Introduction
If you want to create a large-scale simulation of an electric grid that closely matches reality, you generally must calibrate the individual representative models of that grid. This process, power plant model validation (PPMV), can be performed with both offline step tests and online performance monitoring of grid events. Additionally, due to the increasing percentage of wind, solar, and energy storage in the modern grid, accurate models of these renewable energy systems, in addition to traditional generation, are becoming increasingly important. The main objectives of PPMV are as follows:
- Find potential errors and/or fixes in the model
- Understand the sensitivity of parameters to potential model improvements
This task can be challenging, particularly when required by technical regulations such as NERC Standard MOD-026 or MOD-027. This paper covers a workflow for PPMV using MATLAB® and Simulink®, with emphasis on offline step test and online performance monitoring of grid events using phasor measurement unit (PMU) data for both renewable energy and traditional generation. It explores workflows that include both manual adjustments and automated techniques. A solar plant case study demonstrates how to do several things:
- Replay measured data through simulations
- Gain insight into response discrepancies through field data replay
- Use engineering judgement and automated parameter sensitivity to assess and rank the influence of system parameters on system response
- Fine-tune system response using both manual adjustments and automated parameter estimation
Additional templates for traditional generation tests are discussed as well:
- Zero-power factor load rejection test
- Open circuit voltage step tests
- Online step tests
Replay Measured Data Through Your Simulations
Calibration of standardized power system models with real data requires overlaying and comparing both simulation response and field measurement data. One such method of comparing responses is to replay measured field data through simulation models and observing responses. With the influx of renewable energy on the grid, standardized models of plant performance have been developed, similar to the IEEE standard models for traditional generation. Figure 1 (below) depicts how a utility-scale solar plant can be represented at a high level.
Figure 1. Standard model for utility-scale solar plant accommodating multiple types of vendors.
The core attribute of PPMV is to replay measured data through a simulation model. For offline step tests, data replay involves exciting the plant model through control signal measurements. For online performance monitoring, data replay involves exciting the plant model through physical measurements of voltage or current at the grid point-of-connection. It is common to consider voltage (V), frequency (F), active-power (P), and reactive-power (Q) as the four measurements necessary to perform PPMV. For traditional generation, field voltage or field current can also be used if recorded during testing procedures.
In Figure 2 below, a model of a solar facility (using REEC_A, REPC_A and REGC_A) has a VF replay block that allows for replay of field data to confirm plant performance.
Figure 2. Replay of voltage and frequency field data to match active and reactive power.
Gain Insight into Response Discrepancies Through Field Data Replay
The first stage in PPMV is to gain insights into response discrepancies through manual adjustment of parameters. You may have a list of “go-to” parameters to adjust based on previous experience and can also gain additional insights through observing the attributes of a response discrepancy, which can point to certain parameters requiring adjustment. Figure 3 shows the results of VF replay for the renewable energy plant example. As the simulated PQ response does not match the measured PQ response, the model parameters are inaccurate and require parameter adjustment. Since the P and Q responses have dynamic response differences, the voltage feedback loop parameters are likely the culprit of the difference in response and could be manually adjusted.
Figure 3. VF replay results showing the solar plant model parameters need adjusting to match field performance.
Use Engineering Judgement and Automated Parameter Sensitivity to Assess and Rank the Influence of System Parameters on System Response
Manual parameter adjustment can be made, but solar plant models contain over 50 parameters that could be adjusted. To find the parameters to focus on for this voltage step test, we can apply automated parameter sensitivity to assess and rank the influence of additional system parameters on system response. Sensitivity analysis, as shown in Figure 4, will run many simulations with small perturbations of the parameters to identify which parameters improve the results the most. This method helps to identify the most important parameters to focus on.
Figure 4. Sensitivity Analysis interface that connects to the VF replay solar plant model.
The results of sensitivity analysis can be found in Figure 5. This plot shows the four parameters that were tested in this case and shows that the repc_Ki gain parameter will have the largest impact on improving the voltage step response performance.
Figure 5. Sensitivity analysis results identifying the most important controller parameters to adjust.
Fine-Tune Your System Response Using Both Manual Adjustments and Automated Parameter Estimation
Following manual adjustment and automated parameter sensitivity, you can apply automated parameter estimation to fine-tune the response. Parameter value ranges can be constrained in the automated parameter estimation task and multiple tests can be added to ensure identified parameters are accurate for a wide range of operation of the solar plant. Parameter estimation, as shown in Figure 6, uses optimization methods to minimize the difference between the simulation and field data by automatically adjusting the plant parameters.
Figure 6. Parameter estimation of solar plant parameters to match P and Q field data.
It should be noted that the PPMV task should not end after automated fine-tuning. You should assess the result and determine whether further manual adjustments can be made. For example, you can set parameters that do not change significantly back to their original values and compare responses. If the result is comparable, it may be more appropriate to stick with the minimum number of parameter changes.
Traditional Generation Replay Examples
In addition to renewable energy model validation, field data replay can be performed for typical traditional generation facilities. While PMU data can be used, replay can be configured for specific offline tests of plant equipment. The advantage of performing multiple independent tests and validation studies is that each component can be tested in isolation, which minimizes the number of parameters to adjust. With the following configuration examples, both sensitivity analysis and parameter estimation can be performed as demonstrated above with the solar facility example.
For a zero-power factor load rejection test, the generator can be simulated in isolation using PQ replay (see Figure 7). In addition to replaying P and Q, we can replay other measurements, such as the frequency and measured field voltage (Efd). In this case, the data match can focus on the generator terminal voltage.
Figure 7. Example of a zero-power factor load rejection test configuration.
With a validated generator model, additional components can be included with additional offline tests, such as voltage step tests of the excitation system. In Figure 8, an ST1A excitation system has been included. The voltage reference is replayed, and the terminal voltage of the generator can be confirmed to match. Since the model is an offline test, the grid replay of active and reactive power is zero.
Figure 8. Example of an offline voltage step test configuration.
Finally, additional components can be added, such as a power system stabilizer. In Figure 9, a PSS2C has been added to the model and PQ replay is being used to validate all the components using grid-connected data. While not directly added in these examples, a governor model could also be included for frequency replay. Once a template for a plant test is created, the same template can be used for multiple power plants where similar testing is conducted. In the event of different generator or excitation systems, these blocks can simply be swapped for different plant configurations.
Figure 9. Example of an online step test template using PQ replay of field data.
Summary
In this paper, we explored PPMV as applied to online performance monitoring of grid events using PMU data, using a workflow that included both manual adjustments and automated techniques. A utility-scale solar plant case study demonstrated the following workflow steps:
- Replay measured data through simulations
- Gain insight into response discrepancies through field data replay
- Use engineering judgement and automated parameter sensitivity to assess and rank the influence of system parameters on system response
- Fine-tune system response using both manual adjustments and automated parameter estimation
With MATLAB and Simulink, you can efficiently perform power plant model validation with automated techniques. The workflow provides insight and flexibility when addressing technical regulations such as NERC Standard MOD-26.
