Resilient Networks

Permanent URI for this collectionhttps://hdl.handle.net/10125/107475

Browse

Now showing 1 - 9 of 9

Detecting and Mitigating Data Integrity Attacks on Distributed Algorithms for Optimal Power Flow using Machine Learning
(2024-01-03) Harris, Rachel; Molzahn, Daniel
Using distributed algorithms, multiple computing agents can coordinate their operations by jointly solving optimal power flow problems. However, cyberattacks on the data communicated among agents may maliciously alter the behavior of a distributed algorithm. To improve cybersecurity, this paper proposes a machine learning method for detecting and mitigating data integrity attacks on distributed algorithms for solving optimal power flow problems. In an offline stage with trustworthy data, agents train and share machine learning models of their local subproblems. During online execution, each agent uses the trained models from neighboring agents to detect cyberattacks using a reputation system and then mitigate their impacts. Numerical results show that this method reliably, accurately, and quickly detects data integrity attacks and effectively mitigates their impacts to achieve near-feasible and near-optimal operating points.
Sensitivity Analysis of Machine Learning Algorithms for Outage Risk Prediction
(2024-01-03) Baembitov, Rashid; Kezunovic, Mladen; Saranovic, Daniel; Obradovic, Zoran
Severe weather conditions are known for causing forced outages in the electric distribution grid. Recent research efforts were aimed at predicting outages using weather and historical outage data. This paper studies the sensitivity of different Machine Learning (ML) algorithms to the inclusion of weather parameters from adjacent geographic areas and data availability. We analyzed the ability of different ML algorithms to predict electric grid outage State of Risk (SoR). The selected algorithms are trained and tested on actual utility company data. The findings indicate that a bigger size of the training dataset improves the performance of all models, which is measured by the Receiver Operating Curve, Average Precision, and F1 Score. Conducted experiments suggest that at least two years of training data is required to achieve satisfactory performance. Also, we investigate a statistical significance in models’ performance with the inclusion of weather in adjacent geographic areas.
GPU-Accelerated Verification of Machine Learning Models for Power Systems
(2024-01-03) Chevalier, Samuel; Murzakhanov, Ilgiz; Chatzivasileiadis, Spyros
Computational tools for rigorously verifying the performance of large-scale machine learning (ML) models have progressed significantly in recent years. The most successful solvers employ highly specialized, GPU-accelerated branch and bound routines. Such tools are crucial for the successful deployment of machine learning applications in safety-critical systems, such as power systems. Despite their successes, however, barriers prevent out-of-the-box application of these routines to power system problems. This paper addresses this issue in two key ways. First, for the first time to our knowledge, we enable the simultaneous verification of multiple verification problems (e.g., checking for the violation of all line flow constraints simultaneously and not by solving individual verification problems). For that, we introduce an exact transformation that converts the “worst-case” violation across a set of potential violations to a series of ReLU-based layers that augment the original neural network. This allows verifiers to interpret them directly. Second, power system ML models often must be verified to satisfy power flow constraints. We propose a dualization procedure which encodes linear equality and inequality constraints directly into the verification problem; and in a manner which is mathematically consistent with the specialized verification tools. To demonstrate these innovations, we verify problems associated with data-driven security constrained DC-OPF solvers. We build and test our first set of innovations using the α, β-CROWN solver, and we benchmark against Gurobi 10.0. Our contributions achieve a speedup that can exceed 100x and allow higher degrees of verification flexibility.
Optimal Reconfiguration of Distribution Networks via Dynamic Programming
(2024-01-03) Meliopoulos, A. P. Sakis; Cokkinides, George; Yang, Zan; Cui, Bai
Modern active distribution systems are experiencing wide operating conditions due to distributed energy resources and prosumers with variable patterns. Faults can create loss of substantial parts of feeders. In both cases, optimizing the configuration of the distribution system can result in great savings. We define two problems: (a) optimal feeder reconfiguration to optimize operations under normal conditions, and (b) optimal feeder reconfiguration to maximize service following the occurrence of a fault. In this paper, we formulate both problem as an optimization problem of the dynamic programming variety. In the first problem, the objective is to minimize losses or minimize operational cost, while the objective of the second problem is to minimize the number of interrupted customers or minimize the total power not served. The solution for both problems is expressed in terms of an optimal sequence of switch control operations, subject to the operational constraints of the switches. The dynamic programming approach is very flexible and can integrate all pertinent operational constraints. However, it suffers from the Curse of Dimensionality. To mitigate the issue, state reduction technique is adopted in a successive dynamic programming algorithm to limit the dimensionality of the problem. The proposed algorithm has been tested on distribution systems to demonstrate
Stochastic Nodal Adequacy Platform: Spot Pricing of Adequacy
(2024-01-03) Tabors, Richard; Rudkevich, Aleksandr
The modern power system is becoming significantly more reliant on weather-dependent generation technologies. Existing resource adequacy metrics are not adequate for systems with a high penetration of weather-dependent, stochastically behaving renewable resources. This paper provides an overview of the Stochastic Nodal Adequacy Platform (SNAP), a novel approach for evaluating the adequacy of a large-scale electrical grid at the nodal level while accounting for the stochastic nature of weather-dependent system components, the physical operation of the system, and the economics and market design governing unit commitment and dispatch. The output of a SNAP analysis is a set of metrics that quantify the adequacy of the system and the physical contribution and economic value that each individual system component contributes towards overall system adequacy. The latter metric – the SNAP value – is an hourly marginal resource adequacy price at every node in the system that can be integrated into existing power market design.
EDGIE: A simulation test-bed for investigating the impacts of building and vehicle electrification on distribution grids
(2024-01-03) Priyadarshan, Priyadarshan; Pergantis, Elias; Crozier, Constance; Baker, Kyri; Kircher, Kevin
Replacing fossil-fueled appliances and vehicles with electric versions can significantly reduce emissions. However, electric heating and vehicle charging can cause peaks in electricity demand that stress infrastructure in buildings and power grids, jeopardizing reliability or forcing costly infrastructure upgrades. This paper presents the open-source EDGIE (Emulating the Distribution Grid Impacts of Electrification) toolbox. EDGIE matches experiment data from an all-electric home in cold weather. It can simulate many locations and levels of technology adoption, and supports optimization and network power flow simulation. In simulations of a fully electrified neighborhood during the coldest week of 2019 in New York, demand peaks at quadruple today's summer peak. Peaks are particularly sensitive to the use of overnight thermostat set-point reductions and to the efficiencies of heat pumps and building envelopes. Optimal vehicle-to-home coordination with flexible space and water heating reduces peak demand by 35% and transformer degradation by 99%.
Resource Adequacy Assessment from the Ground Up
(2024-01-03) Carreras, Benjamin; Newman, David; Lenhart, Stephanie; Blumsack, Seth; Kouts, Anna; Su, Wenjing
In response to the expanding role of wind, solar, and storage, increasing demand flexibility, and a changing climate, new analytical methods and metrics to assess resource adequacy are needed. A focus has been on identifying ways to reduce risks of failure. Less attention has been directed to how new analytical approaches can inform the design of planning processes, regulatory standards, and markets. Using mixed methods and a community-engaged approach, data on community preferences and uneven distributions of impacts are used in a demonstration of a coupled socio-technical systems model that has been validated in diverse settings. The research is informed by the physical and institutional infrastructures in the Railbelt power grid of Alaska. The findings illustrate how new analytical tools can inform institutional design and facilitate more affordable, sustainable, and equitable outcomes.
Inclusion of Reactive Power into Ecological Robustness-Oriented Optimal Power Flow for Enhancing Power System Resilience
(2024-01-03) Huang, Hao; Poor, H.Vincent; Davis, Katherine
Traditional optimal power flow problems focus on minimizing the operational cost which can result in a fragile system during unexpected contingencies. An ecological robustness-oriented optimal power flow (RECO OPF) problem has been proposed to incorporate ecosystems’ resilient characteristics into power system operations, enhancing their inherent resilience against multi-hazard contingencies. However, the formulation of RECO only considered real power flows but neglected reactive power flows. In this paper, we include reactive power into the formulation of RECO and propose a reactive power flow based RECO OPF (Q-RECO OPF) and an apparent power flow based RECO OPF (MVA-RECO OPF) to guide the distribution of power flows, respectively. By comparing a 200-bus system’s resilience against N-x contingencies using different OPF problems, we observe that both Q-based and MVA-based RECO OPF can provide a more resilient operating state.
Introduction to the Minitrack on Resilient Networks
(2024-01-03) Davis, Katherine; Newman, David

Browse

Recent Submissions