Publications  —  Devendra Shelar

Click on any title to see the corresponding abstract!

2019

  • Resilience of Electricity Distribution Networks - Part I: Cyber-physical disruption models [pdf]
    Devendra Shelar, Saurabh Amin, Ian Hiskens
    IEEE Transactions of Control Network Systems, 2019 (submitted)
    Abstract. ŒThis work contributes to the need for developing a systematic approach to evaluate and improve the resilience of electricity distribution networks (DNs) to cyber-physical failure events. We introduce a network model that captures the joint impact of physical failures in the transmission network and a class of cyberattacks (security failures) on DNs. These failures result in voltage disturbances and supply-demand disturbances at multiple DN nodes. The model is used to formulate a bilevel mixed-integer problem that captures the sequential interaction between an attacker (leader) and the DN operator (follower). The attacker (resp. operator) aims to maximize (resp. minimize) the post-contingency loss resulting from the cyber-physical failure events. We solve this problem by transforming it to an equivalent min-cardinality disruption problem and applying the Benders Decomposition algorithm. Our solution approach relies on a reformulation of the \enquote{coupling constraints} which model the effects of the attacker's discrete actions on the set of feasible operator response strategies. We evaluate the operator's value of timely response as the net reduction in post-contingency loss compared to the case with no response. This reduction can be viewed as the improvement in DN resiliency against the class of cyber-physical failure events.
  • Resilience of Electricity Distribution Networks - Part II: Leveraging Microgrids [pdf]
    Devendra Shelar, Saurabh Amin, Ian Hiskens
    IEEE Transactions of Control Network Systems, 2019 (submitted)
    Abstract. Recent technological advances in microgrids powered by Distributed Energy Resources (DERs) make them an attractive response mechanism for improving the resilience of electricity distribution networks (DNs) to reliability and security failures. This paper presents an approach to evaluate the value of implementing a timely response using microgrid operations and DER dispatch in the aftermath of of a disruption event, which involves strategic compromise of multiple DN components. Firstly, we refine the modeling and resiliency assessment framework in [1] and develop a sequential (bilevel) formulation which models attacker-operator interactions on a radial DN with one or more microgrids. Particularly, the operator response model includes microgrid operations under various islanding configurations (regimes), and single- or multi-master operation of DERs in providing grid-forming services as well as frequency and voltage regulation. Secondly, we introduce a restoration problem in which the operator gradually reconnects the disrupted components over multiple periods to restore the nominal performance of the DN. The first problem, formulated as a bilevel mixed-integer problem, is solved using Benders decomposition method. The second problem, formulated as a multi-period mixed-integer problem, can be solved using a greedy algorithm. Our computational results illustrate the benefit of using microgrids in reducing the post-contingency losses, both immediately after the disruption event and during the restoration process.

2018

  • DER Allocation and Line Repair Scheduling for Storm-induced Failures in Distribution Networks [pdf]
    Devendra Shelar, Derek Chang, Saurabh Amin
    IEEE SmartGridComm (SGC), 2018
    Abstract. Electricity distribution networks (DNs) in many regions are increasingly subjected to disruptions caused by tropical storms. Distributed Energy Resources (DERs) can act as temporary supply sources to sustain “microgrids” resulting from disruptions. In this paper, we investigate the problem of suitable DER allocation to facilitate more efficient repair operations and faster recovery. First, we estimate the failure probabilities of DN components (lines) using a stochastic model of line failures which parametrically depends on the locationspecific storm wind field. Next, we formulate a two-stage stochastic mixed integer program, which models the distribution utility’s decision to allocate DERs in the DN (pre-storm stage); and accounts for multi-period decisions on optimal dispatch and line repair scheduling (post-storm stage). A key feature of this formulation is that it jointly optimizes electricity dispatch within the individual microgrids and the line repair schedules to minimize the sum of the cost of DER allocation and cost due to lost load. To illustrate our approach, we use the sample average approximation method to solve our problem for a small-size DN under different storm intensities and DER/crew constraints.

2017

  • Compromising Security of Economic Dispatch in Power System Operations [pdf]
    Devendra Shelar, Pengfei Sun, Saurabh Amin, Saman Zonouz
    Dependable Systems and Networks (DSN), 2017
    Abstract. Power grid operations rely on the trustworthy operation of critical control center functionalities, including the socalled Economic Dispatch (ED) problem. The ED problem is a large-scale optimization problem that is periodically solved by the system operator to ensure the balance of supply and load while maintaining reliability constraints. In this paper, we propose a semantics-based attack generation and implementation approach to study the security of the ED problem. Firstly, we generate optimal attack vectors to transmission line ratings to induce maximum congestion in the critical lines, resulting in the violation of capacity limits. We formulate a bilevel optimization problem in which the attacker chooses manipulations of line capacity ratings to maximinimize the percentage line capacity violations under linear power flows.We reformulate the bilevel problem as a mixed integer linear program that can be solved efficiently. Secondly, we describe how the optimal attack vectors can be implemented in commercial energy management systems (EMSs). The attack explores the dynamic memory space of the EMS, and replaces the true line capacity ratings stored in data regions with the optimal attack vectors. In contrast to the well-known false data injection attacks to control systems that require compromising distributed sensors, our approach directly implements attacks to the control center server. Our experimental results on benchmark power systems and five widely utilized EMSs show the practical feasibility of our attack generation and implementation approach.
  • Towards Resilience-Aware Resource Allocation and Dispatch in Electricity Distribution Networks [pdf]
    Devendra Shelar, Saurabh Amin, Ian Hiskens
    Book chapter in Springer/IMA volume on The Control of Energy Markets and Grids, 2017 (under review)
    Abstract. ŒThis contribution presents an approach to improve the resilience of electricity distribution networks (DNs) to a class of cyber-physical failures by way of optimal allocation of distributed energy resources (DERs). The approach is motivated by the need to adapt the well-known security-constrained optimal power flow problem to DNs with remotely controllable (and hence, vulnerable) distributed generation sources or loads. To this end, we model the interaction between the system operator (SO) and an external adversary as a three-stage sequential game. In this game, the SO allocates the available resources (Stage 0) and also responds to the adversary’s action by optimally dispatching them (Stage 2). ŒThe adversary, on the other hand, compromises a subset of vulnerable components with the objective of inducing operating bound violations (Stage 1). We consider qualitatively different allocation strategies in the Stage 0, and develop a scalable greedy heuristic to solve the Stages 1-2 (i.e. bilevel optimization problem). We utilize this greedy heuristic to obtain structural insights about optimal adversarial compromises and desirable allocation strategies of the SO.

2016

  • Vulnerability analysis of Electricity Distribution Networks with large-penetrations of PEVs and DERs [pdf]
    Devendra Shelar
    (Master's Thesis), 2016
    Abstract. This thesis focuses on the vulnerability assessment of radial electricity distribution networks (DNs) under large-scale integration of Distributed Energy Resources (DERs) and Plug-in Electric Vehicles (PEVs). We formulate a two-player Stackelberg security game involving an attacker (external threat agent) and the defender (network operator). First, the attacker targets a subset of the insecure DER or PEV nodes, and strategically manipulates their set-points by attacking the DER/PEV controller logic at the nodes. Next, the defender responds to the resulting supply-demand mismatch by triggering network control operations, which includes direct load control and control of available non-compromised DERs/PEVs.
    The attacker's (resp. defender's) objective is to maximize (resp. minimize) the weighted sum of the cost of active and reactive power supply, costs of DER/PEV and load control, and the cost due to loss of voltage regulation. This composite cost captures the key trade-offs that the network operator faces in balancing power supply and quality during a broad range of contingency conditions. The choice of this cost in the security game reflects the attacker's overall goal of comprising the DER/PEV nodes to maximize the minimum composite cost for the network operator.
    Solving the sequential game with nonlinear power flow constraints is a computationally hard problem. To address this challenge, we introduce two auxiliary sequential game problems each with linear power flow constraints. We prove that the values of these relaxed problems upper and lower bound the value of the original game. Next, we introduce a greedy algorithm that can be utilized to efficiently compute an optimal attack strategy for both auxiliary games. Our main result is that, under a set of assumptions, the set of optimal attacker strategies is identical for these games, and hence we obtain a tractable solution to compute an optimal attack for the original game. Furthermore, the optimal attack strategy exhibits an interesting structural property: the downstream nodes are more critical for limiting costs of reactive power supply and maintaining voltage regulation. This insight is useful for vulnerability assessment of DNs under DER/PEV node compromises. Finally, we also exploit the structure of optimal attack to design a distributed control strategy for defender response.
  • Security Assessment of Electricity Distribution Networks under DER Node Compromises [pdf]
    Devendra Shelar, Saurabh Amin
    IEEE Transactions on Control of Networked Systems (TCNS), 2016
    Abstract. This article focuses on the security assessment of electricity distribution networks (DNs) with vulnerable distributed energy resource (DER) nodes. The adversary model is simultaneous compromise of DER nodes by strategic manipulation of generation set-points. The loss to the defender (DN operator) includes loss of voltage regulation and cost of induced load control under supply-demand mismatch caused by the attack. A 3-stage Defender-Attacker-Defender (DAD) game is formulated: in Stage 1, the defender chooses a security strategy to secure a subset of DER nodes; in Stage 2, the attacker compromises a set of vulnerable DERs and injects false generation set-points; in Stage 3, the defender responds by controlling loads and uncompromised DERs. Solving this trilevel optimization problem is hard due to nonlinear power flows and mixed-integer decision variables. To address this challenge, the problem is approximated by tractable formulations based on linear power flows. The set of critical DER nodes and the set-point manipulations characterizing the optimal attack strategy are characterized. An iterative greedy approach to compute attacker-defender strategies for the original nonlinear problem is proposed. These results provide guidelines for optimal security investment and defender response in pre- and post-attack conditions, respectively.

2015

  • A distributed strategy for electricity distribution network control in the face of DER compromises [pdf]
    Devendra Shelar, Jairo Giraldo, Saurabh Amin
    IEEE Conference on Decision and Control (CDC), 2015
    Abstract. We focus on the question of distributed control of electricity distribution networks in the face of security attacks to Distributed Energy Resources (DERs). Our attack model includes strategic manipulation of DER set-points by an external hacker to induce a sudden compromise of a subset of DERs connected to the network. We approach the distributed control design problem in two stages. In the first stage, we model the attacker-defender interaction as a Stackelberg game. The attacker (leader) disconnects a subset of DERs by sending them wrong set-point signals. The distribution utility (follower) response includes Volt-VAR control of non-compromised DERs and load control. The objective of the attacker (resp. defender) is to maximize (resp. minimize) the weighted sum of the total cost due to loss of frequency regulation and the cost due to loss of voltage regulation. In the second stage, we propose a distributed control (defender response) strategy for each local controller such that, if sudden supply-demand mismatch is detected (for example, due to DER compromises), the local controllers automatically respond based on their respective observations of local fluctuations in voltage and frequency. This strategy aims to achieve diversification of DER functions in the sense that each uncompromised DER node either contributes to voltage regulation (by contributing reactive power) or to frequency regulation (by contributing active power). We illustrate the effectiveness of this control strategy on a benchmark network.
  • Analyzing Vulnerability of Electricity Distribution Networks to DER Disruptions [pdf]
    Devendra Shelar, Saurabh Amin
    American Control Conference (ACC), 2015
    Abstract. We formulate a sequential (Stackelberg) game for assessing the vulnerability of radial electricity distribution networks to disruptions in Distributed Energy Resources (DERs). In this model, the attacker disrupts a subset of DER nodes by remotely manipulating the set-points of their inverters. The defender (network operator) responds by controlling the non-compromised DERs and by imposing partial load reduction via direct load control. The attacker's (resp. defender's) objective is to maximize (resp. minimize) the weighted sum of cost due to the loss of voltage regulation and the cost of load control. For the sequential play game where the attacker (resp. defender) is the leader (resp. follower) and under linear power flow equations, we show that the problem reduces to standard bilevel network interdiction problem. Under our assumptions on the attack model, we obtain a structural insight that the attacker's optimal strategy is to compromise the downstream DER nodes as opposed to the upstream ones. We present a small case study to demonstrate the applicability of our model for vulnerability assessment of distribution networks.

2012

  • Simulation of Multiple Line Rail Sections [pdf]
    Devendra Shelar, Priya Agrawal, Narayan Rangaraj, Abhiram Ranade
    International Simulation Conference of India (ISCI), 2012
    Abstract. In this paper, we discuss a simulation-based approach to determine a strategy for the effective use of third line in a 3-line railway section. A railway traffic simulator developed at IIT Bombay has been used to experiment with different scheduling strategies. The simulator provides a detailed picture of the running of multiple trains on a rail section, considering timetables of fixed schedule trains, dynamic assignment of paths of unscheduled trains, detailed infrastructure modelling (stations, signals and block sections) and detailed train running characteristics.We describe a typical railway section consisting of two unidirectional lines and one bidirectional line (also called as the third line). Numerical experiments have been performed considering different positioning of the third line relative to the other lines as this affects the traffic scheduling and track allocation. We have demonstrated the analysis using the Tiruvallur Arakkonam section in Southern Railways, where we have modelled the section layout and the traffic patterns. (of scheduled passenger trains as well as unscheduled freight trains). The performance measures used to quantify the section performance are weighted average traversal time and the maximum delay. The strategies that we analyse include fixed time interval reservation and variable time interval reservation based on traffic indicators. The approach is promising for analysing multiple line rail sections generally