6.2.1 Smart Grid

The introduction of the smart grid increases the amount of information exchange and the functionalities. The smart grid has the following information‐generating and ‐consuming functions.

1. Consumers can now be producers of power. Therefore, power production information should be collected from consumers as well. This makes consumers a new source of information.
2. Accurate load balancing of the production requires dynamic load information exchange from all power generation stations.
3. Smart sensors collecting information from power transmission lines, transformers, distribution substations, etc.
4. Advanced smart meters collecting power usage information.
5. Information collected from electric vehicles (EV) and high‐capacity energy storage devices, such as large batteries in EVs.

With the integration of the IoT to the smart grid, virtually all assets from anywhere can connect, and efficiencies and productivity can increase. This ubiquitous connectivity will transform our businesses; new services that improve our lives will be possible. IoT will offer new value to utilities enabling them to make better business decisions, use their resources more efficiently, and be the best service provider possible to their customers.

There are multiple types of data collected and moved in the smart grid with different requirements. These are: (1) time‐critical data, and (2) volume data. Time‐critical data has stringent time budget requirements for delay and is needed for decisions including protection switching. The volume data does not have such strict requirements for delay but needs high throughput; meter data is one of them.

The geographical footprint of the smart grid communication network is large and therefore employs heterogeneous networks in different sections of the entire coverage. Optical fiber networks are commonly used for connecting generation sites to substations and deployed along with transmission lines and are called core or wide area networks (WAN). In field area networks (FAN) and neighborhood area networks (NAN), the bandwidth is not high; however, flexibility and access are important. Power line communication and/or wireless access technologies are frequently used in distribution areas.

In the traditional power grid, the devices and protocols used in substations in the distribution areas were proprietary and closed systems. To increase the interoperability of many other new devices and functionalities that are coming in to the picture with the smart grid, a standardization of substations automation became a need. The TC 57 working group released a series of standards called IEC 61850 (IEC) family where inter‐ and intra‐substation devices, communication, and functions have been reviewed and standardized.

In the process of modernizing the traditional power grid toward the smart grid, there is an increasing need for communication between machines as well as machine and human. This need puts information and communication technologies (ICT) in a vital role for this evaluation and modernization to happen smoothly.

In the smart grid, there are technology and business challenges. The integration of renewable distributed energy resources (DER) will require frequent and close monitoring of the entire system.

Some of the technology challenges of the smart grid include the following integration of systems and functionalities:

• demand side management (DSM);
• automated meter reading (AMR);
• customer energy management systems (CEMS) to manage new prosumers (consumer and producer) into the smart grid architecture;
• hybrid and plug‐in electric vehicles with large batteries that require coordination for storage from the grid;
• intelligent electronic devices (IED) that are used for collecting, monitoring, and processing data;
• self‐healing capability against failures and disasters (earthquakes, hurricanes, tornados, etc.);
• quick isolation of failures in microgrid to prevent large blackouts;
• providing predictive information to utilities as well as subscribers with recommendations about efficiencies in usage as well as production;
• interoperability among multiple heterogeneous networks and domains; the smart grid should be able to work multiple networks and platforms together with no inconvenience to users and other systems; and
• real‐time analytic information processing systems to determine the current status and predictions for what is coming next.

Business challenges of smart grid include:

• increase energy trade, not only for traditional power plants, but from individual consumers/producers (prosumers); and
• developing business cases for new services and applications (predictive and/or historical analysis of usage and production for example).

The potential solutions for these challenges lead to high volume of data collection, transfer, and processing, which increase the complexity and the requirements of the ICT infrastructure. In addition to this complexity, the ICT infrastructure still requires even higher performance, availability, reliability (Abhishek, 2016), dependability, stability, and security. To combat this complexity, there have been proposals from researchers and practitioners in the industry, which include new networking concepts such as virtualization and the softwarization of functionalities. For a successful implementation of the smart grid system, one needs to seamlessly integrate these new concepts in to the existing grid system with ICT.

Moving to the smart grid requires a carrier‐grade communication network, which enables scalability, self‐healing, and organizing (Nakata, 2016), quality of service, management and monitoring, and security (confidentiality, integrity, and availability). Smart Grid is large scale, heterogeneous and distributed system. In Smart Grid, power distribution networks have large number of Distributed Energy Resources (DER) which are enabled by the new energy market where any consumer can become a producer of power. Information communication infrastructure should be enhanced matching these technological challenges of the Smart Grid.

In the smart grid, large data collection is required to make actionable smart decisions for the dynamic control of the intelligent devices. The new generation of devices such as phasor measurement units (PMUs) measure power quality with high accuracy. The measurements are frequent, resulting in up to hundreds of measurements per second, collecting large amounts of data.

6.3.1 Software Defined Networking (SDN) as Next‐Generation Software‐Centric Approach to Telecommunications Networks

The traditional communication networks are already quite complex and costly in operations and management. They are based on purpose‐built hardware for each network function and use technology‐specific interfaces. These issues and difficulties have motivated many researchers and industry toward new concepts and architectures. A relatively new paradigm change in communications networking seems to be taking the lead. Many startup companies, incumbent communication and networking equipment vendors, and medium and large network operators and service providers have invested both time and money in these new ways of designing and managing communications networks, called software‐defined networking (SDN) and network function virtualization (NFV) as their next‐generation software‐centric approach to networking.

When it first started, there were multiple definitions of it offered by different organizations. Recently, these definitions have been converging based on some common points. Among all the organizations, the Open Networking Forum (ONF) has been one of the main organization behind SDN. They organize the education and implementation of this new paradigm and have working groups to make recommendations for different aspects of it. SDN is a programmable approach to designing, building, and managing reliable networks (Abhishek, 2016). It decouples network control from forwarding in network devices and offloads its functions to logically centralized SDN controller software, as seen in Figure 6.2.

The replacement of the control intelligence into logically centralized SDN controller(s) would capture the global view of the entire network and provide vendor‐independent control to the day‐to‐day operators of the network. An immediate benefit is obvious simplification, which leads to cost reduction for network devices. Operating SDN‐capable networks will be less complex. Network administrators can programmatically configure the entire network at the SDN control layer, instead of having to manually integrate all configurations in to devices scattered around the network. The SDN controller will provide the common network services such as routing, access control, dynamic bandwidth management, QoS, storage optimization, and policy management to different application‐specific functions through open northbound APIs. SDN's unified control plane allows network abstraction and lets these applications be implemented easily and efficiently with required customization and optimization. The expectation is that these abstractions and simplifications will create a fruitful environment for innovation as turning up new services would be an easy task but needs to be invented. The same unified control plane facilitates the creation of a virtualized network and the slicing of pieces of the physical network or any other hardware source for specific functions and their virtualization, called virtualized functions. In this context combined with IT virtualization in data centers, network function virtualization (NFV) has been proposed and supported by major network operators and service providers as a platform to host and operate network services as virtual network functions (VNFs). There have been studies suggesting that NFV results in considerable cost efficiencies in both Capex and Opex (Yu, 2015; Cerroni; 2014).

Flexibility is another benefit that SDN with NFV offers; mix‐and‐match solutions from different vendors would be accomplished easily with SDN. A summary of an incomplete list on how SDN can benefit current network is tabulated in Figure 6.3 (Fujitsu, 2015).

6.3.2 Adaptation of SDN for Smart Grid and City

There has been very focused research on alternative next‐generation networking infrastructures and architectures such as SDN/NFV producing encouraging results; however, these are unlikely to be deployed by utility companies and distribution system operators without an implementation and demonstration of a successful large‐scale prototype under real‐world conditions to build smart cities.

Another reason for the slow pace of adopting these existing new communications concepts and prototypes into the real production network of any utility or energy company is the regulations and hefty fees enforced by NERC (North American Electric Reliability Corporation) for each day that the utility is in violation of standards. This is an understandable risk to be avoided because utilities networks are part of the critical infrastructure of smart cities.

The pathway to the adoption of these new concepts and architectures for smart grid and smart cities should involve the following steps:

1. idea generation on how SDN/NFV can be used to address the problems of utilities networks and smart city applications in general;
2. proof of concept in a lab environment;
3. testing with simulators/emulators;
4. testing under real conditions in a sliced isolated real network in field trials; and
5. Approval of results and migration to production network in incremental phases.

6.3.3 Opportunities for SDN in Smart Grid

Current challenges and push for modernization require open standards and non‐proprietary solutions to mitigate the risk of rapid technology shifts and to reduce high OPEX cost. Following this trend, SDN and NFV not only provide solutions to some of the challenges of the utilities and energy sector for their networks but could also be a solid option for future‐proof modernization including the smart grid.

SDN, as defined, provides a global end‐to‐end view of the network; therefore, with this new approach, the complexity of multiple domains and networks can be dealt with. SDN adopts mostly open standards and introduces technology abstraction, which provides a vendor‐agnostic approach to configuring and maintaining various types of network elements. Hardware virtualization is one of the goals of SDN and is used for piecing up a physical network into virtual slices for ease of managing different networks while using resources efficiently. It will help connect multiple data centers in different physical locations and make their use efficient. Programmatic network control of SDNs serve for service abstraction and will simplify all configuration activities. Due to its holistic view of networks, an SDN network will be able to have a better control over delay and jitter in the network. This will help teleprotection of control traffic including SCADA. The bandwidth‐on‐demand feature of SDNs will solve the problem of elastic bandwidth need for new and emerging apps in the smart grid. This will also create an opportunity to increase revenue through accelerated service velocity in cases where the utility also serves as a service provider in the coverage area. It has been observed that more and more utility companies serve as Internet service providers in rural areas.

In this new networking infrastructure (see Figure 6.4), the utility‐specific apps on the applications layer will have the opportunity to blossom. These apps could facilitate power/energy peak‐shaving, analytics of system and customer data, M2M communication, fast restoration from a failure, better management of heterogeneous network devices, rapid diagnostics on large inventory of network devices, distributed automation enablement, AMI network monitoring, and many more one could think of.

SDN for resiliency and protection in smart grids: The smart grid maintains multiple large‐scale heterogeneous communication networks to deliver services. One challenge is automatic fast recovery when failures occur by manmade or natural causes. There have been studies that demonstrate self‐healing capability to the failures of links for inter‐substation communication by using SDN architecture (Aydeger, 2016; Sydney2014; Ren, 2016; Ghosh 2016). All these failure recoveries are possible using the SDNs feature of having overall topology view of the entire network. To support multi‐network fast recovery, there are proposed algorithms that are integrated into the SDN control layer (Dorsch, 2014). The algorithm proactively configures alternate routes for all possible failure scenarios and coordinates with SDN controller's monitoring function to be notified at a port status change (happens at a failure). Then, automatic switching takes place, and traffic start flowing on a re‐routed path. There are some studies reporting improved recovery times compared to MPLS‐based techniques (Aydeger, 2016). In addition to link failures, self‐recoveries from node failures have been also reported by using the SDN architecture (Zhang, 2016).

Simple management: The simplicity of the management is a direct product of the separation of the control layer from the device infrastructure layer of the SDN architecture (see SDN architecture in Figure 6.2). An SDN controller at the control layer collects all topology information from devices and makes the collective management decision for the entire network. An example implementation of this management simplicity for the smart grid has been reported by Kim (2015) for an AMI application. AMI meters need forwarding information for the metered data they receive. In an SDN‐based architecture, the SDN controller pushes the forwarding policy for all AMI devices in the data/infrastructure level, so that AMI meters do not carry the burden of making their own forwarding decisions; this reduces the computational burden on AMI devices as well as the topological updates AMI devices needed to receive to make those decisions.

Interoperability and programmatic interfaces: The SDN architecture offers standardized interfaces to be used seamlessly for different communications devices at the infrastructure layer. The northbound interfaces are used for the communication between the SDN controller and applications running for different network services. Due to the variety of such applications, there is no single interface for northbound. In fact, there are expected multiples of interfaces in addition to NETCONF and REST. The southbound interface is used for the communication between the SDN controller and the network switches and routers. The interface facilitates the dynamic control of switches and routers according to the network traffic demand by the SDN controller. The OpenFlow is the first and most used southbound interface in the field. Regardless of the vendors and models, these devices could be performing different functionalities as a physical device or implemented in software as a virtual device. All of the information provided by such devices can be transparently interpreted by the control layer and used by the service provider and/or network operator in an interoperable way; no special control stack should be maintained at the control level by the operator. This would give great flexibility to the smart grid operators for all different kinds of devices they need to have in their communications infrastructure. The SDN architecture offers the flexibility of adopting any change in the middleware control layer by the way it is designed based on software. Considering the rapidly changing protocols and new interfaces the smart grid needs in its growth cycle, this feature of SDN could be a good asset.

Network awareness: With SDN, all devices in the infrastructure layer may provide constant feedback to the controller in the control layer. The operator can define the network events based on the collective information, such as low performance, congestion, large delays, etc. These events could dynamically trigger functions such as load balancing, re‐routing, and new arrangements of QoS parameters. The direct application of this in the smart grid could be the readjustments of AMI and IED devices in terms of reading frequencies and reconfigurations. These would not require any human interventions and eliminate any fat‐finger problems.

Hardware as commodity: Currently, there are different types of equipment from multiple vendors in the traditional network architecture requiring frequent installs for updates and bug fixes. With the introduction of the new SDN architecture, hardware becomes a commodity, and types and vendors will not make a difference at the control layer. All changes and updates will be done at the control layer and pushed to all hardware at once by using the standardized interfaces.

Load balancing and quality of service: SDN offers load balancing based on flow information that is shared between the controller and application profiles. The traditional packet‐based load balancing cannot relate the load balancing functionality to the applications through service profiles. This capability makes the new SDN‐based network more application aware and able to prioritize flows for more efficient resource use. This feature of the SDN architecture could be quite useful to secure the minimum bandwidth and resource levels for high‐priority information flows such as power‐quality event warning messages so that these types of emergency event handlings can be prioritized. There are some specific studies that report such benefits of flow‐based priority differentiation in smart grid applications (Sydney et al., 2014). Flow information that is used for load balancing in the SDN architecture can be easily leveraged for QoS and DiffServ applications. Once flow is recognized and prioritized by the SDN controller, depending on the arrival rate and delay requirements of the flow, appropriate queues can be modelled to accommodate any QoS mechanism in a dynamic manner. This flexibility and control in terms of QoS are quite useful in smart grid applications. There are proposals on heuristic methods for smart grid QoS applications where smart grid–specific AMI flows can be treated differently based on the application service profiles (Qin et al., 2014).

Security: A lot of attacks target critical infrastructures including utilities communications networks. More concerning is the fact that these types of activities are increasing, which justifies the ongoing awareness in and around the security of critical infrastructure networks and assets. These networks require large security coverage due to their footprint with multiple domains and numerous end points (Cankaya, 2015). Each end point counts for a potential point of attack and increases the risk and vulnerability of the system. There are many guidelines, best practices, and standards for security of utilities and energy sector administered by multiple organizations.

For a long time, we have been using the same technologies and practices for security, which have been strictly defense oriented. The “block all” vision that we had has been forcing us to drop a firewall appliance wherever the function is needed in the network. SDN has triggered a change for moving away from this trend. It can enable security developers to implement more innovative solutions for perceived threats as it provides better visibility of the entire network for both control and data. With logically centralized control, the confusion about where to place a security appliance will come to an end. SDN's security‐enabled infrastructure allows network security administrators to route all traffic to a logically central firewall that can virtually be present wherever and whenever needed in the network (Ghosh et al., 2016). In addition to conventional brute‐force techniques, the change will feed innovation for intelligent security applications such as smarter quarantine systems, faster emergency broadcasts, advanced Honeynets, and context‐aware detection algorithms.

In general SDN can provide security based on software without a specialized hardware. There are proposals to use role‐based access control systems in the SDN architecture (Pfeiffenberger and Du, 2014), where an SDN controller uses digital signature with shared keys for all control packets. Kloti et al. (2013) provide a good analysis for open‐flow‐based security for an SDN architected network and provides recommendations on security issues within OpenFlow deployments in SDN architectures.

Flexibility in network routing: An SDN centralized controller can continuously keep track of all network devices and maintain the topology of the network. Using this topology, it can also construct a routing information base that can compute any route with the network, instead of letting devices run their own copy of the distributed routing algorithms to build the routes. In this case, if a device needs a route, it receives the direction of the packet to move in terms of the forwarding information from the SDN controller. Dorsch et al. (2014) proposed a purpose‐built SDN architecture for smart grids, called SDN4SamrtGrid. In this new SDN architecture, the topology and the routes between the smart grid devices are constructed by the centralized SDN controller, and the forwarding information is disseminated to all required devices for next‐hop decisions. The article also claims a superior performance of SDN4SMartGrid routing to the traditional open shortest path first (OSPF) routing.

Ease of deploying new services and applications: The SDN architecture makes the new service and application deployment relatively easy. Service providers and operators have the direct control of the SDN controller. The business case for a new service and/or application starts with simply a policy development for the new service. After the policy has been developed, the policy needs to be provisioned directly to the SDN controller without any hardware dependency in terms of installation and configuration on the infrastructure. The controller layer then monitors and manages the service continuously and decides to make changes only on the control layer to be pushed to the devices.

6.4.1 Network Virtualization

Network virtualization (NV) is a technique to create an independent isolated instance of a network service, called virtual network (VN), over a shared physical network infrastructure with customizable performance levels. A VN is composed of virtual nodes and links that form a virtual topology of the VN. A virtual node is mapped to a physical node; however, a virtual link can be mapped to a section of the physical network that may include multiple physical links and nodes. The owner of the VN can choose to use any protocol, mechanism, and/or control and management planes in this virtual network. The benefits of VNs are many. First, multiple VNs can share the same physical network infrastructure or multiple heterogeneous physical network infrastructures without functionally affecting each other. VNs are also isolated from each other meaning that any failure, misconfiguration, and security threats in one VN should not replicate themselves to the others. Another motivation for VNs is to ameliorate heterogeneity, which complicates the management and service delivery. One is the heterogeneity of the underlying network technologies. There may be WDM, Ethernet, and/or wireless technologies used at the physical infrastructure layer together. A VN is, by definition, transparent to these multiple technologies; only the topological connectivity is important to the administrator of the specific VN. Similarly, multiple VNs should be able to run different protocol stacks and deliver different services. In the smart grid, the amount of information being exchanged is quite large compared to the traditional grid. With the open platform that the smart grid provides, the sources of information have also been increasing. Power generation equipment feeds information for their status and load conditions from subscriber sites. All smart sensors monitoring transmission lines, substations, and transformers send information to control centers. Advanced meters are read frequently. Large capacity power storages at subscriber generation sites including electric vehicle batteries need to communicate their status to the controllers. Not only the count and size of these information sources but also where they communicate to matter. There are also different applications, mainly control and operations applications and data collection applications. The requirements for these applications are different. Control and operations applications are small in data volume but require real‐time communication with limits on delay and availability; however, data collection may be in large bursts but do not have strict requirements on delay; however, they may benefit from high throughput.

Information exchange occurs on multiple physical network technologies, i.e., heterogeneous networks. Optical networking technology is used in the transmission part of the grid to connect substations and head‐end power stations in a wide area. Power line communication (PLC) and wireless mesh (WM) are used in the distribution part of the grid due to the favorable cost point and flexibility of these technologies.

In comparison the requirements of the applications, optical transmission has reliability, low latency, and high data volume capabilities; however, it does not cover the distribution part of the neighborhood area and home networks due to high cost and ROI issues. On the other hand, PLC and WM are ubiquitous in home and neighborhood areas but don't have the robustness and reliability by themselves.

Because the smart grid delivers very diverse applications with a different set of requirements over heterogeneous networks and technologies, one can argue that NV is quite applicable to smart grid communication with considerable benefits. For example, a specific application can run on a tailored VN without affecting any other application running on different VNs. The main challenge in NV in the smart grid, like any other implementation of the NV, is the mapping of virtual networks to physical network. There have been some proposals for the mapping problem (Cai et al., 2010; Zhu et al., 2006; Chowdhury et al., 2009; Lischka and Karl, 2009).

In the smart grid, there are even multiple physical networks, i.e., PLS, WM, optical, for this mapping exercise. Houidi et al. (2011) report on mapping to multiple physical networks. Specifically for the smart grid, Pin et al. (2014) implement NV for smart grid applications by considering WM and PLC as underlying two physical networks and report benefits. The study provisions two VNs, one for real‐time applications where latency and reliability are priority and the other is a data collection application where the throughput is more important. This multi‐objective optimization lands itself into NP‐hard problem set, and the authors propose a suboptimal heuristic solution to map these application‐specific VNs to PLC and WM physical network infrastructures. They report considerable improvement in higher reliability by mapping logical links into diverse physical links from both the PLC and WM infrastructures. A similar diversification has been employed to increase throughput for data collection application VNs.

6.4.2 Network Function Virtualization

In traditional methods, network functions have been delivered by hardware products and devices; for example, various vendors make routers, firewalls, switches, load balancers, etc., for specific network functions and to be placed and installed at certain physical locations in networks. This traditional way of accomplishing network functions has been challenged by a new concept called network function virtualization (NFV) supported by the SDN framework. Other than being just a concept, the number of successful implementations in data center networks, wireless backhaul networks, and content delivery networks has been in the rise along with obvious benefits (Network Function Virtualisation, 2013; Chiosi et al., 2012).

The NFV framework has an extended coverage that includes NV and softwarization of functions (implementation of functions in software) to deliver network services. This will liberate offering services from strict dependency on purpose‐built hardware devices in the traditional networking. The trend is to load the special functionalities to software and run the software on commodity compute, storage, and network hardware resources. The benefit is flexibility and agility in offering new and more complex services by using virtualized function chains. In this context, the coordination and life cycle management of VNFs are done by the VNF orchestrator/manager. The responsibility includes creating an instance of VNF and monitoring its performance and termination.

Niedermeier and Meer (2016) use NFV‐based virtualization to implement virtualized advanced meter infrastructure (AMI) that reports increased reliability and availability with reduced cost, which were pain points for the smart grid AMI application without the implementation of virtualization.