M. Kezunovic (P.I.) S. S. Luo D. Ristanovic Texas A&M University Fault Location in Transmission Networks Using Modeling, Simulation and Limited Field Recorded Data M. Kezunovic (P.I.) S. S. Luo D. Ristanovic Texas A&M University PSerc Review Meeting College station, Nov. 7, 2002 PSERC

Overview Objectives Testing procedure data requirements and test results Potential problems and improvement Design and User documentation Conclusions Future Research This presentation includes three parts: Part I introduces the summary report Part II introduces the technical report Last section is conclusion. Mainly introduce the status of the project

Objectives Defining procedure to be used for testing Defining requirements regarding input data Testing using fault data collected from Center Point Energy Testing using fault data collected from other utility Analyzing and evaluating the performance and proposing potential improvements Improving the software and developing user interface Developing the design and user documentation for facilitating future upgrades and practical use of the software

Testing procedure Using static system model Obtaining the fault data file from utility and then converting into COMTRADE format Producing the input data file based on the available DFR files Checking the corresponding substation interpretation files based on the DFR configuration and system file Running the software and obtaining the estimated location Comparing the estimated result to the actual or calculated result When using updated system model Extracting parameters from the model and producing the topology data file before executing the above steps System file refers to the one consisting of the topology, load and generation data of the system and short circuit parameters given in the specific format, such as PSS/E format, hlppssemodel.sav

Data requirement When using static system model Fault case data fault data file in COMTRADE format (DFR raw data is need to be converted) fault report (optional) actual fault location (optional) System data Load flow and short-circuit model including topology Interpretation file for each monitored substation

Data requirement, Cont. Input data file generated by user based on available DFR files Necessary fault information including the fault type, affected fault circuit Options how to produce a list of possible faulted branch candidates Algorithm parameter file Including iteration number, crossover and mutation possibilities, population number… When the more accurate model is required Additional real power flow data is needed

Testing 15 fault cases obtained from Reliant Energy H&LP were tested Power system simulator PSS/E is utilized PSS/E models in versions 26, 27, 28 are tested Sensitivity of results under different options is analyzed Static model and tuned model are used

Testing Using fault data collected from TVA or other utility to test fault location The item was changed because data is not obtained Instead, different power system simulation software was used CAPE is a new selection PSS/E system model is converted into CAPE database Short circuit results obtained from the PSS/E and from CAPE are compared The fault location software is customized for CAPE Test result is not available because current version of CAPE is not perfect

Test results (1) Case # Number of DFR triggered Actual or calculated fault location Estimated fault location Error 1 2 41111-41700 0.40 mile 0.2 mile 0.2 48402-40590 3.32 miles 0.3 3 41300-48386 3.49 miles 3.15 miles 4 40570-41405 2.50 miles 2.47 miles 0.0 5 46262-48306 2.0 miles 1.18 miles 0.8 6 46570-48219 1 mile 0.1 mile 0.9 7 2.8 miles 46512-4830 6.1 miles 3.3 8 3 miles 5.7 miles 2.7

Test results (2) Case # Number of DFR triggered Actual or calculated fault location Estimated fault location Error 9 1 5915-9073 66.0 miles 66.9 miles 1.0 10 45840-40180 3.8 miles 40180-40620 0.4 mile 0.9 11 3 40620-48295 2.36 miles 2.13 miles 0.2 12 2 46020-3390 7.77 miles 6.54 miles 1.2 13 46020-3391 6.2 miles 1.6 14 4.77 miles 3.0 15 7.09 miles 0.7

Problems and improvements Some factors affected estimated fault location accuracy: Fault cycle Faulted branch candidates Phasor calculation Model GA result Statement work mainly presents all activities for the project. Defining the procedure to be used for testing the fault location software----one of two primary goals of the project is to evaluate the performance of the software.To get a comprehensive assessment, a more systematic testing is needed. Defining the requirements regarding the input data needed for testing----It refers to the requirements for various power system Testing the fault location software using fault data collected from RE HL&P---- Testing the fault location software using different power system simulation software----

Problems and improvements Fault cycle - Problems For each triggered DFR, correct cycle to calculate the during-fault phasor should be used. For several triggered DFRs, the the same cycle to calculate the during-fault phasor should be use - Improvements The criteria of determining fault cycle is improved Additional measurements are taken to avoid using different fault cycles In the user interface part, a new feature is added for user to specify the exact fault cycle This is a schedule of the project. Note that the task #4, the original task is to use the fault data obtained from other utilities. Because of lack of the fault data from other utilities, the schedule is changed slightly. The task is to use new system simulation package instead of the PSS/E to test the fault location software

Problems and improvements Branch Candidates - Problems The produced list of candidates must include the faulty branch, which creates a large number of choices - Improvements Additional options are added for user to choose the method of producing candidates user can check the detail list of possible faulted branch candidates and edit it before the fault location software runs

Problems and improvements Phasor calculation - Problems • Waveform includes DC offset and high frequency components, which affects the accuracy Improvements Using improved Fourier algorithm for obtaining the during-fault recorded phasor Filter the noise contained in recorded waveforms The section is to introduce the technical report. There are six parts: Introduction. Background mainly presents the former activities before the last review meeting. These activities include 1.testing the initial version of the fault location software through using the fault data obtained from the RE HL&P, 2.analyzing the performance of the fault location algorithm and software 3. Implementing the identified improvements. These improvements include correction of the phase sequence, correct calculation the fault inception cycles, correcting the design problems, algorithm improvements and model improvement. For the algorithm, signal processing is introduced to remove the noise found from the recorded waveforms, improved Fourier algorithm is introduced to eliminate the DC component. For GA algorithm, some measures are taken to overcome the small population problem, convergence problem, search space problem and so on. For model problem, a tuning strategy is applied to get a better updated model. To facility the usage of the fault location software, user interface was developed. 4. Several versions of interim testing reports were submitted. Activities refers to the activities since last review meeting Results refers to the results obtained from the new version of the fault location software which utilizes the new strategy of tuning system model Next step-activities for coming months Conclusion

Problems and improvements Model - Problems The static PSS/E model provided by utility may not reflect the real system operation condition when a fault occurs Tuning generator and load power as well as tuning the system topology is required - Improvements Using different version of PSS/E model with different topology and parameters Tuning static parameters. Two situations are considered: No additional real data is available. The concept of pre-fault phasor matching is introduced. Some cases show that tuning is effective Additional real data is available, generator and load power scaling is utilized Activities Developing the graphical user interface In order to facilitate the usage of the fault location software and further upgrades , a limited user interface was developed. Much benefits Tuning the static system model Converting PSS/E data into CAPE database

Problems and improvements Genetic algorithm convergence - Problems Fixed iteration number may not always approach the final solution For different runs, GA result may vary within a specific range - Improvements • Using fitness scaling to solve the small population • Using multi-point crossover to increase the search space • Using new replacement of “weak parent” to make GA more robust • Studying behavior of the fitness value add a criteria using the average fitness • Adding a feature to give an exact result after using GA limit search range in the user interface part

Design and user documentation Limited development of user interface for practical use Fault location software design documentation and user’s guide are produced for software upgrade

Conclusion(1) The test results show that the scheme of matching waveforms to locate a fault is feasible Multiple triggered DFRs are helpful for improving location accuracy It is suggested to use all the recorded currents and voltages for matching It is suggested to use the same fault cycle to calculate during-fault phasors for each DFR

Conclusion(2) Tuning system model and making it fit the operation condition when the fault occurs helps producing more accurate results, especially when additional real data is available. It is suggested that the fault resistance is set within a reasonable range, especially in 345KV system Producing a right list of faulted branch candidates before running fault location software is very helpful

Future research How to obtain and incorporate more accurate model data How to make the user’s knowledge more useful How to incorporate an iterative approach between running the program and having the user look at the results and make some practical choices How to interface the program to variety of short circuit programs How to obtain more data for further evaluation of the performance

Thank you!