Securing substations from cyber attacks is essential to safeguard critical power infrastructure. However, digital substations that are based on the IEC-61850 standard have Generic Object Oriented ...Substation Events (GOOSE) messages and Sampled Value (SV) messages that are time-critical and thus cannot be protected using encryption techniques. This work presents a study on deception technology (decoys) for mitigating cyber attacks on GOOSE message virtual LAN (VLAN) which is a non-observable strongly connected bigraph. In this paper, the deployment of defender decoys is proposed by defining observable subgraphs in the VLAN. The defender-attacker interaction is modeled as a single-leader single-follower game with the defender as the leader. The optimal allocation of decoys for asset protection and attack detection is then formulated as a bi-level optimisation problem. Simultaneous allocation and sequential allocation of protection and detection decoys are considered for defender resource allocation. The existence of equilibrium of the defender-attacker game is proven. The model is illustrated in a 3-IED VLAN and performance is evaluated in a 12-IED VLAN system in the PSRC-I5 protection relay report. The results are compared with the zero-sum game model and it is found that the proposed model is capable of mitigating attacks in the GOOSE VLAN.
In deregulated power systems, reactive power ancillary service through electricity market is becoming relevant where private generation companies participate in maintaining system wide bus voltage ...within the permissible limits. Marginal cost price (MCP) based real time reactive power ancillary service market faces several challenges due to the localized nature of reactive power. In this paper, a market mechanism for real time reactive power ancillary service market based on Stackelberg game model is proposed considering voltage-apparent power coupled subsystems. In the proposed Stackelberg game model, Independent System Operator (ISO) is considered as the leader, and GENCOs as followers. In the formulation, each GENCO is associated with a relevance factor in the partitioned subsystem so as to consider the real time voltage support requirement in the system. The market is then formulated as Mathematical Program with Equilibrium Constraints problem (MPEC). Existence of equilibrium, incentive compatibility, and individual rationality of the proposed market mechanism is then analysed in this work. The numerical examples are illustrated in PJM 5-bus system, and tested on IEEE 30- bus system, and Nordic 32 Bus-system. The mechanism induces truth-telling behavior of GENCOs, yields a non-negative profit, and the system wide bus voltage is improved.
Modern power system is moving towards a smart and competitive grid, with competing generating companies, power retailers, and strategically behaving consumers playing a crucial role in the daily ...operation of the system. Independent System Operator (ISO) monitors these daily operations and procure required services through market operations. Active and reactive power pricing in real-time at the wholesale and retail level is considered an efficient energy management method. However, procurement of reactive power through a market mechanism in real-time is not being implemented, despite its crucial role in maintaining system parameters within the permissible limits. This paper attempts to detail the challenges faced in implementing such reactive power markets and presents a review of existing reactive power market mechanisms to address these challenges. A framework suitable for reactive power ancillary service in a smart grid is also detailed in this paper.
•The paper presents the discussion on a wide range of literature related to efficient allocation of reactive power provision in power systems.•The paper also covers related topics like market design and coupling across different levels of the power system.•The authors present an ideal framework for reactive power ancillary service market suitable for smart grid.
This work proposes a novel relative electrical distance measure that provides information of coupling between voltage and apparent power between two buses in power systems. Relative electrical ...distance measure is derived from the bus admittance matrix which can be obtained in real time using Phasor Measurement Units. Based on the relative electrical distance measure, in this work, an isoperimetric clustering based algorithm for partitioning power systems into voltage–apparent power coupled areas is proposed. The advantage of the partitioning algorithm proposed in this work is that large networks can be represented as a weighted graph with number of vertices equal to number of generators in the system which is much lesser than the size of system, thereby reducing the computational effort for partitioning. Isoperimetric clustering technique along with k-means is then applied to the graph to obtain voltage–apparent power coupled areas. Simulations carried out on New England 39-bus system and IEEE 118-bus system demonstrate the effectiveness of the proposed methodology for partitioning the system into voltage–apparent power coupled areas, subject to changes in the operating condition of the system. The quality of clustering is analysed and compared with Cheeger inequality bounds, which ensures that power system is well partitioned.
In real time electricity markets, the objective of generation companies while bidding is to maximize their profit. The strategies for learning optimal bidding have been formulated through game ...theoretical approaches and stochastic optimization problems. Similar studies in reactive power markets have not been reported so far because the network voltage operating conditions have an increased impact on reactive power markets than on active power markets. Contrary to active power markets, the bids of rivals are not directly related to fuel costs in reactive power markets. Hence, the assumption of a suitable probability distribution function is unrealistic, making the strategies adopted in active power markets unsuitable for learning optimal bids in reactive power market mechanisms. Therefore, a bidding strategy is to be learnt from market observations and experience in imperfect oligopolistic competition-based markets. In this paper, a pioneer work on learning optimal bidding strategies from observation and experience in a three-stage reactive power market is reported.
Traffic signal control is an important problem in urban mobility with a significant potential for economic and environmental impact. While there is a growing interest in Reinforcement Learning (RL) ...for traffic signal control, the work so far has focussed on learning through simulations which could lead to inaccuracies due to simplifying assumptions. Instead, real experience data on traffic is available and could be exploited at minimal costs. Recent progress in offline or batch RL has enabled just that. Model-based offline RL methods, in particular, have been shown to generalize from the experience data much better than others.
We build a model-based learning framework that infers a Markov Decision Process (MDP) from a dataset collected using a cyclic traffic signal control policy that is both commonplace and easy to gather. The MDP is built with pessimistic costs to manage out-of-distribution scenarios using an adaptive shaping of rewards which is shown to provide better regularization compared to the prior related work in addition to being PAC-optimal. Our model is evaluated on a complex signalized roundabout and a large multi-intersection environment, demonstrating that highly performant traffic control policies can be built in a data-efficient manner.
Implementation of IEC-61850 in the electrical substations has transformed them into digital substations. However, this has also exposed the communication network of the substation to cyberattacks, ...where an attacker can temper with GOOSE messages. To protect digital substations from potential cyberattacks, an effective intrusion detection system is very much required. Hence, in this work an unsupervised learning based intrusion detection system is proposed, which can detect the anomalies in GOOSE packets transmitted within the substation. Two unsupervised learning techniques, DBSCAN and autoencoder, are used in this work to develop an intrusion detection system, and their performance in detecting payload corruption is evaluated through numerical simulations.
Reactive power service is considered as an important ancillary service, due to its contribution towards maintaining system wide bus voltage. In this work, a value based reactive power pricing model ...suitable for real time market is proposed. In the proposed reactive power market model, bids(operation cost and lost opportunity cost(LOC) of reactive power) are received from the Generating Companies (GENCOs) participating in the real time market. From the bids received, a three component reactive power value function is formulated. The components of value function being load serving component, voltage support component and reserve component of reactive power requirement in the system. The objective of the Independent System Operator (ISO) is to minimise the value function of reactive power requirement in the system subject to generator limits, bus voltage limits and transmission line limits. From the Lagrange function of the optimisation problem, the marginal value (MV) of reactive power is derived to calculate the marginal price (MP) of reactive power. The NLP formulation of the market model is solved using DICOPT solver in GAMS and anlaysed on IEEE 24-Bus system. The simulation results prove that the proposed algorithm is efficient in clearing the reactive power market in such a way that the system wide bus voltage deviation is minimal and sufficient reactive power reserve is maintained in the network.
Demand response is a dynamic mechanism which helps to manage customer participation in response to power supply conditions under smart grid environment thereby contributing in system operation ...studies like Automatic Generation Control. This paper, which consists of two parts, presents Automatic Generation Control scheme under a deregulated environment using demand response. The first part proposes a transfer function model for power gain that can be achieved from price based demand response of thermostatically controlled loads. The second part proposes an Automatic Generation Control scheme under a deregulated system involving competing generation companies. In this paper Automatic Generation Control has been modeled as a collaborative stochastic game using Reinforcement based Learning. The advantage of this model is that the generation companies produce optimal amount of power that is required to minimize their cost of power production, while preserving the spirit of deregulation. Reinforcement Learning based Automatic Generation Control scheme incorporating demand response transfer function was simulated. The simulation results show that this approach can minimize the frequency deviation during load frequency control. The main contribution of the paper is the proposal of an AGC scheme that is suitable under smart grid (Demand Response) and deregulated environment (Competition among Generation companies).
Under deregulation, Automatic Generation Control (AGC) scheme may consider economic dispatch of Generating Companies (GENCOs) as well in addition to minimising the system frequency deviation. This ...will result in increased participation from GENCOs in AGC. GENCOs being private utilities, will have the private information like cost function etc which will not be shared with the system operator. This imposes a challenge on implementing classical economic dispatch problem in AGC scheme. The contribution of the paper is a game theoretic based model of AGC scheme that will ensure an optimal dispatch to competing GENCOs. This is achieved by defining an optimal participation factor for GENCOs that will minimize their cost of production. A game theoretic approach towards AGC as a dynamic game is presented and GENCOs are considered as Reinforcement learning agents. A Single Agent Q-Learning method along with pursuit algorithm is used at each GENCO agent to achieve the equilibrium point in AGC. The algorithm was studied with three competing GENCOs. Results show the suitability of the proposed AGC model and its ability to minimize the deviation in system frequency while ensuring economic operation of the system.