PNNL

A paper describing an open-source tool suite used to validate power plant models and calibrate inaccurate parameters was selected as one of the best conference papers submitted to the 2017 Institute of Electrical and Electronics Engineers’ Power and Energy Society General Meeting.

PNNL’s Ruisheng Diao, one of the authors, presented the paper, “An Innovative Software Tool Suite for Power Plant Model Validation and Parameter Calibration using PMU Measurements,” at the meeting on July 17.

Validation—at a Cost

Our nation’s power plants keep the lights shining, the heat and cool air flowing, and appliances performing domestic duties. But the production and distribution of power to our homes would not occur reliably without the stability of power plant infrastructure, such as generators.

Computer models for power plant stability assessment typically consist of equations representing the dynamic nature of numerous plant components. These models provide insight needed for planning and operations for the power grid. For power plant generators with capacities of greater than 10 megavolt amps, the North American Electric Reliability Corporation requires all models to be validated—or checked for accuracy—every five years. However, traditional validation methods, including staged testing, have been expensive ventures for power plant owners. Each test requires them to take their generators offline at a cost between $15,000 and $35,000 per generator.

A Three-Pronged Approach

In search of a lower-cost approach, PNNL and the Bonneville Power Administration developed and tested an innovative, open-source tool that performs power plant model validation, parameter calibration, and model verification.

The tool leverages the ever-growing deployment of phasor measurement units—or PMUs—by power plant and system operators. PMUs remotely measure electrical waves on multiple points of an electricity grid and provide operators with situational awareness.

The tool works like this:

• When an anomaly is detected in PMU measurements, a dynamic simulation is launched to check model performance against the actual measurements.
• If a deficiency in the model is identified, the calibration module is used to find and correct the inaccurate parameter values in the model. This involves three steps: (1) sanity check, which rules out obvious parameter issues; (2) trajectory sensitivity analysis, which identifies typical parameter errors from mismatching patterns; and (3) automatic parameter calibration, which corrects the inaccuracy.
• After the calibrated parameters are obtained and corrected, more model validation studies are launched using different event information to verify that the parameter set is correct and robust.

The tool successfully validated and calibrated parameters for a real hydro power plant using 12 different “events.” This testing proved to the team that the tool is easily used to validate models. The team also found that the trajectory sensitivity analysis is effective in suggesting inaccurate parameters, and that the automatic parameter calibration method was robust in correcting the inaccuracies.

The paper was developed in collaboration with General Electric, who is adopting these methods in their commercial tool, and Peak Reliability, a potential user who will perform testing and validation of GE’s tool. The research was sponsored by the Bonneville Power Administration through its Technology Innovation Program and the Advanced Grid Modeling Program under DOE’s Office of Electricity Delivery and Energy Reliability.