Towards the Automation of Data-Driven Remedial Action Scheme Design
Abstract
The emerging advancements in the electric power industry are constantly redefining the scope of power system planning and operations, where engineering decisions made are reliant on the most up-to-date knowledge of the grid. Managing and operating a conventional electric grid is a task in which engineers and operators have a century of collected experience. However, the understanding we obtained from previous experiences might not be sufficient for the grid of the future.
Traditional remedial action scheme (RAS) design takes a holistic approach that requires years of expertise in a specific power system. Building on the industry’s current practice, this dissertation develops a power system tool called Auto-RAS that provides a systematic approach to create remedial action schemes in a robust, effective, and automated manner. Leveraging data-driven techniques, responses and control practices of the
current power system are contextualized as statistical characteristics and mathematical expressions to guide the design of remedial action schemes.
To maintain a similar level of size and modeling complexity as the real power system while
still be share the research result publicly and freely, synthetic electric grid network models are used as test cases for the development and testing of Auto-RAS. In this dissertation, a chronological power system operation simulation framework is developed to provide large amount
of scenarios representing a wide spectrum of system operating conditions as input data for the Auto-RAS process.
The determination of Auto-RAS condition logic starts with operational scenario analysis, where the need of remedial action schemes are identified as a list of severe violation system elements. A two-stage linear SVM algorithm is implemented to selected features that can best represent the operational scenarios, and learns a hyperplane that can optimally divide the scenarios with and without risks of severe violations. The selected scenario features, along with the learned hyperplane, and list of violation-causing contingencies are then leveraged as Auto-RAS condition logic.
A sensitivity-based methodology is developed in this dissertation to create the corrective actions that can be deployed adaptively for remedial action scheme. Leveraging network connectivity analysis, a subset of power system elements that can be controlled as part of the corrective action scheme is selected. Sensitivity analysis such as line outage distribution factor and transmission loading relief are used to quickly determine the most effective controllable elements and the corresponding corrective actions to address specific operational violations.
To evaluate the performance of remedial action schemes developed by the Auto-RAS framework, a checklist is developed to follow typical industrial system planning standard, and to ensure that the designed remedial action schemes can operate to perform their intended functionalities, and do not introduce unintentional or unacceptable reliability risks to the bulk electric power system.
Citation
Li, Hanyue (2021). Towards the Automation of Data-Driven Remedial Action Scheme Design. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /195244.