By Michael B. Robinson, IBM
Smarter electrical grids are getting a lot of attention these days. Utilities want to meet the growing demand for electricity while improving electric service delivery and being good environmental stewards. Telecommunications is key in making this happen.
As the smart grid concept matures, utilities and network designers will see a myriad of network technologies. They will need a way to organize their thoughts on these complex interconnected networks. The smart grid telecom reference model (SG-TRM) offers a conceptual framework that supports smart grid planning and provides the necessary organization.
Smart grid thinking could be organized around technologies (wireless, fiber, etc.), but this way of distinguishing interconnected networks can be confusing. For example, many forms of wireless technologies exist, but they operate differently. In addition, fiber can refer to long-haul networks or simply networks that connect servers to network switches in data centers.
Conceptual frameworks can better facilitate an understanding of how things fit together to form the whole. With this in mind, the SG-TRM refers to the purpose or function of its domains instead of the technologies. Consistent references to functional domains provide clarity during discussions that cross organizational boundaries. The open systems interconnection reference model (OSI model) achieves a similar goal. Engineers of varying expertise might refer to the application layer, network layer and so on and achieve a clear understanding of the topic. The lead artwork depicts an end-to-end SG-TRM consisting of six network domains, each having distinct purposes and attributes, which are described below.
A Look at the Networks
The extranet: All utilities interconnect with third parties; they may be suppliers, operating partners or the public at large. The extranet makes interconnections effective, safe and reliable.
When security risks associated with outside environments exist, application and network devices must be implemented on extranet connections. In contrast, other domains may not require the same kind of treatment, except to secure operational environments from other internal environments.
Historically, extranet connections were centralized at hosting centers (data, control, operation and call centers); however, as smart grid technology matures, changes in security architecture will be needed to accommodate distributed extranet environments.
Extranet network technologies include base band and T1 voice circuits, cellular connections (voice and data, including wired and wireless connections), data technologies (multiprotocol label switching (MPLS), frame relay, digital and analog private lines) and Internet connections. Although some commonalities exist among extranet technologies and those in the core (discussed below), it is important to treat these domains differently because of functional (mostly security) concerns.
Core networks: Core networks provide high-performance, highly reliable connections between application hosting centers. Traffic volume is heaviest in the core because it also includes traffic from distant locations. It is critical for utilities to keep the core network operating because the loss of network connectivity can result in serious operational problems. Accordingly, carrier-class equipment and technologies are used in the core.
Fiber is most often used for the core because it supports the highest capacity. Core networks with relatively low fiber counts may require some form of wave division multiplexing (WDM) to increase the number of available light channels. Synchronous optical networking (SONET) was frequently used, but because of its scalability and cost, gigabit-Ethernet is now being used.
A critical feature of SONET is its ability to support alternative path failover within 50 milliseconds, which is transparent to most applications. Gigabit-Ethernet can only perform this quickly if MPLS fast re-route (MPLS-FRR) is deployed. MPLS is rapidly becoming the choice packet switching technology for core networks. Failover dictates that diverse paths must be available between devices, which can be expensive to construct. Because of its higher cost structure and for the sake of performance, utilities should be judicious when adding nodes to an operating core.
Backhaul networks: Backhaul networks are typically linear spurs that reach out to locations remote from the core. Backhaul economics generally preclude construction of redundant paths for failover. From a smart grid operations perspective, this constraint dictates that backhaul be treated differently.
For instance, when a link failure occurs on backhaul, all connectivity to downstream devices is lost. If grid applications (like demand management) were network-aware, then they would not spend time attempting to communicate with off-net devices. Instead, applications could recompute and determine if other portions of the grid might deliver demand reduction.
Scenarios like this, and the backhaul network’s value to the business plan should be considered during telecom planning to ensure the proper network technology (fiber, microwave, spread spectrum, etc.) is selected. Typically, private backhaul extensions are justified link by link as part of a public circuit telecom expense-reduction program. The emphasis on cost reduction can result in implementing equipment with minimal capabilities (compared with that found in the core). When utilities are considering performance, it is important to remember that long backhaul link concatenations–each representing a router hop–are not friendly to some real-time applications, especially if the equipment has medium-class performance ratings. Careful studies of trade-offs and break-even points are required to determine the optimal solution.
Access networks: The access network’s objective is to connect remote devices. The volume and density of these devices are important when choosing the appropriate access technology.
Wireless access network connections generally occur at a substation or an existing radio tower (or cellular tower if public services are used). Access networks must reach in all directions to grid devices located many miles away from the substation, which can be a challenge. (As alternatives to wireless access, power line carrier and broadband over power line have had a degree of success with some electric utilities.)
Wireless reach depends on many criteria; frequency, signal power and terrain are critical. Unfortunately, utilities have no access to spectrum that is conducive to long-range, areawide operation in diverse terrain (e.g., spectrum below 1 GHz) that can support broadband applications. Spectrums available for utilities include the industrial, scientific and medical (ISM) and the unlicensed national information infrastructure (UNII) bands.
Although these bands have been used in innovative ways, their operation using a hub-and-spoke architecture makes covering a substation’s service territory a serious technical challenge. Furthermore, the utility assumes some increased operational and financial risk controlling the nation’s electrical grid with any spectrum shared by the general public. Nonetheless, advancements must be made with the available spectrum. Some vendors have introduced access networks based on mesh radio architectures. Others are holding out for licensed spectrum that can use higher power levels.
Access networks are most needed in the distribution grid–a new frontier for networking. The distribution grid will see many changes as new applications and increased functionality are introduced. Treating the access network as a distinct domain with its own functional requirements helps contain risks.
Neighborhood network: The neighborhood network domain is another new frontier, with its roots supporting meter reading. Because of interference issues associated with ground-level communication, mesh radio attempts to deal with power limitations by leveraging nearby radio devices. Mesh routing algorithms move traffic upstream or downstream as required by hopping from one device to another.
Although neighborhood networks predominantly support metering functions, vendors recently began trials to support distribution grid devices such as capacitor bank controllers, voltage regulators, switches and sensors. Whether mesh radio performance can adequately support these devices remains to be seen. If so, they might take pressure off the access network by supporting end nodes of any type that are in close range.
Home area networks: The future of smart grid likely will incorporate some form of home area network (HAN) that can directly or indirectly control electricity usage at the appliance level. In-home displays (IHDs) will communicate pricing signals and usage information from the home’s electric meter. Air conditioners, furnaces, pool pumps and other high-energy-consuming devices may be temporarily turned off for short periods as demand approaches the utility’s ability to supply. Most HANs use wireless networks that leverage the ZigBee protocol. HAN ownership and support ultimately will be the homeowner’s responsibility. The electric meter simply will participate in this network.
The SG-TRM described here does not compete with other frameworks or models. For example, it is consistent with the GridWise Architectural Council’s context-setting framework, facilitating discussion and choices among networking alternatives. It is also complementary with the open systems interconnection (OSI) reference model. Within SG-TRM domains, network designers can discuss the session layer, transport layer and so on. Physical and data link layers of each SG-TRM domain might share similar technologies yet have different functionality.
For instance, the access domain might be supported by IEEE 802.16 wireless technology whereas the neighborhood domain might be supported by proprietary wireless mesh, yet the two domains have different purposes. Although the OSI reference model typically conveys a vertical hierarchy (layers), the SG-TRM predominantly describes a horizontal interconnection of functional domains.
The SG-TRM’s six domains make two things apparent. First, each domain serves a unique purpose. All domains support smart grid communications in unique ways. The core network serves hosting centers and not in-home displays. Access networks gather traffic from wide coverage areas and are inappropriate for communications between two hosting centers.
Second, each domain has certain technologies that fit its purpose, yet other technologies are precluded. For example, wireless mesh networks seem ideal for meters in neighborhoods but would not be a good choice for the core network. Similarly, fiber is a great technology for the core network but not for IHDs. These examples are extremes and exceptions to what is intended to be a general rule.
Networking is a field with constant changes and improvements, therefore, utilities should regularly reconsider applicable choices. The SG-TRM can be used to frame such discussions as they relate to the smart grid, especially regarding interoperability, security and requirements for domain functionality.
Michael B. Robinson is a network architect on the IBM energy and utilities industry team. Reach him at firstname.lastname@example.org.