Aranya Chakrabortty

This DOE proposal will develop new machine-learning based algorithms for detection, localization and mitigation of different types of cyber-attacks in electric power systems.

Microgrids (MG) deliver highly resilient power supply to local loads in the event of a power outage, while improving distribution system reliability by reducing the load on the system under stress conditions. Networked microgrids that coordinate with each other and the distribution system operator increase reliability and resilience further by improving the diversity of the generation assets and loads. Microgrids are also widely considered as an essential technology to address the challenges of climate change.

The proposed project will apply risk segmentation, adaptive credit scoring and network-based portfolio analysis techniques from financial engineering and risk management for risk analytics of power systems at both asset and system levels. At the asset level (Thrust 1), the project will introduce risk segmentation of an asset??????????????????s throughput by applying tranching similar to collateralized debt obligations. The risk-free to most risky tranches will be assessed for their risk profile in terms of risk scores taking into account the variability of the renewable resource (wind or solar), presence of storage units or services that they may be equipped/associated with, and the asset??????????????????s locational specification. This risk scoring will be designed to be adaptive based on system level (Thrust 2) feedback at different contractual time-scales, starting from sub-seconds to tens of minutes, to determine asset suitability as an energy, regulation, spin, non-spin, or replacement reserve. Novel copula-based probabilistic risk models will be developed for the joint correlation structures between different contract tranches of assets for asset and system level risk assessment.

Dr. Chakrabortty will be serving as a program director in the ECCS division of NSF under an IPA agreement.

In the current state-of-the-art, data-driven machine learning based control of large-scale complex multi-agent networks systems is largely bottlenecked by the curse of dimensionality. Even the simplest linear quadratic regulator design demands cubic numerical complexity in real-time. The problem becomes even more complex when the network model is unknown, due to which an additional learning time needs to be accommodated. In this project, we seek to take a different approach, and investigate a hierarchical reinforcement learning based control scheme for extreme-scale multi-agent swarm networks. Here, control actions will be taken based on low-rank information from data instead of models. The approach will be to decompose a control objective into multiple smaller hierarchies, such as group-level microscopic controls that can be learned using dense but local data only, and a broad system-level macroscopic control that steers the swarm in its desired direction but using only high-level sparse data. Each hierarchy will have its own learning loop with local and global reward functions. The control goal of the network will be decomposed accordingly into local (microscopic) and global (macroscopic) reward functions. Local controllers will be designed via private group learning, and the global controllers via model reduction and averaging. Sparse controller structures will be imposed on top of the local controllers to reduce their communication complexity. Deep learning algorithms based on historical events will be used to train recurrent neural networks so that they can rapidly predict these sparsely structured projections following any disturbance event in the network. One driving example, which will be used as a benchmark for simulations, is joint target-tracking and interception using swarms of ground vehicles and air vehicles. The ground vehicles will be divided into teams that are autonomously formed through learning. Agents in each team will learn a local reinforcement learning control that can track detailed microscopic dynamics of a target while respecting individual formation constraints. The air vehicles, on the other hand, will serve as higher-level coordinators that learn a global reinforcement learning control using low-resolution and low-rank data to provide a macroscopic view of the target motion in terms of the movement of centroids.

Cybersecurity for next generation ECUs

This project will develop different cyber attack detection and mitigation scenarios for smart grid technology.

The objective of this research will be on improving dynamic performance of New York State (NYS) power grid using supplementary Wide-Area Damping Control (WADC) with shunt-connected FACTS devices and Wind farms as the control actuators. The NY state is moving towards more renewable generation. The state of New York already undertaken a comprehensive energy strategy, known as Reforming Energy Vision (REV), for building a clean, more resilient, and affordable energy system. One of the major goals of REV is to reach 50% renewable generation by the year 2030. Bulk wind power integration will be a major contributor in reaching this goal. Different research conducted by New York Power Authority (NYPA) and FREEDM Systems Center have shown that high wind penetration can impact the grid oscillation properties adversely and can induce poorly damped inter-area oscillatory modes. This could result in destabilization of the power grid. In wake of this scenario, NYPA is considering to implement Wide Area Controllers within their territory. NYPA has already installed multiple Phasor Measurement Units (PMUs) all across the grid. Currently, data stream from PMUs is being used to provide Wide-Area Situational Awareness (WASA) for the system operators; however, this data is not being used for any decision making and control-based remedial action for grid operation. Recently more emphasis is given to use this Wide-Area Measurement Systems (WAMS) in order to design oscillation damping controllers. This research will look into explore the feasibility, constraints and possible solutions for the real-life implementation of wide-area damping controllers through FACTS and Wind farms in New York State Grid.

This project will develop a distributed hierarchical control architecture for next-generation power transmission networks by using ideas from real-time learning and decision-making. The vision is to consider the future grid with numerous inverter-interfaced generation, FACTS devices, distributed energy resources such as wind, solar, and storage, as well as convention synchronous generators with primary/secondary/tertiary controls, all of which collectively generate thousands of ????????????????control points??????????????? over wide geographical expanses of the grid. The goal is to design an online hierarchical grid control architecture, starting from the substation with plant controllers, extending to local areas with local-area controllers and going up to a system level for wide-area control and optimal positioning for security constrained real and reactive-voltage setpoint dispatch.

This 2-year NSF project is a collaboration between MIT (main lead), NC State University, and University of Notre Dame. The main purpose of the project is to develop a suite of novel numerical computational algorithms by which very large complicated mathematical models of large power system networks can be constructed and solved in real-time, or even faster than real-time. The study will involve complex network models of power grids with high penetration of wind and solar power, and their associated stochasticity, and make use of new ideas from algebraic topology theory to develop solutions of those models. Validation will be done using the RTDS-WAMS testbed at FREEDM systems center.