and STRIFE are within 0.21s of the times that they opened at during the disturbance. Breaker opening at FROST was 0.34s faster in the simulation. Also note that 1s after the intercept valves close, frequency has already started to recover in the simulation after the first minimum, whereas it is still reducing in the actual event and does not start to increase until sometime after that. However, the first minimum frequency seen in the simulation and disturbance are fairly close to each other – 59.55Hz. Georgia Tech Protective Relaying Conference 2014 31 Intercept valves close FROST Opens GINGKO & STRIFE Open GINGKO & STRIFE Open FROST Opens Intercept valves open Figure 28: Frequency comparison at the PEKING 230kV interchange bus. Simulation results are shown on top; actual recorded frequency is shown at the bottom. The time scales are the same. Beyond this first minimum, frequency recovers to 59.85Hz in the simulation, where as it recovers to a little over 59.6Hz in the disturbance. Then, the frequency starts to reduce again and this is when under-frequency relays start to operate – between 2 and 3s after closure of the intercept valves. In the simulation, the frequency drops to 58.75Hz, whereas it drops to only 59.16Hz in the disturbance. Also, this second minimum persists for almost 0.75s in the disturbance, while it lasts for only about 0.25s in the simulation. Georgia Tech Protective Relaying Conference 2014 32 When FROST opens on out-of-step, the frequency at PEKING has recovered to 60.1Hz in the simulation. In the disturbance, frequency recovers to 59.46Hz when FROST opens. Beyond this point, the frequency during the disturbance reduces to 59.32Hz, and then finally starts to recover. In the simulation, this third frequency minimum is not seen. Note that the intercept valves opened at approximately 3.25s after they closed (4.25s into the simulation) and generator output started to recover. If the valve opening actually took place later than 3.25s after they closed, frequency recovery in the simulation could take longer, and match up better with actual data. This remains to be tested. These observations suggest that a more detailed investigation of the load and generation profiles in effect at the time of the disturbance is necessary – not only in the utility’s own service area but also in the neighboring system. However, protective relay behavior was replicated to a large extent, and in that sense, the case study’s attempt to recreate the event and validate the integrated platform’s simulation technique can be deemed a success. With better modeling of the primary network components, it should be possible to achieve even better agreement between the simulations and the event. VIII-E. Other Comments After the event occurred, the utility performed an analysis and determined that several improvements could be made to their relay settings and settings philosophy. Some of these are: • • • Enable out-of-step protection on all tie lines. Ensure that both ends of the tie lines block phase distance elements for stable power swings. If tripping on out-of-step is to occur, ensure that it occurs when angular separation between the equivalent sources at the ends of the tie line is less than 90° and reducing, so that stress to the circuit breaker can be minimized. Several relay settings were changed since 2008, but a thorough validation of the new settings has not been carried out yet. The integrated protection-planning simulation platform is an ideal tool for this purpose. Georgia Tech Protective Relaying Conference 2014 33 IX. Other Applications of the IPPS Platform In this section, a few additional applications of the IPPS platform are presented. In each example, the protection and planning models are concurrently simulated: the protection model indicates when and where circuit breakers should operate, while the planning model simulates machine dynamics and computes the resulting positive-sequence voltage profile. All examples are from real utility networks, although not all are actual events that occurred, unless explicitly stated. IX-A. Delayed Fault Clearing Leads to Generator Out-of-Step Condition Figure 29 shows a generating station with 6 units, operating at 13.2kV. The high-voltage 161kV bus is connected to several 161kV lines. Multi-phase faults on these lines, if cleared in zone 2 time (approximately 0.5s), will cause a generator out-of-step condition. Zone 1 Zone 2 Figure 29: Multi-phase fault on 161kV line out of a generating station is cleared in zone 2 time (~ 0.5s) due to failure of all communication channels. This leads to a generator out-of-step condition. The plots in Figure 30 show the rotor angle of one of the generators and the voltage and frequency at the terminals of the same machine over a 1.3 second period. Both the stable case (pilot relays operate correctly) and the unstable case (failure of all communication channels) are shown. This event has not actually occurred in the utility’s network, but performing the integrated protectionplanning simulations gives the utility the ability to study the situation, and check if protection modeled on the generators behaves properly. The utility can also check how relays on neighboring lines respond to the changing machine dynamics. This would not have been possible in a routine coordination review using classical short-circuit analysis. Georgia Tech Protective Relaying Conference 2014 34 Fault: 0.033s Zone 1: 0.100s Pilot: 0.108s Fault: 0.033s Zone 1: 0.100s Zone 2: 0.600s (A) (B) Zone 1 Zone 1 Pilot Zone 2 Fault Fault (C) (D) Zone 1 Pilot Zone 1 Zone 2 Fault Fault (E) (F) Figure 30: (A) Machine rotor angle – stable, (B) Machine rotor angle – unstable, (C) Machine terminal frequency – stable, (D) Machine terminal frequency – unstable, (E) Machine terminal voltage magnitude and angle – stable and (F) Machine terminal voltage magnitude and angle – unstable. Annotations on the graph indicating relay operation (zone 1, pilot, zone 2) have been placed at the times at which network topology changed due to circuit breaker operation. Clearly, the failure of communication channels and the subsequent fault clearing in zone 2 time (0.6s, inclusive of circuit breaker interrupting time) results in an unstable situation. For this example, generator protection relays were not part of the protection simulation environment, and thus did not remove the generator from service. Georgia Tech Protective Relaying Conference 2014 35 IX-B. Absence of Breaker Failure Protection and Long Time Delay on Generator Backup Relay Results in Generator Continuing to Feed the Fault In this example, a multi-phase fault is applied on a 115kV line close to the Mimosa generating station as shown in Figure 31. The circuit breaker on the 115kV line is stuck (fails to operate). The two 13kV units are equipped with backup distance relays, looking into the 115kV network. The time delay on these distance elements is 1.4s (84 cycles). Stuck Breaker Figure 31: Fault on 115kV Line Close to Generating Station with Stuck Breaker on the Line. The utility in question does not apply breaker failure protection at voltages below 230kV due to a rigorous maintenance schedule on primary and secondary equipment. Several zone 2 relays at remote stations operate in an attempt to clear the fault. The backup distance relay on the generator fails to operate because the impedance measured by the relay does not stay inside the element characteristic for the set time delay of 1.4s as shown in Figure 32. Mechanical overspeed protection is set to trip the generator when its speed reaches 3960rpm (10% over nominal). This mechanical protection was not modeled in the simulations. The utility is investigating adding out-of-step protection for the generators. Georgia Tech Protective Relaying Conference 2014 36 1 5 4 2 3 Figure 32: Trajectory of the apparent impedance measured by generator backup distance relay is shown at 1.3s into the simulation. The impedance does not stay inside the relay characteristic for the requisite 1.4s (84 cycles). On the last entry into the characteristic (end of arrow marked 3), it appears that the trajectory might stay inside. Note that the 1.4s clock will reset to 0 each time the impedance leaves the characteristic. Figure 33: Frequency at the terminals of the two units at Mimosa. Mechanical over-speed protection would operate to disconnect the generators (not modeled in the simulations). Georgia Tech Protective Relaying Conference 2014 37 IX-C. Under-frequency Load Shedding Event This example describes an actual under-frequency event , which was subsequently simulated (partially) in the IPPS environment. • • • • • • • • • A high-resistance single-line-to-ground fault occurred on a 138kV transmission line. Lack of adequate ground fault impedance coverage in the 21N/67N relays prevented this fault from being cleared quickly. Stage 1 under-frequency relaying operated (49.2Hz, 0.15s) and shed 41MW of load. Nominal system frequency is 50Hz. The fault evolved into a two-phase fault, and phase distance relays operated to clear the fault. Frequency recovered to its nominal value. After a manual reclose of the line, the high-resistance fault developed again, more than 60s afterward. The fault was perceived as a load on the system. The 21N/67N relays were not sensitive enough to detect the fault. Stage 2 under-frequency relaying then operated (48.9Hz, 0.15s) and shed another 43MW. The fault became less resistive and was cleared by ground distance relays. The simulations as shown in Figures 34 through 36 were stopped after the operation of the first stage of under-frequency relaying because of the more than 60s duration between the first and second stage relay operation. Figure 34: 30Ω Single-Line-to-Ground Fault 82% from OLDHARB1. Figure 35: Ground distance relay is unable to detect the fault. Directional ground time-overcurrent 67NT is present, and also not able to detect the fault. Georgia Tech Protective Relaying Conference 2014 38 21P operates to clear fault Several UF relays operate Figure 36: Frequency at a Load Bus. Several UF relays operate to disconnect loads at around 1s and frequency starts to recover. The fault changes to a B-C fault, and phase distance relays operate to clear the fault at around 1.1s. Frequency continues to recover. IX-D. Check of Loss-of-Field Relay Settings The upcoming NERC standard PRC-019 requires loss-of-field protection relays to coordinate with machine capability curves and any voltage regulating system controls, including excitation limiters. Figures 37 through 39 demonstrate the use of the IPPS platform to simulate a loss-of-field condition and subsequent loss-of-field relay operation, thereby verifying the relay settings. Figure 37: 18kV Generating station. Loss-of-field condition is simulated on generator S1. Georgia Tech Protective Relaying Conference 2014 39 Figure 38: Loss-of-field relay characteristics along with machine capability and excitation limit curves. This display provides static verification of the loss-of-field relay settings. 9.886s 5.682s 0.17s Delay 5s Delay Figure 39: Apparent impedance trajectory is shown entering the larger loss-of-field circular characteristic at 5.682s. This element is delayed by 5s before it is allowed to issue a trip signal. The smaller loss-of-field element operates at 9.886s; it is additionally delayed by 0.17s so that it issues a trip at 9.886 + 0.17 = 10.056s to disconnect the generator. Note that the larger element would have issued a trip only at 5.682 + 5 = 10.682s, and is preceded by the smaller element (at 10.056s). Georgia Tech Protective Relaying Conference 2014 40 X. Conclusions Utilities have long recognized the need to include protection system behavior in transient stability planning studies. However, this behavior has not been modeled systematically. The action of protective relays for both balanced and unbalanced faults is assumed, rather than simulated. The information presented in this paper describes a new simulation platform that combines the separate protection and planning environments and makes them interact with each other. The planning model simulates the dynamics of the system and generates a positive-sequence voltage profile of the network. The protection model, using its knowledge of the zero- and negative-sequence networks, recreates the unbalanced set of voltages in the network. Relay operation is evaluated in the protection model and breaker operations are communicated back to the planning model. This type of integrated simulation allows high-fidelity evaluation of the impact of protective relay operations on network stability. It can be used as a post-mortem analysis tool and to tune relay settings with a view to reducing cascading outages. The integrated platform also allows modeling and testing of special protection schemes (remedial action schemes). To be sure, there are several steps that must be performed before such a simulation can be embarked upon. One of the most critical is the alignment of the protection and planning network models. But the alignment, or mapping, once developed, just has to be incrementally adjusted as the networks change. NERC and FERC have begun calling for better representation of protective devices in planning studies. The integrated simulation platform presented here is one way to meet that requirement. Acknowledgments The work presented in this paper was performed in part under a US Department of Energy (DOE) grant for Small Business Innovation Research (SBIR), Phases I and II, DE-FG02-08ER85193, . The efforts of John J. Quada of Electrocon International, Inc. in programming several important features of the IPPS platform are gratefully acknowledged. Georgia Tech Protective Relaying Conference 2014 41 References  FERC and NERC Staff, “Arizona-Southern California Outages on September 8, 2011 – Causes and Recommendations,” April 2012.  U.S. – Canada Power System Outage Task Force, “Final Report on the August 14, 2003 Blackout in the United States and Canada: Causes and Recommendations,” April 2004.  Paul F. McGuire, John J. Quada, and Daryl B. Coleman, “A Stepped-Event Technique for Simulating Protection System Responses,” VI Seminário Técnico de Proteção e Controle, Natal - RN, Brazil; September 27 to October 2, 1998.  Sergio Porto Roméro, Ricardo Diniz Rangel, Fernando Hevelton Duarte Oliveira, and Sergio Luis Varricchio, “Modelo de Seqüência Positiva de Linhas com Abertura Monopolar para Estudos de Estabilidade Transitória,” 9th Symposium of Specialists in Electric Operational and Expansion Planning, Rio de Janeiro, Brasil; May 23-27, 2004.  I. Rossi, S. P. Romero, A. R. Carvalho, and O. A. Cunha, “Implementação e Validação de Modelos de Seqüência Positiva para Estudos de Estabilidade em Linhas com Religamento Monopolar”, 10th Symposium of Specialists in Electric Operational and Expansion Planning, Florianópolis, Brasil; May 21-25, 2006.  P. M. Anderson, “Power System Protection,” chapter 21, “Protective Schemes for Stability Enhancement,” IEEE Press, 1998.  Russell W. Patterson, Hortnel Johnson and Marvin Watson, “Analysis of 138kV Tree Fault in Jamaican Public Service Company System,” 12th Annual Fault and Disturbance Analysis Conference, Atlanta, GA, April 20-21, 2009.  Electrocon International, Inc., “Simulating the Smart Electric Power Grid of the 21st Century,” Final Report submitted to US Dept. of Energy, under DOE grant DE-FG02-08ER85193, March 30, 2012. Biographies Ashok Gopalakrishnan is a Research Engineer at Quanta Technology. For the past 14 years, he has been involved in developing advanced software tools for relay protection and coordination studies, for use in the Computer Aided Protection Engineering (CAPE) program. His current focus is on combining protection and planning models and their associated calculations to study the dynamic behavior of the electric grid and system stability in response to protective relay operations. Ashok is a graduate of the Birla Institute of Technology & Science, Pilani, India and Texas A & M University, College Station, Texas, USA. He is a member of the IEEE and CIGRE. Sandro G. Aquiles-Pérez has worked as an Application and Development Engineer at Electrocon International, Inc. since 2007. His main work is the development of detailed protective relay models for phasor- and time-dependent environments, as well as the production of software tools for protection studies for CAPE. He is a graduate of the Instituto Tecnológico de Oaxaca, Mexico; he received his M.S. Georgia Tech Protective Relaying Conference 2014 42 and Ph.D. degrees from the Instituto Politécnico Nacional, Mexico and the University of Saskatchewan, Canada, respectively. Donald M. MacGregor is a Lead Engineer at Electrocon International, Inc. He received his B.A. degree in Mathematics with Honors from St. Catharine's College, Cambridge, England in 1970. Then he attended University College of North Wales in Bangor, where he earned his Ph.D. in Electronic Engineering in 1973. He joined Electrocon in 1973, initially analyzing helix traveling-wave amplifier tubes and electron guns. More recently he has made significant contributions to software for fault studies, phasor models of power transformers, and the automatic setting of multi-function relays for power system protection. Daryl B. Coleman has worked as an Application and Development Engineer at Electrocon International, Inc. since 1993. During the last 20 years he has contributed to many aspects of CAPE development, with a primary focus on data conversion, data management and user interface. He holds a B.S. (EE) from Michigan State University. Paul F. McGuire is currently the President of Electrocon International, Inc. He has participated in every phase of Electrocon’s activities since 1974 and has been the Product Manager for the Computer-Aided Protection Engineering (CAPE) software since its inception in 1986. He holds a B.S. (Engineering Physics) from the University of Maine; M.S.E. (EE) from the University of Michigan; and E.E. (Professional Degree, Electrical Power Systems) from the University of Michigan where he was a Detroit Edison Fellow. Mr. McGuire has been a registered professional engineer since 1984. He is a member of the Institute of Electrical and Electronics Engineers, Conference Internationale des Grands Réseaux Électriques a Haute Tension (CIGRÉ), Tau Beta Pi, Sigma Pi Sigma, Phi Kappa Phi. Kevin W. Jones is a Principal Engineer in the System Protection Engineering Department at Southwestern Public Service Company (SPS), an Operating Company of Xcel Energy. He has worked in various departments at SPS over the last 24 years, including Distribution Design, Substation Commissioning, Transmission Operations, and System Protection Engineering. Kevin specializes in highvoltage transmission line relaying, event analysis, and system stability relaying. Kevin graduated from the University of Missouri—Columbia in 1989 with bachelor’s degrees in Electrical and Computer Engineering. He is a registered professional engineer in the state of Texas and a member of IEEE. Jay Senthil is a Power Engineer with a background in electric power system analysis, modeling, and software development. He has over fifteen years of experience in writing production grade software for power system applications. He has developed advanced applications including load flow, fault analysis, trouble call outage analysis for Distributed Management Systems (DMS), scheduling applications for the deregulated power industry, and EMTP-type programs with detailed models of generators, turbinegovernors and HVDC converters and converter controls. Since joining Siemens PTI in 2001, he has been responsible for the stability component of PSS®E; for development, maintenance and customer support. James W. Feltes is a Senior Manager in the Consulting Department with Siemens, PTI. He received his BSEE degree with honors from Iowa State University in 1979 and his MSEE degree from Union College in 1990. He joined Power Technologies Inc. (PTI), now Siemens PTI, in 1979. At PTI, he has participated in many studies involving planning, analysis and design of transmission and distribution systems. He is an instructor in several of the courses taught by PTI. He is a member of several IEEE committees, working groups, and task forces dealing with power system stability and control and has also participated in the development of CIGRE technical brochures. He is a Senior Member of the IEEE and is a registered professional engineer in the State of New York. Georgia Tech Protective Relaying Conference 2014 43 Glenn Pietrow is a software engineer with a background in CAD modeling and interface design. He has been part of Siemens PTI for 23 years. He began as a GUI designer, programming for all PSS® products, eventually focusing on the PSS®E GUI in 1996. After working extensively with the PSS®E product, Mr. Pietrow became the PSS®E Product Manager in 2004. Today, Mr. Pietrow is still very much involved with the GUI development for PSS®E. Anjan Bose received the B. Tech. degree from IIT, Kharagpur, the M.S. degree from University of California, Berkeley, and Ph.D. from Iowa State University. He has worked for industry, academia and government for 40 years in electric power engineering. He is currently Regents Professor and holds the endowed Distinguished Professor in Power Engineering chair at Washington State University, where he also served as the Dean of the College of Engineering & Architecture from 1998 to 2005. Dr. Bose is a Member of the US National Academy of Engineering, a Foreign Fellow of the Indian National Academy of Engineering and a Fellow of the IEEE. He is the recipient of the Herman Halperin Award and the Millennium Medal from the IEEE, and was recognized as a distinguished alumnus by IIT Kharagpur and Iowa State University. Georgia Tech Protective Relaying Conference 2014 44 Appendix A: Flowcharts Summarizing the Behavior of the IPPS Environment The interaction between the protection and planning simulation programs is summarized here. There are three phases to consider. In the flow charts below, PS indicates the Protection Simulation Program and TS indicates the Transient Stability Program. Initialization Phase PS: Invoke the Link PS: Specify -TS files for load flow and dynamics -Bus Map and branch map file -∆t for the simulations (in dynamics file) -Initialize the TS program TS: -Move simulation time to -2∆t (steady state) -Calculate new voltages at all buses -Return control to PS Figure 40: Initializing the Integrated Simulations. At the end of the initialization, the protection and planning networks are aligned with each other. That is, given a bus or a branch in the protection model, we know the corresponding bus or branch in the planning model. Disturbance Application Phase: PS: -Read and update all mapped bus voltages from TS -Initialization complete, ready to run scenario -Specify fault type, location, total simulation time T -If unbalanced fault, convert to positive-seq. equivalent -Run PS to time = 0.0 TS: -Run simulation to time = 0.0 -Apply disturbance and simulate dynamics -Determine voltages at all buses -Return control to PS Figure 41: Application of the Disturbance. Georgia Tech Protective Relaying Conference 2014 45 Main Simulation Loop: PS: While present time < Total Simulation Time PS: -Read new voltage profile from TS -If unbalanced fault, convert positive-sequence voltage to unbalanced set -Calculate relay currents -Advance PS simulation by one time step -Evaluate relay operation and determine if any breakers are going to operate -Pass breaker operation information to TS TS: -Open (or close) breaker (line, transformer, machine, etc.) -Advance TS simulation by one time step -Calculate new positive-sequence voltage profile and return control to PS Figure 42: Main Simulation Loop. The simulation loop can run in several modes: • • • Unattended with appropriate breaker operations, until the total simulation time elapses. Under user control with user asking the simulation to advance to the next time step. Under user control with user asking the simulation to advance to the next breaker operation. Georgia Tech Protective Relaying Conference 2014 46 Appendix B: Modeling Standard Faults The four standard fault types that protection engineers simulate are the single-line-to-ground (SLG), line-to-line (LTL), double-line-to-ground (DLG), and three-phase (TPH) faults. As explained earlier, to simulate unbalanced faults in the positive-sequence domain of transient stability, we have to provide a suitable fault impedance comprising appropriately connected Thévenin zero- and negative-sequence impedances to the transient stability calculation. The transient stability program will then use this information to simulate the dynamics, and calculate a new positive-sequence voltage profile. From the positive-sequence voltage profile computed by the transient stability program, we can generate zero- and negative-sequence voltages so that the unbalanced set of phase voltages can be calculated in the protection network model. This last step requires injecting zero- and negative-sequence fault currents into the faulted bus in the protection network. The values of the zero- and negative-sequence fault currents depend on the fault type and the positive-sequence fault current. In this section, we state without derivation for each of the standard faults, the following: • • Fault impedance needed to simulate the standard fault types in the positive-sequence domain of transient stability. The negative- and zero-sequence fault currents that need to be injected into the protection network’s faulted node, so that negative- and zero-sequence voltages may be reconstructed at all buses in the protection network model. In the table below, TS indicates the Transient Stability Program. Georgia Tech Protective Relaying Conference 2014 47 Fault Type Fault Impedance to apply in TS Positive-Seq. Voltage Computed by TS ZF V1F In the Protection Network Model Positive-Seq. Fault Negative-Seq. Zero-Seq. Fault Current Fault Current Current Three-phase fault A B Impedance of external fault. C ZF ZF TS simulates dynamics and calculates this voltage. ZF I1F does not need to be computed in the protection network. Directly apply positive-sequence voltages calculated by TS at all buses. I2F = 0 I1F = 0 I2F = I1F I0F = I1F Inject this current into the negativesequence network to calculate postfault negativesequence voltages. Inject this current into the zero-sequence network to calculate postfault zerosequence voltages. Single-line-to-ground fault A Z2 + Z0 + 3ZF V1F B C ZF where Z2 and Z0 are the negative- and zerosequence Thévenin impedances at the fault point. ZF is the external fault impedance. Georgia Tech Protective Relaying Conference 2014 TS simulates dynamics and calculates this voltage. 48 I1F = V1F Z 2 + Z 0 + 3Z F Fault Type Fault Impedance to apply in TS Positive-Seq. Voltage Computed by TS Z2 + ZF V1F In the Protection Network Model Positive-Seq. Fault Negative-Seq. Zero-Seq. Fault Current Fault Current Current Line-to-line fault A B where Z2 is the negativesequence Thévenin impedance at the fault point. ZF is the external fault impedance. C ZF TS simulates dynamics and calculates this voltage. I1F = V1F Z2 + ZF I2F = –I1F I1F = 0 Inject this current into the negativesequence network to calculate postfault negativesequence voltages. Double-line-to-ground fault Z1F = ZF + ( Z2 + ZF ) ⋅ ( Z 0 + ZF + 3Z G ) Z TOTAL A B C ZF ZF ZG where Z2 and Z0 are the negative- and zerosequence Thévenin impedances at the fault point. ZF and ZG as shown in the figure to the left. I2F = V1F TS simulates dynamics and calculates this voltage. ZTOTAL = Z2 + Z0 + 2ZF + 3ZG Georgia Tech Protective Relaying Conference 2014 49 V I1F = 1F Z1F See Z1F definition two columns to the left. −I1F I0F = ( Z 0 + ZF + 3Z G ) Z TOTAL Inject this current into the negativesequence network to calculate postfault negativesequence voltages. −I1F ( Z 2 + ZF ) Z TOTAL Inject this current into the zero-sequence network to calculate postfault zerosequence voltages.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project