Peer-to-Peer Networks Applied to Power Grid - Semantic Scholar

13 downloads 29144 Views 259KB Size Report
are named as servant (SERver + cliENT). This means .... due to expenses such as expensive, dedicated ... Agents buy and/or sell power trying to minimize the.
Peer-to-Peer Networks Applied to Power Grid* Hakem Beitollahi [email protected]

Geert Deconinck [email protected]

Katholieke Universiteit Leuven, Electrical Engineering, Kasteelpark Arenberg 10, Leuven, Belgium

Abstract In order to improve and deploy distributed control systems for the power grid a robust ICT infrastructure is required partially based on dependable middleware that exploits redundancy and diversity in communication networks. The dynamic nature of power grid applications makes necessity of using peer-to-peer (P2P) networks in this system. The main requirements of P2P networks for power grid are that they are scalable, self-organizing and distributed. Furthermore they degrade gracefully in the advent of failures, restore automatically after repair and need to cope with dynamic environment. This paper at first classifies P2P networks according their architecture and compares them. Then it analyzes P2P networks based on their architectures for controlling power grid components. It shows hybrid decentralized indexing P2P networks are best choice and most suitable architecture for power grid.

1. Introduction The power grid transports and distributes electricity from the power plants to the consumers [1] (figure 1). Improving performance of power grid (optimize power quality, cost, energy loss, etc.) needs precise management and distributed control [2]. Since Sep 11, 2001 the security of major international and national infrastructures in all countries has became a more critical concern to governments and industry. So nowadays a dependable networked ICT system for the management and control of the electric power grid is desired [3]. CRUTIAL (CRitical UTility InfrastructurAL resilience) is a recently started European IST research project addressing dependable network for the management of the electric power grid, in which agents controlling the *

Figure 1: power grid with a small number of power plants connected to the meshed high voltage grid and distributed generation and loads in the radial low voltage distribution

physical process of electricity distribution need to be connected with information infrastructures, through corporate networks, which are in turn connected to the internet. The dynamic nature of applications on power grid (e.g. the Autonomous Electricity Network applications [4]) makes that information infrastructure on power grid is needed not only for static configuration but also for modification during run time of application. For example the set of components that needs to communicate varies over

Acknowledgements: This project is partially supported by the K.U.Leuven Research Council (project GOA/2007/09) and by the European Commission (projects IST-4-27513 CRUTIAL and IST-4-026923 GRID).

time because of switching of generators and loads in the distributed generation and hence, the logical communication topology has to follow accordingly [2]. This dynamic nature of applications provides opportunities to use peer-to-peer (P2P) networks. The main requirements of P2P networks for power girds are that they are scalable, self-organizing and distributed. Furthermore they degrade gracefully in the advent of failures, restore automatically after repair and can cope with dynamic environment. They allow construction of a communication structure that requires little communication and also allow implementation of distributed algorithms for the control and coordination among nodes as well as for the aggregation of measured data. So studying and analyzing P2P networks carefully and studying whether they are sufficiently dependable for power grid is very necessary. This paper classifies the P2P networks based on their architectures and enumerates the advantages and disadvantages of them. It then evaluates and analyzes P2P networks for controlling power grid components and shows the hybrid decentralized indexing P2P type is the good and suitable choice for power grid. This paper provides guidance for researchers and practitioners to design robust and effective P2P networks applications requiring basic knowledge about P2P networks, accessibility, fault resilience, control mechanisms of power grid and how a computer network like P2P is applied to power grid. The rest of this paper is organized as follows: in section 2 definition of P2P is presented. Section 3 presents the P2P architectures. Section 4 briefly studies the basic control base for new power grid system. Section 5 shows P2P network is a new opportunity for controlling power grid components and section 6 presents conclusion.

2. Peer-to-Peer systems Many different definitions for P2P exist [5-10]; we can extract several principles about P2P networks from these definitions that will be useful in the context of power grid. 1) The principle of resource sharing: all P2P networks have an aspect of file or resource sharing (both physical resources and logical resources). As a result of resource sharing, applications can be realized which it could not be set up by a single node. 2) The principle of decentralization: Parts of the network or even the entire network are no longer operated centrally. Decentralization is a

particular character of P2P networks in order to avoid single point of failure as well as performance bottlenecks in the network. 3) The principle of self-organization: as a result of decentralization, there exists no longer a node that can centrally coordinate its activity or a database to centrally store global information about the system. Therefore nodes have to selforganize themselves, based on whatever local information is available and interacting with locally reachable nodes. In the last few years, P2P systems have rapidly evolved and emerged as a promising platform to deploy new applications and services in the internet [11, 12]. One of the reasons P2P systems are seen as an excellent platform for large scale distributed systems is their resilience in the presence of node failures. However malicious faults (attacks) are serious threats in these systems.

3. Peer-to-peer Architecture According to differences of query search mechanism, logical topology and degree of decentralization, the P2P architecture can be classified into four types.

a) Purely unstructured decentralized This P2P system is a distributed system without any centralized control. In such systems all nodes are equivalent in functionality. In these systems nodes are named as servant (SERver + cliENT). This means all nodes of a P2P system can act at the same time as server as well as a client. The logical P2P topology in these systems is often random and unstructured mesh. The query is executed hop-by-hop through the mesh till success/failure or timeout. An example of these systems is Gnutella [13]. Routing algorithm: Purely unstructured systems like Gnutella use flooding broadcast of queries for routing. In these systems each query from a peer is flooded (broadcasted) to directly connected peers, which themselves flood their peers and etc., until the request is answered or a maximum number of flooding steps occur. Different flooding policies have been implemented to improve search in these systems. In Gnutella to find a file, a node issues queries to its neighbors. The most typical query method is flooding where the query is broadcasted to all neighbors (hops) within certain TTL (Time-To-Live) value. At each hop the value of the TTL is decremented, and when it reaches zero the

message is dropped. In order to avoid loops, the nodes use the unique message identifiers to detect and drop duplicated messages. When the file was found in a certain node, it initiates a direct out-ofnetwork download, establishing a direct connection between the source and target node (figure 2).

Examples of this system are Chord [14], CAN [15], and Pastry [16]. Routing algorithm: In these systems the routing model adds structure to the way information about resources are stored using distributed hash tables. This structure provides a mapping between the resource identifier and location, in the form of distributed routing table. With this method the queries can be efficiently routed to the node with the desired resource. This method also reduces the number of hops that must be taken to locate a resource. Among examples of these systems we have chosen the Chord routing algorithm. Figure 3 shows Chord routing model. N8+1 N8+2 N8+4 N8+8 N8+16 N8+32

N1 N8

Finger table

Source

N54

Query Response

N14 N14 N14 N14 N32 N42

Destinstion

N14 N51

N48

Figure 2 : Flooding Search

N21

N42

• i)

Advantages of these systems: They are inherently fault-tolerant, since there is no central point of failure and the loss of a peer or even a number of peers can easily be recovered. ii) They are flexible and robust. • i) ii) iii)

Disadvantages of these systems: They are slow in routing. No guarantee about quality of service. Predicting system behavior is hard because of the lack of a global view at the system level. iv) They are not scalable. v) Managing the system is hard.

b) Purely structured decentralized They are similar to P2P systems of group A with this difference that the logical P2P topology is a structured topology such as mesh, ring, d-dimension torus or butterfly. These structured topologies are usually constructed using distributed hashing tables (DHT) techniques. The query is also executed hopby-hop through the structured topology, and is sure to be successful after some deterministic hops under ideal case.

N38

N32

Figure 3: Chord routing model

• i) ii) iii) iv) v)

Advantages of these systems: They have a quick and efficient query search or in other words they are quick in routing. They provide load balancing. They have a deterministic search guarantee. They are scalable and robust. Managing the system is easy.

• i)

Disadvantages of these systems: They are vulnerable under malicious attacks. In comparison with group A they are worse under malicious attacks because they have deterministic and structured architecture and this is a important point for attackers. ii) They are not flexible.

c) Hybrid centralized indexing In these systems there is a central server that maintains directions of information about registered

users to the network. In these systems each arriving node needs to actively notify the server about its information, then other nodes only need to search peer’s address from server about its wanted objects. There is the end-to-end interaction between two peer clients. Napster [17] is an example of these systems. Routing algorithm: In these systems the peers of the community connect to a centralized directory server, which stores all information regarding location and usage of resources. Upon request from a peer, the central index will match the request with the best peer in its directory that matches the request. The best peer could be the one that is cheapest, fastest, nearest, or most available, depending on the user needs. Then the data exchange will occur directly between the two peers. Napster uses this method. Figure 4 shows routing in Napster as an example of these systems.

and proxy search requests on behalf of these peers. In such systems peers are automatically elected to become super-nodes if they have sufficient bandwidth and good processing power. Like systems of group C there is end-to-end interaction (data exchanging) between two peer clients. Figure 5 shows these systems. Kazaa [18] and Morpheus [19] are two decentralized indexing systems.

Super-node

Server Peer “Where is X.doc” A

Download X.doc

Figure 4: Routing algorithm in Napster

• Advantages of these systems: i) They are very simple and easy to be deployed. ii) They are quick and efficient for information discovery (routing or query search). iii) Searches are comprehensive and they can provide guarantee in searches. iv) Managing the system is easy. • i) ii) iii) iv)

Figure 5: Hybrid decentralized indexing

A has X.doc

Disadvantages of these systems: They are vulnerable to malicious attacks. They are single point failure in server. They have bad scalability. They are not so flexible and robust.

Routing algorithm: In these systems every super-node is associated with a set of nodes and every node connects to a supernode to which it belongs. When a search for a data item issued by node n1, lookup message will follow the path from a super-node of node n1 to all super nodes, the operation will repeat until success or until all paths are completely searched. • i) ii) iii) iv)

d) Hybrid decentralized indexing: In these systems some nodes that called super-nodes are central servers and they register users to the system and also facilitate the peer discovery process. Super-nodes maintain the central indexes for the information shared by local peers connected to them

v)

Advantages of these systems: They are quick and efficient for search query (routing). In comparison with other groups they reduce the traffic on message exchanging. In comparison with group C they reduce the workload on central server because they have several servers. In comparison with group C there is no unique point of failure. In these systems if one or more super-nodes go down, the nodes connected to them can open new connections with others, and the network can continue the operation. In fact after super-nodes go down, the existing peers become super-nodes themselves. They are scalable.

vi) They are flexible and robust. vii) Managing the system is easy. • i)

Disadvantages of these systems: In comparison with group A and B they are more vulnerable to malicious attacks because of supernodes. ii) In comparison with group C, they present slower information discovery.

4. Basic control base for new power grid system Today, there is an important trend to use small dispersed generators in low or medium voltage, also referred to as Distributed Generation (DG), for producing electricity [20, 21, 22]. However, this dramatically changing in power grid has changed the traditional distributed grid topology, control systems, security measurements, etc [20, 21]. All these changes put new and extra stress on power grid where power is one of the most important commodities for industrial, economical and everyday activities. Control of power elements in power grid has three control levels [4, 22, 23] A) Primary control is used to balance both active and reactive power, based on frequency measured locally. This kind of control has no need for communications. B) Secondary control is mainly used for maintaining rated voltage levels or rated frequency and scheduled power transfers C) tertiary control optimizes generators output for economic criteria. The last two control levels both require some form of coordination and communication with other generator controllers. By adding DGs to power grid for these levels of control, traditional central control systems are not suitable due to expenses such as expensive, dedicated communication lines and a large number of load balancing servers. So we are looking for less expensive ICT-infrastructure for control paradigms. Agent based control systems Evolutions in the power grid have convinced us to design an intelligent, distributed control scheme [2, 24]. In this scheme power grid components (e.g. generators, dispatching loads) are equipped with an autonomous control entity (agent). These agents are implemented on some electronic devices (e.g. PCs or DSPs). Agents can supervise the grid component at hand, fetching state parameters of the components of the component from various sensors. Agents interpret these parameters and aggregate these into a high

level conclusion on the current state of the component. Agents may also send control parameters to the control system (e.g. SCADA) of the component. An important aspect of agents, besides autonomy, is that they communicate with a society of (similar) agents, from which they may fetch external knowledge of their interest. This way of exchanging information enables an agent to not be only aware of the state of its own component, but also to have a notion of the global environment in which it is operating. These societies can be built easily by setting up a peer-to-peer network.

5. P2P networks for Microgrid control applications Agent based control scheme by using P2P communication opens new opportunities for controlling power grid elements. Various applications of agent based control schemes (smart grids) have been proposed in literature. Here we give a short overview of some noteworthy applications and role of P2P network on them. Intelligent Protection: paper [25] has shown that malfunctioning or wrongful switching of protection systems are at basis of 63% of all major system disturbances. Therefore various schemes are presented to control these circuit breakers by intelligent agents, using for example fuzzy logic and communication with other agents to decide the best action when a near short circuit current is observed [26]. By the P2P network, the agent floods the problem to other agents and then by collective agreement algorithms they can decide the best action. All types of P2P network can be useful for this application especially type B and D for some time deadlines of best action. This application can also be done by type D very efficiency. The agent encountering a problem sends a message to its server and then the server wants other servers (any server has some agents) to send their commands for solving the problem. So the server can decide the best action intelligently based on all commands. Agents in Power Market: Multi-agent technology can be used for real-time online power markets. Agents buy and/or sell power trying to minimize the costs and/or maximizing income while meeting the demand of their master (e.g. generation facility, factory, household, etc.) [27]. P2P technology provides directly communication between any two agents and also provides communication and discussion of an agent with all other agents. So with

these properties of P2P network and also file sharing property of this network, agents can minimize the costs and/or maximize income according the best decision in power market. Supply demand matching: Intelligent forecasting of demand and generation (e.g. for solar cells or wind turbines) and matching these by adjusting generator output or intelligent loads, may reduce regulating needs and costs [27]. Intelligent load shedding: loads may be switched off when network load is high, according to some properties. Heating or air conditioning may be switched off temporary without anyone noticing, while hospitals or traffic signals may only be shed when there are no other options. As nodes (agents here) are autonomous and self-organized in P2P networks, they can switch off some loads like heating or air condition temporarily based on some conditions to balance the network load. Autonomous electricity grid (AEG): An AEG is low voltage net with a decent amount of DG installed, that is self-regulating and possibly able to continue operating, if necessary in a degraded way, when it is disconnected from the high voltage transmission grid [2, 28, 29]. We can consider the control sites of power plants (high voltage plants and DGs) as LANs where any LAN has a super-node to communicate with other LANs. Super-nodes communicate together by a P2P network (recall type D of P2P architecture). As Super-nodes are members of P2P network they are autonomous and selforganized and can leave the network based on their interest or power grid conditions. With this architecture the AEG application is implemented. For example if DG resources are disconnected from the high voltage grid, we can consider that super-nodes of distributed generators’ LANs have been disconnected from super-nodes of high voltage generators’ LANs and peer-to-peer connection still exist among super-nodes of distributed generators’ LANs. Importance of P2P network for secondary and tertiary control As mentioned before secondary and tertiary controls need some form of coordination and communication with other generator’s controllers. For this agents (generator’s controllers) set up a P2P network which directly or indirectly interconnects all controller agents in the distribution grid. The topology and fault-tolerance mechanisms of this overlay network make it suitable to be used in unbounded systems using an unreliable open network such as Internet. A

simple example of P2P network formed between agents has shown in figure 6.

Low Voltage Distribution Grid with embedded DG

Corresponding Peer-to-peer Network

Househould (with smart loads) Wind torbin Fossil fuel driven small generator Photovolic generator Peer-to-peer network links Public communications LV power line Grid connected transform Controller agent

Figure 6: Example of P2P network that applied to power grid

Agents communicate together by gossiping algorithm. In gossiping algorithm, every agent exchanges information at fixed time intervals with one of its neighbors (chosen randomly). If that neighbor exchanges this new information with one of its neighbors (and so forth), the news spreads in the network. All types of P2P architectures implement the gossiping algorithm. In type A it is enough any node (agent) exchanges information with one of its random neighbors and so forth. In type B as any node has its finger table list it is enough any node exchanges information with members of its finger table list and then the members exchange information with members of their finger table list. In types C and D any node just exchanges information to its server (super-node) and then server (super-node) spread information over the network by exchange information to other its nodes or other servers (supernodes) and so forth. Evaluation of the four P2P types for power grids Power grid network is a scalable network and multiple power plants maybe are added to the grid. Some applications of Power grid have time deadlines, which mean that the correctness of a system operation depends not only on its logical results, but also on the time at which these results become available. So power grids need a network with quick and efficient routing. Power grid also requires efficient management for three control levels and power components. With these interpretations and situations type A of P2P is not suitable for power grid because it is not scalable; it is slow in routing; and also management is hard. However this type is good for security and fault-tolerance. Type B of P2P networks for power grid is efficient as it is scalable; it is quick and efficient in routing and also management is easy. However, this type is more vulnerable from security outlook due to deterministic and structural architecture. Type C is not good for power grid as it has bad scalability; it is vulnerable from security point of view because it is enough that the attacker only considers the server, which is a single point of failure (SPF). However, it is simple and easy for management and it has quick routing. As power grid applications shows, type D of P2P is the most suitable and best for power grid because it is scalable and flexible; it has quick and efficient routing. It has not SPF points like type C; management is easy and it is more secure in comparison with other types except type A.

6. Conclusion Considering advantages and disadvantages of different architecture of P2P networks, system designer can choose a topology for their system according to their requirements. Evaluations in the power grid show it requires an intelligent, distributed and dynamic control system. P2P networks are good and suitable choice for power grid due to dynamism, scalable, self-organizing, and decentralized nature. Preliminary studies on P2P types (considering advantages and disadvantages of P2P types and resilience of P2P types to random faults and attacks) show type D, the hybrid decentralized indexing type, is the most suitable and best for power grid among P2P types because of scalability, flexibility, quick and efficient routing and easy management.

7. References [1] K. Vanthournout, G. Deconinck, R. Belmans, “A middleware control layer for distributed generation systems”, IEEE Power Systems Conference and Exhibition (PSCE), New York City, USA, Oct. 2004. [2] K. Vanthournout, G. Deconinck, R. Belmans, “Agora: Distributed Tertiary Control of Distributed Resources”, 15th Power Systems Computation Conference, 2005, Liege, Belgium. [3] K.Lin, K.E.Holbert, “PRA for vulnerability assessment of power system infrastructure security”, Proc. of the 37th Annual North America, Power Symposium, 2005. [4] K.Vanthournout, “A semantic overlay network based robust data-infrastructure, applied to the electric power grid”, Ph.D thesis, Department of Electrotechniek-ESAT, Katholieke universiteit Leuven, April 2006. [5] P. S. Munindar, “peering at peer-to-peer computing”, IEEE Internet computing, 5(1): 4-5, 2001. [6] K.Kant, R.Tyer, V.Tewari, “A Framework for classifying peer-to-peer technologies”, Proc. of the 2nd IEEE/ACM International symposium on cluster computing and the Grid (CCGRID’02), 2002. [7] K.Aberer, P.cudr, M.Hauswirth, “The chatty web: emergent semantics through gossiping”, Proc. of the 12th Inter. Conf. on word wide web, pages 197-206, ACM press, 2003. [8] A.oram, “Peer-to-Peer: Harnessing the Power of Disruptive Technologies”, O’Reilly & Associates, Inc., Sebastopol, CA, USA, 2001.

[9] R.Schollmeier, “A definition of Peer-to-Peer networking for the classification of peer-to-peer architectures and applications”, Proc. of the 1st Inter. Conf. on peer-to-peer computing (P2P2001), pages 101-102. IEEE computer Society, Aug 2001. [10] K.Aberer, M.Hauswirth, “An overview on peer-topeer information systems”, Proc. of the workshop on Distributed data and structures (WDAS 2002), Paris, France, 2002. [11] R.Buyya, H.Stockinger, J.Giddy, D.Abramson, “Economic Models for management of Resources in Peerto-Peer and Grid computing”, Proc. of the SPIE Inter. Conf. on Commercial Applications for High-Performance Computing, 2001. [12] T.Reti, M.Valimaki, “Business Models for content Distribution on Mobile Peer-to-Peer Networks”, Working paper, Helsinki Institute for Information Technology, 2002. [13] Gnutella: http://www.gnutella.com [14] I.Stoica, R.Morris, D.Karger, F.Kaashoek, H.Balakrishnan, “Chord: A Scalable Peer-to-Peer Lookup Service for Internet Applications”, Proc. of the Conf. on applications, technologies, architectures, and protocols for computer communications (SIGCOMM’01), Pages 149160, San Diego, California, USA, August 2001. [15] S.Ratnasamy, P.Francis, M.Haudley, R.Karp, S.Shenker, “A scalable content addressable network”, Proc. of the Conf. on applications, technologies, architectures, and protocols for computer communications (SIGCOMM’01), pages 27-31, San Diego, California, USA, August 2001. [16] A.Rowstron, P.Druschel, “Pastry: scalable, decentralized object location and routing for large-scale peer-to-peer systems”, Proc. of the 18th IFIP/ACM Inter. Conf. on Distributed Systems Platforms (Middleware 2001), Heidelberg, Germany, November 2001. [17] Napster: http://www.napster.com [18] S.G. Nathaniel, A.Krekelberg, “Usability and privacy: a study of Kazaa P2P file-sharing”, Conf. on Human factors in computing systems (CHI’03), pages 137-144, ACM Press, 2003. [19] Morpheus: http://www.morpheus.com. [20] J. Cardell, M. Ilić, R. D. Tabors, “Integrating Small Scale Distributed Generation into a Deregulated Market: Control Strategies and Price Feedback”, MIT Energy Laboratory Technical Report, MITEL-98-001, April 1998. [21] S. Massoud Amin, B. F. Wollenberg, “Toward a Smart Grid”, IEEE Power & Energy Magazine, Vol. 3,umber 5, September/October 2005, p. 34-41.

[22] T.Rigole, K. Vanthournout, G. Deconinck, " Interdependencies between an Electric Power Infrastructure with Distributed Control, and the Underlying ICT Infrastructure," International Workshop on Complex Network and Infrastructure Protection 428-440, Rome, Italy, March 28-29, 2006; pp. 428-440. [23] T.Rigole, K. Vanthournout, G. Deconinck, " Distributed control systems for electric power applications," 2nd International Workshop on Networked Control Systems: Tolerant ro Faults , Rende (CS), Italy, Nov.23-24, 2006; 7 pages. [24] A. Massoud, S. Wollenberg, “Toward a smart grid” IEEE Power and Energy Magazine 3(5), 34–41, 2005. [25] NERC, “Review of selected electric system disturbances in north america”, Report of NERC Disturbances Analysis Working Group, 1995. [26] J. Huang, S.S. Venkata, “Wide area adaptive protection: Architecture, algorithms and communications”, International Journal of Critical Infrastructures 1(1), 8–19, 2004. [27] J.K. Kok,, C.J.Warner, I.G. Kamphuis, “Powermatcher: Multiagent control in the electricity infrastructure”, Proceedings of Autonomous Agents and Multi-Agent Systems, Utrecht, Netherlands, 2005. [28] A.L. Dimeas, N.D. Hatziagyriou, “Operation of a multiagent system for microgrid control”, IEEE Transactions on Power Systems, 2005. [29] P. Lund,, S. Cherian, T. Ackerman, “A cell controller for autonomous operation of a 60kv distribution area”, International Journal of Distributed Energy Resources, 2006.