A Realistic ICT Network Design and Implementation in the Neighbourhood Area of the Smart Grid ()
1. Introduction
In recent years the conventional power grid faces difficulties due to the lack of intelligence, energy crisis and environmental degradation. Consideration of these issues led to the birth of the Smart Grid. According to the US Department of Energy (DOE) [1] desirable characteristics of the Smart Grid have been identified as; self-healing, consumer friendly, reliable with good power quality, and resistant to cyber-attack. It should also be able to accommodate all the storage and generation options and facilitate new service and markets. The Smart Grid proposes to accomplish the mentioned goals by incorporating Information and Communication Technology (ICT) in the power network to bring more intelligence to the Grid. Intelligence here addresses the ability to monitor and control a range of industrial appliances and functional components to optimize energy generation and consumption [2].
The Smart Grid can be viewed as resembling the internet [3]. Instead of uploading and downloading data, customers upload and download electricity. Rather than using a modem representing the data usage, the customers have smart meters indicating the electricity they use or generate and the price accordingly.
As mentioned earlier, the Smart Grid is the integration of a power network and a communication network to meet the future needs of energy. A power Grid, responsible for transmitting the energy from generation point to the customers, consists of three layers; generation, transmission, and distribution. Similarly, the current proposal for the communication network, responsible for collecting and routing data between different sectors in power Grid, has three layers; Wide Area Network (WAN), Neighbourhood Area Network (NAN), and Home Area Network (HAN) [4]. The WAN covers the generation, transmission, and distribution network of the power Grid. The NAN and HAN covers the rest of the distribution grid ending in the consumer premises.
Until now most research has focused on monitoring and controlling at the WAN and HAN levels. By contrast we have particularly investigated the monitoring of the NAN, and describe how it is being implemented in a test bed of monitoring devices installed in a 6.6 kV sub-Grid on a university campus. Until now it is not understood how a NAN functions as a Grid, thus NANs are controlled from higher levels. Our research aims to provide such monitoring architecture at the NAN level of the Grid.
In this paper we have considered two related architecture within the ICT in frastructure, namely communication network architecture and a software architecture. The former encompasses the networking aspects of the NAN and considers energy efficiency of the communication system. The proposed architecture is based on hybrid communication technologies which integrate sensing, computation and decision-making to enable control and prediction of the future state of the sub-Grid in the realtime. A Wireless Sensor Network (WSN) is considered as an essential component of the monitoring function. The WSN is responsible for monitoring and collecting real-time data from the field. It will send live data to a NAN Control Unit (NCU) to provide more accurate prediction. These sensors are envisaged to be battery powered and have power shortage problems. Our communication network design is targeted towards the solution of two key problems. The first is the energy constraint problem of general WSN. The second problem addresses the weaknesses of a centralized architecture where data collected from the entire system are stored at a sub-Grid database, where controls are applied.
The second architecture is the software architecture, this describes how we will process and analyse data gathered by the communication system. This is the architecture that will be responsible for the control of the NAN and for integrating different NANs to provide information to higher layers of the Grid.
Three key points of our approach are: firstly we distinguish the communication architecture and software architecture. Secondly we address how monitoring and data processing needs in a NAN could be met. Thirdly we are deploying our architecture in a real NAN environment namely The University of Manchester campus sub-Grid. This is similar to the size and complexity of a NAN in an urban distribution Grid.
The remainder of this paper is organized as follows: Section 2 presents the background research. Section 3 presents our proposed communication architecture and issues of its deployment on the actual experimental test bed. Section 4 surveys available communication technologies and describes our choices. Section 5 discusses the software architecture and analyses its response in a series of emergency scenarios. Finally, section 6 summarises the current status of our work.
2. Background
In this section, we discuss the available intelligent monitoring in place and general issues of what needs to be added to the power Grid to convert it to the Smart Grid.
We have applied principles from the theory of distributed computing systems, such as locality (attempting to process information locally as much as possible), scalability, fault tolerance, and minimisation of communication. We have utilised work on monitoring and control of other types of distributed networks, such as water distribution Grids [5-7]. The water distribution Grid in the UK has a modular structure, being divided into District Metered Areas (DMAs) that can be controlled in isolation from the rest of the network. A DMA is similar to the size of a NAN in the electrical distribution grid. An example of the water grid that uses sensor data for monitoring purpose is the Neptune project1. Our work extends this approach to the electrical Grid where the monitoring systems take account of much more rapid changes in the state of the network (seconds rather than hours).
Another similar study [8] describes the flood monitoring system which allows the integration of the WSN and the remote fixed-network for computationally-intensive tasks and as well as performing on-site grid computation to support timely prediction and flood warning. This system is able to adapt to different network topology and switch from low power physical network to high power consumption physical network. These adaptations are based on awareness of data about the internal system. However, monitoring the external environment also offers useful information that can support better adaptation of the system. In our proposed ICT architecture we have utilised both internal information such as electrical attributes and external information such as weather, light, and other environmental data to support better control of the NAN.
In electrical distribution grid, the SCADA (Supervisory Control and Data Acquisition) that provides communication infrastructure across the electrical grid from 11 kV to 132 kV is used as the intelligent monitoring system in place. Yang et al. [9] explained the implementation of SCADA in the Distribution Network Operators (DNOs). This system contains a master terminal and many Remote Telemetry Units (RTUs). The RTUs are responsible to gather network measurements from the grid and transmit commands to control devices, and the master node. The master node is located at the control centre of the DNO and is in charge of processing and storing the received data. There can be a heterogeneous communication channels and various physical medium between the RTUs and the master terminal. In this system the up-to-date data are sent every 10 - 20 seconds to the control centre, which means it cannot provide continues data delivery and real-time applications. Therefore, in our architecture, the substation data are sensed and transmitted 1 to 4 times a second. Moreover, some control functions for DNO management require low latency such as millisecond to several second. While the SCADA supports operations from few seconds to a few hundred second [10]. Additionally, it has been observed that SCADA has limited data update rate from few hundred bits per second to few thousand bits per second [11]. These facts make SCADA inappropriate to support realtime applications and fast acting control functions. The other limitation of this system is that it fails to monitor the whole sub-Grid, and only monitors the critical areas of the network.
Our architecture, like others such as [2,11] uses sensor networks instead of SCADA. We propose to collect three sources of information: from the environment, from the electrical network and from consumption data via metering.
If all this information across the whole Grid is collected at a single database, we face the problem of information overflow. Thus we adapt clustering methods developed in studies of sensor networks [12-14]. Further more, we follow recent studies in the Smart Grids [9,15, 16] in developing distributed control at the NAN level.
Published surveys of available communication technology [17,18], confirmed our initial analysis that it is not feasible to use one single technology for the whole grid and heterogeneous communication technologies should be provided. Since we do not assume that sensors can necessarily be powered from the power network itself energy efficiency is a major driver in the design of our communication architecture. This assumption was confirmed in the practical implementation decisions we made in developing the University campus NAN monitoring.
3. Communication Network Architecture Design
We now analyse the communication architecture in terms of networking aspects and energy efficiency.
3.1. Related Work
In our design, energy efficiency is a prime driver. We have also used a combination of centralized and distributed approach. By contrast [19] proposed the Gossip algorithm which is a distributed approach to provide fault tolerance and guarantee the delivery of the messages. Gossiping can provide robust communication for the subGrids and it is a good choice when the global view of the network is not available or the real-time data is not needed. Therefore it is not the optimal solution for our proposed architecture since we know the topology of our test bed. Knowing the addresses of the devices and using direct communication will prevent hand shaking before establishing connections. This will result in consuming less communication energy. Moreover, time to fulfill thorough message dissemination in Gossip algorithm depends on the graph connectivity, whereas in our architecture we need to support real-time data.
The design of a client server architecture for the Smart Grid using TCP/IP for information transmission has been discussed in [20]. The first difference between this design and our proposed design is that they have not considered the monitoring and control of the NAN in detail. The other main difference is in the design of the server side of the architecture which has a central controller containing a database, whereas we have proposed an architecture utilising distributed data centres communicating in a peer to peer fashion.
Aalamifar et al. [21] use the network simulator-2 to propose a three layer hierarchical communication architecture as an economic architecture for the smart grid utilising power line communication technology. The difference between this work and our proposed architecture is the integration of hierarchical and peer-to-peer architecture which will be discussed later in this section. Moreover, in reality we have experience that a hybrid communication technology (wired and wireless) should be used to improve the reliability of the communication network in the electrical grid.
Another three layer ICT architecture has been proposed [22], which utilities various communication technologies within and between each layer of the architecture. These three layers are AMR networks layer, AMI networks layer, and AMI + networks. Then they have classified the communication network into Consumer Premises Networks (CPNs), Neighbourhood Area Network (NAN), Access Area Network, Backhaul Network, Core and Office Network, and External Access Networks. The main difference between our communication network design and this research is that in this research the substation monitoring is located in the core network, whereas our substation monitoring is located in the NAN which offers information that has not previously been collected. The other difference is that in our architecture we are not only monitoring the electrical attributes which are collected through smart meters and substation monitoring devices, but also we are collecting information about external systems such as environmental data.
Another study [23] suggested three layers architecture for an ICT in the Smart Grid. The first layer is the local area network (WLAN) utilising Wi-Fi to provide communication inside the data centres and between metering devices to the next layer. The second layer is metro network (WMAN) using WiMAX providing the communication between data centres and transmission substations to the utility generators. The final layer is the wide area network (WAN) utilising fibre optic between utility generators and the data centre utility control. The energy efficiency of the communication system is not considered in this research.
A recent work published in 2013 proposes a hybrid network architecture integrating a wired infrastructure, WSN, and a PLC for the Smart Grid [24]. It is divided into three subsystem; data acquisition subsystem, communication subsystem, and supervisory control subsystem. The difference between our architecture and other communication architectures in the Smart Grid is that we are particularly linking it to the area in the sub-Grid, where we are concerning about the energy efficiency of the communication system. Although the mentioned research has considered the energy efficiency of the radio transmission in to account by putting the nodes into sleep mode after each transmission, but they did not propose an optimal network topology in terms of energy efficiency [25] and energy efficient communication system [26] which is believed to be a vital requirement for the ICT network of the Smart Grid.
Having studied the available literature on this subject, we noticed that no other research has considered energy efficiency of the communication architecture in detail and have therefore proposed an energy efficient communication architecture. Moreover they have not distinguished the communication architecture from the software architecture. This is important, since each communication architecture can be implemented in different software architectures. Therefore designing and implementing both of these architectures which are part of an integrated system for the NAN in the Smart Grid is necessary.
3.2. Principles of a Communication Architecture
Our architecture is being implemented on the medium voltage power network of the University of Manchester campus which owns its own distribution grid. This allows us to check our design on real equipment, real data, and input from experts in power engineering.
Our architecture (Figure 1) is a modular architecture that combines the peer-to-peer and hierarchical architectures, tailored to hybrid communication technologies for transmitting data. It contains five layers that cooperate to provide four main functions of monitoring, data movement, data storage and control. The meaning of layers here is different to that used in Section 3.1 since all the layers are now within a NAN. The layers refer to different functions in the total information-processing network. The reason that five layers map onto four functions is that the lowest three layers are three different sensor nets (c.f. Section 1) that all represent the same monitoring function. The remaining two layers represent the two functions of data storage and control and the fourth function refers to the connectivity between the layers, thus is not represented as a layer itself. This layered architecture is designed to address modularity, scalability, fault tolerance, energy efficiency and future proofing against changes in networking technology as follows. Modularity: since each layer can be considered as a separate component (module) which are connected together to represent the whole system. Scalability: since we are able to add or remove each of these layers (modules) without affecting the whole architecture. Fault tolerance: we prevent the single point of failure caused by the centralized architecture and move to a more distributed architecture. Energy efficiency: Given that the energy for data transmission is higher than energy for data computation [27], by reducing the transmission range and adding more computation unit such as CHs or other sub-layers to the system we may achieve an energy efficient architecture. Future proofing: the design is conceived at an abstract level independent from particular communication technologies.
This system is designed for the future intelligent power grid, in that it is able to incorporate different control strategies at various level of the NAN. The first three layers are mainly responsible for sensing, measuring and collecting data. The other two layers present the database and control layers responsible for computing, storing, visualising, and controlling the NAN.
3.3. Implementation on a Campus Sub-Grid
Figure 1 also shows the deployment of a communication architecture based on the principles discussed in the previous section. The actual physical distribution of the sensor networks and the substations determines the detail of the communication patterns between the layers. The first layer of the architecture consists of smart meters monitoring system, which are the gateway from the HAN to the NAN and are used to monitor the building level data. These devices are located in all the buildings in our campus test bed, transmitting data every 30 minutes. They are already connected to the power network and transmit data through wired connections to the data base layer. These monitoring devices provide information about power usage and permit the management of the power generation and consumption. These data can be integrated by the real-time energy prices to offer effec-