Electric energy is currently an essential resource all around the world. With the increase in the use of new technologies in all sectors of human activity, it is easy to predict that the consumption of this type of energy will grow considerably in the near future. For this reason, great research and development efforts have been made, with the aim of improving the generation processes, transport networks and storage systems for this energy. The smart grid networks are the result of the work carried out to obtain improvements in the management, operation and maintenance of the transport infrastructure, as well as in the efficiency with which the energy is used. Their main objective could then be considered as achieving the best use of electrical energy through an improvement in the management and maintenance of the energy sources and the transport infrastructure. At the same time, new services are offered to both supplying companies and consumers.
With these objectives in mind, one of the main advances is being made in the improvement of the data networks associated with the electricity transport infrastructures. These data networks are responsible for carrying and delivering all the control, management, maintenance, and security information of the electricity network infrastructure, as well as the applications that allow a better management of the available resources. The data networks are therefore a fundamental part of the smart grid, and so, their reliability, availability, and security have to be guaranteed in all situations, taking into account that they can stop providing correctly their service due to intrinsic (hardware, software, communications protocols, etc.) or extrinsic (weather conditions, malicious agents, terrorist attacks, etc.) failures [1
The smart grid data communication network is made up of several subnetworks. The different Smart Meters (SM), devices, and other utilities present inside the customers’ homes are interconnected by the Home Area Network (HAN). These HANs are in turn interconnected through the Neighborhood Area Network (NAN), and finally, the information can reach the control centers through a Wide Area Network (WAN). To implement all these subnetworks, different technologies can be chosen. Selectable technology standards for HANs can be, among others, IEEE 802.15.4 (Low-Rate Wireless Personal Area Networks (LR-WPAN)) or IEEE 802.11 (Wireless Local Area Networks (WLAN)). For NANs, Power Line Communication (PLC) technologies or standards such as IEEE 802.15.4g or IEEE 802.11s (Wireless Mesh Networks) can be considered.
Guaranteeing that the quality of service offered is that required by every application, while maintaining a high efficiency in the use of resources, is one of the most important issues to be taken into account both during the planning process and during the maintenance and operation of the network. It is therefore essential to have deep knowledge of the services that will be provided. These issues are especially important when working with smart grids, given the critical importance of the power grid infrastructure. The offered services belong to very different types, and therefore, their service quality requirements are also different [2
]. Generally speaking, most smart grid applications have strong security and reliability requirements. However, their bandwidth needs, as well as their behavior in high packet losses or high latency situations can be very different. Thus, some applications, such as Substation Automation Systems (SAS) or overhead transmission line monitoring, present very strict requirements in terms of latency, but they are not as demanding in terms of bandwidth. On the other hand, demand response management or Advanced Metering Infrastructure (AMI) generally consumes more bandwidth, but can allow greater delays.
In the context of WLANs standardized by the IEEE, the proposal for multi-hop mesh networks was published in 2011 as Amendment Number 10 to the 2007 general standard, with the name of IEEE 802.11s [3
]. In the revision of the general standard published in 2012, as well as in the current revision [4
], mesh networks have been directly incorporated, although a large number of researchers continue to refer to them as IEEE 802.11s mesh networks. The main differentiating characteristic of this type of networks is that, from the upper layer point of view, all the stations appear as connected at the MAC level, although they might not be within the range of coverage. To make this possible, a layer 2 path search mechanism called the Hybrid Wireless Mesh Protocol (HWMP) was designed.
In this article, some modifications on the basic mechanisms and protocols used by the IEEE 802.11 mesh networks are proposed, in order to improve their performance when using this technology as the implementation of the smart grid NANs. Mainly, a new multi-path mechanism is proposed and implemented in conjunction with a multi-channel allocation of the different available paths. These paths are assigned differently according to the quality of service demanded by every traffic. With this strategy, we intend to take better advantage of the available network resources, guaranteeing an adequate service to as much traffic as possible. In this way, the supplier companies can obtain greater benefits by being able to provide service to a higher number of clients or applications, while at the same time, they can offer more competitive prices to their customers. The study and comparison of the benefits obtained is based on simulations carried out with the ns-3 network simulator [5
]. Thanks to the improvements obtained, especially in situations of high load, NAN networks could offer service to a greater volume of traffic and also allow a correct and differentiated quality of service for each application. In this way, new development challenges are created for both device manufacturers and electricity supplier companies, which will be able to offer new and better services to their customers. From our point of view, these new services should have a major impact on greater efficiency in the use of electricity and a faster and improved reaction in front of emergency situations.
Smart grid networks have attracted the attention of numerous researchers in recent years. Among these investigations, several proposals have been presented in order to improve the benefits offered by the NANs, where both wired and wireless technologies have been taken into account [6
]. Within the wired technologies, PLC stands out especially in this environment due to its ability to use the existing infrastructures. However, the available bandwidth with this technology is quite limited, and it also presents drawbacks when data signals must pass through electrical transformers. When the number of nodes in the network grows, as well as the bandwidth needs, other technologies must be considered. In this sense, wireless networks in general [7
] and wireless multi-hop networks in particular [9
] present a series of advantages that make them ideal candidates. For instance, they do not require previous infrastructures, and their bandwidths are constantly increasing. Besides, they have a great flexibility to modify the network topology and to take advantage of multi-channel and multi-path mechanisms that increase their performance in terms of, among others, availability, packet delivery ratio or network transit time. For these multi-hop wireless networks, a new and precise analytical model, which takes into account the hidden nodes problem, has been presented in [11
], an enhancement of the Optimized Link State Protocol (OLSR) in order to satisfy the required level of reliability in NANs is presented. The possibility of offering an adapted quality of service to the different data traffics transmitted through the network is taken into account. To this end, the authors proposed the use of a combination of different basic metrics: Expected Transmission Count (ETX), Minimum Delay (MD) and Minimum Loss (ML). They chose Relevant Link Metric Types (RLMTs) for each application (traffic type), assigned different weights to each of them, and used a pruning technique to reduce the number of considered paths to a given destination. The best link to send each traffic was then calculated by means of an AHP (Analytical Hierarchy Process) algorithm. The proposal was evaluated by means of ns-2 simulations, over a usual network environment consisting of a grid of smart meters transmitting (receiving) information to (from) a data concentrator and taking into account four basic CBR traffic types. Moreover, the topology was modified by increasing the number of smart meters (from 25 to 64) and changing the data concentrator position. The network performance was measured in terms of dropped packets, packet delivery ratio and average delay, showing a better behavior when compared to a basic OLSR implementation. The same authors previously presented in [13
] a performance evaluation and comparison of the OLSR and HWMP (IEEE 802.11s) routing protocols, together with a classification of the main AMI application traffic.
A multi-gate communication network, based on IEEE 802.11s, was proposed in [7
] for smart grids. The authors took into account the possibility of having more than one node acting as a gateway, together with a real-time traffic scheduling and a multi-channel-aided routing protocol. Besides, the authors proposed a heuristic backpressure scheme, where every node evaluated the state of its neighbors before selecting one of them as the best next hop, which implies that some information (the backpressure metric) must be periodically exchanged between nodes. Otherwise, to avoid loop problems, a hop-count limit was imposed on the data packets. Besides, in order to reduce the effect of co-channel interference, a multi-channel protocol was also introduced. To evaluate the proposals, three simulation scenarios were taken into account: (a) three separated sub-networks where every one had its own gateway, (b) a multi-gateway network where the three previous networks shared their three gateways and where the nodes were uniformly distributed, and (c) the previous configuration, but with an asymmetrical distribution of the nodes. The results showed the better behavior offered by the proposed backpressure scheme in terms of overall throughput, average end-to-end delay, and adaptation to malfunctioning nodes. On the other hand, the benefits of the multi-channel protocol were also clearly shown.
A cross-layer mechanism that combines information from the physical, MAC, and network layers was presented in [14
]. Based on that mechanism, the authors defined a new routing metric (Expected Path Throughput (EPT)) and a distributed routing protocol, which was evaluated with the help of the ns-2 simulator. The results showed the good behavior of the proposal when compared with other classical metrics and protocols.
], the authors proposed the HWMP-NQ protocol, a modification of HWMP to ensure the Needed Quality of Service (QoS) of several smart grid traffic types. To this end, the airtime link metric was modified by considering the packet size and the transmission rate. However, the needed number of channel measurement could be excessively increased. To avoid this, a frame error rate computing algorithm based on a single measurement was also proposed. Besides, the benefits provided by a multi-gateway backup routing scheme were also analyzed. Moreover, to reduce the routing overhead in case of link failures, a modification of the path error mechanism was introduced. To evaluate the benefits of their proposals, the authors built classical NAN grid topologies with the help of the ns-3 simulator and ran multiple simulations to measure the average throughput, packet delivery ratio, end-to-end delay, and routing control information overhead. The results showed the benefits of the multi-gate routing scheme presented in [7
] and the HWMP-NQ protocol, in front of the classical HWMP implementation, for different NAN grid sizes (from 9–64 smart meters). What is more, the influence of the node failure rate was also studied, showing that the performance improvements obtained with the authors’ proposals increased when that failure rate was higher.
In order to improve the network throughput and reliability, another modification of the airtime link metric calculation method was presented in [9
]. One of the contributions of this work was to give more importance to the upstream transmission status (from smart meters to the concentrator), since most data were transmitted in this direction. Besides, a modification of the path selection mechanism was provided to avoid the classical problem of route fluctuation. With this modification, not only the current airtime link metric value, but also its variations were taken into account to select (or not) a new route between two network nodes. ns-3 simulations were presented to show the achieved benefits in terms of packet delivery ratio, end-to-end delay and data retransmission count. The results also highlighted the need for congestion control mechanisms when the network size was increased.
Some of the same authors of [9
] performed in [10
] a study of the HWMP routing protocol, with the goal of identifying its weakness, both from the HWMP protocol itself (route instability and route recovery) and from the integration with smart grid networks (oversimplified calculation of airtime link metric and the need for traffic differentiation). Here, a modification of the airtime link metric computation was also proposed, as well as a proposal for the path selection mechanism. Besides, to get a better performance in terms of packet losses, reserve routes were stored in the network nodes. This idea gave rise also to a reduction in the traffic management traffic needed when a path was broken. Moreover, in order to provide a better quality of service to some applications, a delay-tolerant traffic management method based on the concept of delay-tolerant networking was proposed. The improvements obtained with the application of these new solutions to the protocol, called HWMP-reliability enhancement (HWMP-RE), were checked and shown by means of ns-3 simulations. Grid topologies were considered, from 9—49 nodes, where every node generated traffic (belonging to seven different applications) to two root mesh stations (gateways). HWMP-RE was compared with the basic HWMP and with the previous proposal in [9
], showing a better behavior in terms of packet delivery ratio, end-to-end delay, number of PERR/PREQ generations, throughput and reliability.
Other proposals based on the modification of the HWMP metric can be found in [16
]. In [16
], a QoS-aware and load-balance routing scheme was proposed, which was complemented with an EDCA-based adaptive priority adjustment scheme, with the goal of satisfying the QoS requirements of different NAN applications. The modification proposed for the airtime link metric consisted of including the packet size and calculating the frame error rate separately for the different NAN applications. Besides, to avoid congested paths, the queuing delay was also added to the metric. What is more, the dynamic adjustment of the packet priority allowed a better resources utilization under low load conditions and improved the reliability under heavy load conditions. ns-3 simulations were carried out to evaluate the obtained performance, which showed an increase of both the packet delivery ratio and the throughput, as well as a reduction of the average end-to-end delay. The network scenario consisted of a grid topology where the number of nodes varied between nine and 64.
On the other hand, the metric modification proposed in [17
] (Interference-Aware Expected Transmission Time (IAETT)) was oriented to reduce the impact caused by inter- and intra-flow electromagnetic interferences. Besides, traffic differentiation was also considered. Based on this metric, an interference-aware QoS routing protocol was proposed and evaluated. The performance evaluation was carried out again by ns-3 simulations, over a scenario consisting of 100 nodes arranged in a 10 × 10 regular grid, where both the gateways (nine nodes) and the traffic generating nodes were randomly chosen. Results showed the improvements obtained in terms of average end-to-end delay and packet delivery ratio.
As already mentioned, this paper presents a new proposal for improving performance in smart grid NANs when using IEEE 802.11 mesh network technology. Although the modification of the routing metrics is a good idea to differentiate the service offered to different traffic in the network, we preferred to maintain the basic airtime link metric and focused our efforts on the modification of the mechanism used by the HWMP protocol for the selection of the most appropriate path each time a data packet must be (re)transmitted. By its own nature, the default metric informs about the congestion state of the different network areas, which is the most relevant measure for our approach. Moreover, it is important to keep in mind that working with more complicated metrics usually leads to higher CPU and memory requirements in the network nodes, as well as to protocols that generate more network control traffic. In this way, a modification of the HMWP protocol is proposed and implemented, to allow an efficient selection of paths among multiple possibilities, depending on the service quality needs of the different traffic flows. The proposed mechanism is complemented with the assignment of different frequency channels to each available path. In addition, to avoid packet losses due to the formation of unwanted loops, the proposed technique is combined with a criterion of the minimum number of hops when choosing the paths. This technique reduces the number of selectable paths, but avoids the need to use packet hop counters (which are used by the nodes to discard packets after a given number of hops, with the added disadvantage of using network resources for a certain number of retransmissions in a completely useless way). On the other hand, as will be seen in the Results Section, we considered it of great importance to provide not only the average values of the performance parameters under study, since this way, the real network performance is not obtained and would probably lead us to an erroneous network planning.
The rest of the paper is organized as follows. The next section presents the modifications proposed for the HWMP protocol. Details about the multi-path and multi-channel mechanisms implementation are provided, as well as the route selection and assignment algorithms. Section 3
presents and analyzes the results obtained through the simulations, and finally, the conclusions, as well as the future lines of our research are summarized in Section 4