Presentation on theme: "Validation and Accreditation of Transient Stability (TS) Results Tom Overbye Fox Family Professor of ECE University of Illinois at Urbana-Champaign Jim."— Presentation transcript:
Validation and Accreditation of Transient Stability (TS) Results Tom Overbye Fox Family Professor of ECE University of Illinois at Urbana-Champaign Jim Gronquist Bonneville Power Administration (BPA)
Overview Goal of project is to perform validation of the transient stability (TS) packages Commonly thought that different TS packages give different results for the same system conditions Software validation is (according to DOD), The process of determining the degree to which a model or simulation and its associated data are an accurate representation of the real world from the perspective of the intended uses of the model. This project is just focused on validating the packages against each other, as opposed to with real world data (which would be a natural follow-on) 2
Project Importance TS looks at the time-domain response of a system following a disturbance (contingency) for several seconds to minutes Integration used here was ¼ cycle Assessing angular and short-term voltage stability System stability is a growing concern, partially because of more wind integration PMUs providing more dynamic observability 3
Background Project grew out of PSERC effort from 2010 to 2011 funded entirely by BPA involving UIUC and WSU to research the validation of the PSLF, PSSE, TSAT, and PowerWorld Simulator TS packages UIUC took the lead with PSSE and PowerWorld, WSU with PowerTech, and BPA with PSLF Most of the PSLF solutions were done by BPA, with just several of the earlier (2010) PSLF runs provided by WSU. Just solution results were provided to UIUC. None of the results presented here involve PSLF solutions done by UIUC. 4
Project Challenges Key challenges include the sheer number of models that need to be validated. WECC case has more than 3300 generators, 8200 loads, more than 70 different dynamic models (in PSLF) Individual dynamic models can have dozens of parameters that can substantially affect the behavior of the models The PSLF and PSSE packages often use different models to represent the same generators/loads PowerWorld Simulator implements both sets Model parameters are sometimes incorrect, with values often automatically corrected 5
Model Example: GGOV1 6
PSERC Project Caveats Project required a deep knowledge of the TS solution process and the ability to solve large systems using the different TS packages Net gain to industry when students graduate, but there is also a loss of university expertise Project focused primarily on PSLF and PowerWorld, and on the WECC system Pointing out problems is not meant to imply they are worse than others, such as the Eastern Interconnect model; it just wasnt studied 7
Data Confidentiality Many of the studies done for this project utilized confidential information that was provided to UIUC and WSU under NDAs Data presented in these slides uses WECC results. Public release can be made at the discretion of WECC 8
The Starting Point 9 In May 2010, as the project was just beginning, we were able to do comparisons for a full WECC case (provided by BPA) between PSLF, TSAT and PowerWorld Simulator. The plot on the right shows the variation in a bus frequency following the double Palo Verde generator trip contingency for the three packages We moved forward with both a top-down (full system) and bottom-up (individual model) approach
The Top-Down Approach 10 Within a week or two we were able to determine that a bug in how the frequency deviations for the induction motor loads were being handled by PowerWorld was causing some of the frequency variation. The new run is shown on the right. The oscillation in the PowerWorld frequencies were tracked down to a model error. Model error was associated with a generator line drop compensation; fixed by auto-correction
Bottom Up Approach 11 The bottom-up approach consists of creating two bus equivalents for the most common WECC generator models, and then running them in the PowerWorld and PSSE packages. This has resulted in several PowerWorld changes with the new code often giving quite close matches Graph shows a comparison of the field voltage at generator for a fault on the two bus equivalent system between PSSE, and two versions of PowerWorld Simulator
The Bottom-Up Approach While there are lots of models, project initially focused on the most widely used models. WECC case has a total of 17,709 models in 77 model types. But the 20 most common model types contain 15,949 (90%) of these models. These are the key focus areas for the bottom-up analysis. 12
Post-PERC Work Since the PSERC project finished in August 2011, UIUC and BPA have continued the validation work focusing on comparisons between PowerWorld and PSLF The bottom line is the two packages now agree quite closely for the studied scenarios, certainly within the bounds of variability due to input data errors. This close match allows for detailed comparisons! 13
Power Flow Solutions The power flow solution provides the starting point for the TS. PSLF and PowerWorld have very similar solutions. Key difference is the distribution of reactive power at multi-generator buses. The net bus injection is the same, resulting in the same power flow solution. But the algorithms to distribute reactive power to the generators differ, resulting in slightly different initial transient stability generator field voltages 14
Initial TS Variable Values Power flow operating point is back-solved to determine the initial TS states and related variables. Ideally this should result in an equilibrium point, so without a disturbance none of the variables should change. But some may if there is an initial limit violation. Largest field voltage change in two seconds was 0.34 pu in both packages at gen 42177 due to an initial Vr limit violation. 15
Efd at Gen 42117 16 PowerWorld includes an option to automatically modify the limits to provide a flat solution
Initialization Pmech: Match is Good The initial values for the governor mechanical power input and Efd were compared for all generators. Very little difference was observed Largest differences for mechanical power were 4 MWs at the wind turbines (wndtrb), which was associated with how the rotor losses are modeled 17
Initialization Efd: Match is Good For the Efd values, the average error was just 0.003761 per unit (typical Efd values of 1 to 3) Just 200 units (out of a total of 2322) had initial differences greater than 0.01 per unit. In a small number of comparisons between PSLF, PowerWorld and PSSE; PowerWorld and PSSE matched closely Example gen #QF at 24815 PW had Efd=1.8409, PSSE 1.8404 and PSLF 1.8691 (Qg=-72.4 Mvar) Example gen #G1 at 24815 PW had Efd= 1.0459, PSSE had 1.0471 and PSLF 1.013.(Qg=-2.9Mvar) 18
Simulation Comparison Methodology Study involved the Double Palo Verde contingency The case comparison methodology involved looking at time results for selected values, and differences between the min/max values for different time periods for all key variables (Efd, Pmech, stabilizer output) Case is quite stressed so small variations can significantly impact the results 19
Validation Issues: IEEEG1 20 With the IEEEG1 governor, PowerWorld auto-corrects high K1 to K8 values. Gens 56503, 57503 and 58503 have K1=20.4, which causes very high Pmech outputs in the PSLF results
56503 Output PSLF vs PowerWorld 21 Graph compares the Pmech between PSLF and PowerWorld with K1=20.4 or auto-corrected. Note the turbine rating is 71.2 MVA. Behavior is similar at 57503 and 58503.
Unexpected PSLF Behavior 22 Graph compares the Pmech between PSLF and PowerWorld for G1 at 40559 (Ice Harbor). Unit is modeled with an IEEEG3 Governor. Turbine rating is 103 MVA. PowerWorld cannot compensate for this difference. To some extent it offsets the K1 issue.
40687 (Malin) Freq. Comparison 23 Note whether the K1 issue is corrected has a large impact on the resultant frequency. The two PowerWorld results mostly bracket the PSLF results. The 40559 issue also has a significant impact.
40687 (Malin) Volt Comparison 24 Again the two PowerWorld solutions mostly bound the PSLF results
24801 (Devers) Freq. Comparison 25 Results are similar for a bus in Southern California. Again the PowerWorld results (with the K1 issue either high or fixed) bracket the PSLF results.
24801 Voltage Comparison 26
Potential EXST1 Issue In comparing the initial field response for the four units at 36405/6 (Moss Landing), PSLF may be ignoring the feedback term for the EXST1 exciter when Ta = 0 27
Gen 36406H Efd for Varying Ta 28 Note changing Ta slightly has little impact on the PowerWorld results but gives a significantly fast PSLF response. The rate feedback should limit the change in Efd.
Key Differences: Modeler Intent Some key differences in the packages arise in how they handle modeler intent in cases in which the input data is either ambigous or wrong. Example: At gens 56503, 57503 and 58503 the IEEEG1 governor has K1=20.2, which means the maximum output of the units is 20 times the normal value PowerWorld catches this as an error, allowing the value to be autocorrected 29
Key Differences: Modeler Intent Other examples include Handling exciter saturation when SE1=0 and SE2 > 0. PowerWorld fits a quadratic to these two points Issues when GV/PGV data is not increasing. PowerWorld raises a validation error. Reading in models with blank IDs. PowerWorld had defaulted if there was one gen at the bus WSCCST stabilizer behavior when vcutoff < 0. PowerWorld ignores the parameter, PSLF sets stabilizer response to zero (33141/2/3) 30
Possible Issue with PSLF EXAC1 31 Graphs compares the field voltage for the generator at 15914. This generator uses an EXAC1 exciter. The limit on Ve should prevent a negative field voltage. The PSLF results show negative values, with PowerWorld clamping the value at zero.
Conclusion PSLF and PowerWorld match quite closely and both can be used for WECC studies. Using multiple packages can help to track down issues either within the packages or within the data. Hence there is a net gain. Individual package innovation, such as improved error checking or options, can result in better results for all. 32
Moving Forward The PSERC project has concluded, but this work is too important to stop. We are continuing to move forward with funding provided by the Illinois Center for a Smarter Electric Grid (ICSEG) Industrial participants are invited to join us BPA has already expressed an interest to continue participating in this effort Email email@example.com if firstname.lastname@example.org 33