TRAST, Preventative Measures for Real-World Reliability

Battelle Number: 32107 | N/A

Technology Overview

Remedial action schemes (RASs), also known as special protection schemes, are used throughout the electric utility transmission systems as a non-wire method of responding to abnormal or predetermined system conditions to maintain overall system reliability. PNNL’s patented Transformative Remedial Action Scheme (TRAST) software enhances and automates the process of identifying and evaluating RAS scenarios and settings used by electric utilities to plan for contingencies and emergencies.

Machine Learning, Data-Driven Analytics

TRAST uses a novel approach to automatically generate use cases. It combines advanced statistical data analysis tools and machine learning algorithms to analyze, validate, and help create RAS plans. PNNL’s parallel computing platform and the Microsoft cloud environment are used for steady state and dynamic simulations under massive contingencies and operating conditions. The result is more realistic settings of RAS systems in U.S. Western Interconnection, ultimately making the grid more reliable and resilient.

Video: Pacific Northwest National Laboratory

Due to a lack of automated tools and computation power, RAS settings are traditionally determined by time-consuming offline studies that don’t allow for adaptive learning. Limitations like these result in overly conservative RAS settings, causing unnecessary flow curtailment, generation tripping, or misoperation of RAS. These actions can affect the revenue of generator owners and the economic operation of the entire network. PNNL’s technology bridges this gap. TRAST uses machine learning to enhance the design, study, and evaluation of RAS for the power system, thereby improving grid reliability.  

Significantly Simplify the RAS Design and Study Process

TRAST, a prototype research tool, originated from data analysis for utilities and evolved with domain knowledge from power engineers. The technology uses advanced statistical data analysis tools to automatically generate RAS use cases. Machine learning algorithms then analyze, validate, and aid in the creation of RAS plans with preventive emergency controls.

Automatic and semi-automatic functionalities integrated in TRAST significantly simplify and shorten the RAS design and study process. In addition, the TRAST evaluation methodology enables continuous improvement and validation. With the development of faster computational techniques at PNNL, including parallel dynamic simulation and massive contingency analysis powered by high-performance computing techniques, it becomes feasible to validate the preventive controls and adjust parameter settings in near real-time to improve security and efficiency. With adaptive RAS settings, the power grid can be operated less conservatively and more reliably.

Quickly Perform Full-Scale Studies

Today’s commercial tools are not fast enough to perform a full-scale study to calculate RAS parameters and validate the control performance in a preventive way. Even so, the primary barrier is not computing tools but convincing utilities and regulators to embrace the advantage that high-performance computing techniques allow, in this case, the convergence of data and models. TRAST bridges this gap.


Grid Operators and Utility Planners Would Benefit from Using TRAST

TRAST uses machine learning to enhance the design, study, and evaluation of RAS for the power system, thereby improving grid reliability.
TRAST uses machine learning to enhance the design, study, and evaluation of RAS for the power system, thereby improving grid reliability.

RAS represents one of the core resilience components of the electric power system and is critical in maintaining overall system reliability. More importantly, regulators require utility companies to report the commission, update, and revision of RASs as a valid strategy for rapid response to system emergencies in advance of field implementation.

In partnership with PacifiCorp and Idaho Power, PNNL researchers have identified the technology gaps in current RAS modeling practices for planning and operation and are using machine learning to better understand grid operation patterns and improve RAS designs. Building from real-world system data, TRAST algorithms will determine the arming levels of RAS to develop use cases that demonstrate the benefits of adaptive RAS parameter settings.

The team is working closely with the power industry, including PacifiCorp, Idaho Power, former Peak Reliability, and the Western Electricity Coordinating Council. Both grid operators and utility planning engineers will benefit from the technology, and a better RAS modeling process will increase interconnection-level system reliability and resilience.

Technology Features

Key features of the TRAST tool include the following:

  • Advanced statistical data analysis
  • Automated power flow case generation using optimization algorithms
  • Customized dynamic simulations in a high-performance computing cloud platform
  • Machine-learning-based RAS coefficient prediction
  • Reliable RAS validation strategy in multiple commercial platforms.

State of Development

In partnership with PacifiCorp and Idaho Power, PNNL researchers identified technology gaps in current RAS modeling practices to build TRAST. The tool is protected with patents and copyrights, which are available for licensing in all fields of use.


Available for licensing in all fields


power grid, remedial action schemes, power grid emergency planning, machine learning, utilities, power operators


Electricity Infrastructure

Market Sectors

Energy Infrastructure

Research topics