PHY and MAC Layer Design of Hybrid Spread Spectrum Based Smart Meter Network

Shravan Garlapati , Haris I. Volos , Teja Kuruganti, Michael R. Buehrer , Jeffrey H. Reed
of Electrical & Computer Engineering, Virginia Tech Blacksburg, VA, USA - 24060 {gshra09, hvolos, buehrer, reedjh}@vt.edu Computational Sciences & Engineering Division, Oak Ridge National Laboratory Oak Ridge, TN, USA - 37831 {kurugantipv}@ornl.gov
Dept

AbstractThe smart grid is a combined process of revitalizing the traditional power grid applications and introducing new applications to improve the efciency of power generation, transmission and distribution. This can be achieved by leveraging advanced communication and networking technologies. Therefore the selection of the appropriate communication technology for different smart grid applications has been debated a lot in the recent past. After comparing different possible technologies, a recent research study has arrived at a conclusion that the 3G cellular technology is the right choice for distribution side smart grid applications like smart metering, advanced distribution automation and demand response management system. In this paper, we argue that the current 3G/4G cellular technologies are not an appropriate choice for smart grid distribution applications and propose a Hybrid Spread Spectrum (HSS) based Advanced Metering Infrastructure (AMI) as one of the alternatives to 3G/4G technologies. We present a preliminary PHY and MAC layer design of a HSS based AMI network and evaluate their performance using matlab and NS2 simulations. Also, we propose a time hierarchical scheme that can signicantly reduce the volume of random access trafc generated during blackouts and the delay in power outage reporting.

robust security, low cost of ownership, high performance and high scalability. Although 3G cellular technologies satisfy the requirements of smart grid applications, an analysis on the use of the current state of the art 3G cellular technology for smart grid applications gives an indication that their usage results in data overhead, latency and high power consumption for data transfer (section III). Therefore, instead of using the existing 3G/4G cellular technologies, we propose a HSS based custom smart grid distribution side communication network design that can reduce the overhead, latency and power consumption in data transfer.The following are the main contributions of this paper: 1) We identify some issues/inefciencies of using 3G/4G technologies for distribution side smart grid applications and discuss them in detail. 2) A preliminary PHY and MAC layer design of a HSSbased smart metering network that reduces the overhead and latency in smart meter data collection. 3) Implementation of the proposed PHY and MAC layer in the packet-based network simulator NS-2. There are many issues (see section V) in implementing a new wireless protocol in NS-2 as the current implementation of wireless PHY and MAC layer of NS-2 relies heavily on IEEE 802.11 and IEEE 802.15.4. In order to overcome this problem a generalized architecture is proposed in [2] and was successfully tested using UWB radio simulations for adhoc-networks. We made few modications to the architecture presented in [2] to simulate DSSS and HSS based networks like cellular CDMA and AMI. 4) A Time Hierarchical Scheme (THS) to reduce the high volume of random trafc generated during a large scale power outage. Also THS reduces the latency in power outage reporting and the power consumption of smart meters (relies on battery during a power outage) when compared to cellular CDMA technologies. The remainder of the paper is organized as follows: Section II gives a brief description of the AMI architecture and the data types handled by the AMI network. The main issues of using 3G/4G technologies for distribution side smart grid applications are discussed in Section III. The preliminary PHY and MAC layer design of a HSS-based AMI network is explained in Section IV. In Section V, PHY and MAC layer simulation scenarios and results are discussed. THS is explained in section VI. Energy savings due to THS are discussed in section VII. Section VIII concludes the paper.

I. I NTRODUCTION Transforming the ageing electric power system to smart grid is an active area of research in industry and the Government. The main objective of the smart grid is to improve the efciency of power generation, transmission and distribution and also to improve the stability and the reliability of the grid. In order to achieve this, various processes involved in power generation, transmission, and distribution (GTD) should be armed with advanced sensor technologies, computing, communication and networking capabilities which the power grid has never seen before. A few specic areas in power system where these high speed computational and data transfer abilities can be applied are supervisory control and data acquisition system (SCADA), Advanced distribution automation (DA), phasor measurement unit (PMU), power system state estimation, demand response management system (DRMS) for efcient power management during peak load hours, advanced smart metering infrastructure (AMI) and better Outage management System (OMS) etc. The main focus of this paper is on the distribution side of the smart grid applications i.e. DA, AMI and DRMS. In a recent study, a reliable source compared the key characteristics of different communication and networking technologies like GPRS, RF mesh, PLC and 3G cellular and concluded that 3G cellular technologies provide an advanced and cost effective solution for smart grid communications [1]. The main reason for this is that the 3G cellular gives ubiquitous coverage, high reliability, high capacity and data rates,

DAP DAP

III. 3G/4G I SSUES According to [1], 3G cellular technology seems to be a good choice for smart grid applications. In this section we argue that the CDMA based 3G technologies is not an appropriate choice for smart grid applications. The main issues of using the current state of the art 3G/4G cellular technologies for smart grid applications are as follows: A. TCP/IP data overhead CDMA 2000 and WCDMA both employ a Transmission Control Protocol/Internet Protocol (TCP/IP) based data protocol stack [4]. There are two versions of IP i.e., IPV4 and IPV6. IPV6 is the successor of IPV4. The header overhead of TCP/IPV4 and TCP/IPV6 is 40 bytes and 60 bytes respectively. Nowadays, it is not uncommon to use smart phones to access internet applications like email, web and video streaming. The data transfers involving these applications is generally very high on the order of hundreds of KBytes to MBytes. As a result, the relative overhead (overhead with respect to payload) of TCP/IP packets is around 5% to 10% [5]. But CDMA cellular technologies uses RObust Header Compresssion (ROHC), which renders the overhead to be negligible [6]. Compared with smart phone applications, the data transfer involving smart grid applications is very low i.e., hundreds of Bytes [7]. Even with the use of ROHC, the percentage of TCP/IP overhead for smart grid data is very high because the TCP/IP overhead not only comes from the header size but also from the way TCP operates. TCP uses a 3-way handshake for connection establishment (SYN,SYNC-ACK,ACK) and connection termination (FIN,FIN-ACK,ACK) between a source and a destination. In order to transfer 100 Bytes of data from smart meter to DAP, six additional packets have to be transferred. Using the Wireshark network packet analyzer we found that the TCP/IP connection establishment/termination packet sizes are 48 Bytes (40 Bytes Header + 8 Bytes data) [8]. ROHC reduces the TCP/IP header size from 40 bytes to 4 bytes [6]. Even with the use of ROHC, in order to transfer 100 Bytes of data, one TCP/IP data packet of size 104 Bytes (100 Bytes data + 4 Bytes header) and six TCP/IP connection packets of size 12 Bytes (8 Bytes + 4 Bytes header) are exchanged between source and destination. Assuming a CDMA 2000 PHY layer frame size of 24 Bytes, total 5 data frames and six connection frames are sent i.e. 120% overhead due to connection frames. In this analysis we considered only the transport layer overhead because the overhead associated with the lower layers like MAC and PHY is less and approximately the same for most of the technologies. It is obvious from this simple analysis that the use of TCP/IP for data transfers of the order of few hundreds of bytes (common in smart grid applications [7]) results in very large overhead compared to the payload. Undoubtedly TCP offers reliability in data transfer but the overhead associated with it has to be taken into account while adapting it for an application that involves small size data transfers. Irrespective of the type of the network being designed i.e. wired or wireless network, reducing the data overhead decreases the latency in data transfer from source to destination and it also results in bandwidth savings which further reduces the expenses in system design. TCP/IP based data transfer is appropriate for smart phones because the smart phone applications communicate with TCP/IP servers that are located in different parts of the world. On the other hand the data collected from smart meters is restricted to a local area. As shown in Fig 1, smart meters

DAP - Data Aggregation Point - Smart Meter - Single hop wireless link - wired link Utility/ Control center

Fig. 1.

Advanced Metering Infrastructure

II. P RELIMINARIES A. Advanced Metering Infrastructure AMI is a technology where the traditional electricity meters are going to be replaced by smart meters at the customers site, which give the utilities the ability to monitor and collect information related to the amount of electricity consumed by the customer. In addition to customer power consumption data collection, the communication network set up for AMI can also be used for the other distribution side smart grid applications such as DA, DRMS and OMS. As shown in Fig 1, we assume a two layered communication architecture for the AMI network. The control center collects the data from the smart meters and the Data Aggregation Point (DAP) acts as a relay between the smart meter and the utility. we assume that the AMI employs wireless communication network between the smart meters and the DAP and a wired/wireless network between the DAP and the utility. In the process of data collection, smart meters send the consumers load consumption data to the utility at a preselected time interval. The time interval could be 15 minutes or 30 minutes or 1 hour. The number of smart meters that a DAP handles can vary from as few as 500 to as many as 30,000 [3]. B. Data Types Taking into account all the distribution side smart grid applications, the data handled by the AMI network can be divided into two categories: 1) Scheduled Data: If the time instance at which the data needs to be collected from or sent to the smart meter is known in advance, those data are categorized as scheduled data. Examples of scheduled data are customer power consumption data and DRMS data. 2) Random data: If the time of arrival of the data to the DAP or the smart meter is unknown and occurs at a random time instance, that data are categorized as random data. Power outage data is an example of random data. Power outage is considered as a rare event in most of the developed countries in North America and Europe. On the other hand, power consumption and DRMS data transfer occur periodically with a time interval of 1 hour. Therefore we can consider that the scheduled data contributes to a high percentage (> 90%) of the trafc in the AMI network and the percentage of random trafc is relatively small (< 10%).

send data to or receive queries from a single source i.e. database located at the control center. Therefore IP based communication may not be required for smart meter data transfer. In Fig 1, assuming that there is a database at the DAP, the wireless portion of the network can be considered as a smart metering sensor network. No wireless sensor networks use TCP/IP based stack as it results in data overhead and latency. If the data collected at the DAP has to be sent to any long distant control centre, TCP/IP based stack can be used for the portion of the network between DAP and the control center. Taking the above factors into consideration, it appears that the TCP/IP based 3G cellular technology is not an appropriate choice for smart grid applications. One of the solutions to address the above discussed issues may be to design a custom network for distribution side smart grid applications. Consider the AMI network shown in Fig 1. A high level view of the AMI network gives an impression that its architecture is similar to that of a cellular network. This aspect of AMI network is misleading communication and networking engineers to suggest the use of 3G cellular network for smart meter data collection. It is more appropriate to consider AMI network as a long distance sensor network (single hop) rather than a cellular network. Therefore designing an efcient (low overhead and low latency) custom long distance wireless sensor network would be more appropriate rather than using a 3G cellular network. Similar to any other network, once set up, AMI sensor network is also expected to be used for several years. Therefore from the utilities perspective, it is worth considering the idea of designing a custom AMI network rather than using an already existing 3G/4G network. Also, setting up their own network gives the utilities improved privacy and secrecy in the data transfer. In general, most wireless sensor network designs are limited to a small coverage area of 2 to 3 km. In recent years, a lot of research has been done in the area of long distance wireless sensor network design and experimental test beds were set up using off the shelf equipment to transmit and receive upto 13.2 km [9] [10]. The current state of the art radios that are used in wireless sensor network design can transmit up to 40 miles with a data rate of 115.2 kbps [11]. Motivation can be drawn from these research studies to design a long range AMI sensor network that can exactly meet the demands of smart grid applications. As mentioned above, with the current state of the art RF modems, designing a long range wireless sensor network is not difcult to achieve but the main issue is to reduce the data overhead. Unlike the TCP/IP based CDMA stack, a data stack similar to the Zigbee sensor network can be used in the custom AMI sensor network design to reduce the overhead. Zigbee protocol stack does not include a transport layer [12]. Therefore it is up to the designer to ensure an end-to-end reliable data delivery in a Zigbee data transfer. We are not suggesting the use of Zigbee protocol stack for AMI radios because MAC and PHY layers of Zigbee are insecure [13]. But we are suggesting that a solution to reduce control data overhead is to design an AMI network protocol stack similar to Zigbee i.e. without transport(TCP) layer and ensure end-toend reliable data delivery at the application layer. Apart from the transport layer, MAC layer design also plays a major role in control data overhead reduction. The 3G cellular MAC layer overhead and latency issues are discussed in Section IV-C.

TABLE I D ATA R ATE R EQUIREMENTS Parameter Max Application Payload Number of Smart Meters Latency Data Collection frequency Reliability Data Rate Uplink 2400 50 5 sec 30 mins 99% 232 kbps Downlink 255 50 5 sec 30 mins 99% 45.3 kbps

B. Closed Loop Power Control Issues In order to overcome the near-far problem and combat Doppler shift, 3G cellular technology employs closed loop power control. Power control is a continuous process and it operates at a frequency of 1500 Hz [14] i.e. every 0.66 msec Base Station (BS) performs a computation related to the transmit power of an active Mobile Station (MS) in its cell and sends power control information to the MS to decrease/increase the transmit power by dB during transmission of next frame. In the case of AMI, most of the smart grid devices are stationary. Hence there is no chance of smart meters experiencing doppler shift because of mobility but there is a slight possibility of doppler effect due to changes in environment. Therefore, like CDMA, AMI network may not require faster power control. Power control may be done once per connection based on the currently received pilot and past performance. A closed power control that operates at a low frequency (< 100 Hz) should be sufcient for this task. Thus, reducing the power control operating rate not only reduces the computation burden on the base station but also saves power. The downlink capacity of a CDMA based 3G cellular network is limited by the transmit power at the base station. Therefore for a given transmit power, a custom AMI network designed with a slower power control loop can achieve a higher system capacity when compared to that of the current state of the art 3G cellular technology which operates at 1500 Hz [14]. C. Spread Spectrum vs LTE In the near future, 3G cellular technologies like CDMA and GSM will be obsolete and are expected to be replaced by 4G technologies like LTE and LTE advanced. LTE relies on orthogonal frequency division multiple access (OFDMA) and MIMO and it is targeted at achieving high data rates around 100 Mbps. According to our analysis, as shown in Table I, smart grid application data rate requirements are less than 250 kbps. Compared to LTE, spread spectrum based technologies excel at these data rates and also exhibit good multiple access properties. Therefore designing a spread spectrum based network is more appropriate for smart grid applications rather than using OFDMA based 4G technologies. IV. HSS BASED AMI N ETWORK D ESIGN It is aforementioned that the number of smart meters handled by a DAP can vary from as few as 500 to as many as 30,000 [1]. Therefore a multiple access scheme is required to support a high number of smart meters. CDMA is selected as the multiple access scheme with hybrid spread spectrum (HSS) as the physical layer technique. CDMA offers many advantages over TDMA and FDMA but the main reason for the selection of CDMA for the AMI network design is that the CDMA outperforms TDMA and FDMA in noise, interference and jamming environments. In order to provide the smart meters access to the AMI network, using CDMA as a multiple access scheme we design a MAC protocol.

Similar to multiple access scheme, there are many options to select the PHY layer spread spectrum technique. They are Direct Sequence Spread Spectrum (DSSS), Frequency Hopping Spread Spectrum (FHSS) and Time Hopping Spread Spectrum (THSS) or any combination of these i.e. Hybrid Spread Spectrum (HSS). HSS can combine the good features of both DSSS and FHSS while overcoming some of their shortcomings [15]. For example, a HSS system can coalesce the good antipartial-band-jamming effectiveness of a FHSS system with the anti-multipath feature of a DSSS system. Both these features are very important for smart grid radios because the electric utility environments are harsh (Electromagnetic Interference 300 MHz - 1 GHz) and noisy [16] [17]. Also, HSS systems use shorter signature sequences and hopping patterns, which reduces the overall acquisition time. The above advantages of HSS comes at the cost of the increased transmitter and receiver complexity. In this section, we discuss the details of preliminary MAC layer design of HSS based AMI network that can reduce the overhead and latency in data transfer from the smart meter to the DAP when compared to the 3G cellular technologies. The PHY layer performance of HSS is already a well researched area therefore our concentration is more on MAC layer design [18] [19]. However we provide some insights into the PHY layer. A. AMI PHY Layer The performance of the MAC layer also depends on the PHY layer. Therefore, in order to design a MAC layer the details of the PHY layer have to be known in advance. Hence the following assumptions are made about the AMI physical layer in order to design the AMI-MAC layer. 1) Uplink: (a) Frequency hopping patterns are generated by memoryless random uniform distribution of frequencies whereas long m-sequences are used for direct sequence spreading. Just for the purpose of simulations we used random frequency hopping but in practice we intend to use psuedo-random hopping. (b) Multiple data rates can be achieved by varying the spreading gain and modulation scheme. (c) We assume a Slow Frequency Hopping (SFH) pattern. As SFH is used, coherent demodulation is possible because the frequency hopping rate is much slower than the data rate. Therefore Binary Phase-Shift-Keying (BPSK) is selected for modulation. Half-rate (1/2) constraint length 7 convolutional coding is used as forward error correction technique. 2) Downlink: (a) Walsh codes of length 64 are used for direct sequence spreading and four different frequency bands are used for frequency hopping, resulting in a total spreading gain of 256. (b) The spreading gain or modulation scheme can be varied to achieve multiple data rates. (c) Similar to uplink, SFH, BPSK modulation and 1/2 rate convolutional coding is used. B. Channels The main objective of the AMI-MAC layer is to perform the mapping between the logical and the physical channels. We assume that the AMI network has pilot, synchronization, paging and trafc channels in the downlink and random access channel, trafc channel in the reverse link. The above

mentioned channels exist in any cellular technology but the number and the type (Dedicated and Control) of these channels can differ from one technology to other. C. Cellular CDMA MAC vs AMI-MAC The proposed AMI-MAC is a variant of the cellular CDMA MAC. Therefore we explain the operation of the AMI-MAC by comparing it with the operation of Cellular CDMA-MAC. The main differences between the cellular MAC and the AMI-MAC are as shown in Fig 2. The system initialization, system idle and trafc channel states are the same whereas the behaviour of the system access state is different in the AMI network when compared to that of the cellular network. It is aforementioned that the AMI network carries both scheduled and random trafc and the percentage of scheduled data (> 90%) is very high compared to the random data (> 10%). As the AMI network can schedule the data to be collected from smart meters, it is known in advance when a smart meter sends data to the DAP, whereas in cellular network the time of call origination or data transfer from mobile is not known in advance. The proposed AMI-MAC exploits the knowledge of the time of occurrence of the data transfer to reduce or eliminate the control overhead required for data transfer set up for the scheduled data collection. Unlike the CDMA MAC which incurs control overhead and latency for data transfer set up. As shown in Fig 2, in 3G cellular networks (CDMA), a MS initiates a call by sending a request on the access channel to the BS to allocate (reserve) a trafc channel for the purpose of voice/data communication. Moreover while sending an access request, every MS competes with many other mobiles in the cell that are sending requests. If a mobiles access request collides with the request from an other mobile, it has to retransmit the request which results in longer connection delay. BS acknowledges the MSs request by sending a message on the paging channel. This is an indication that the system resources are reserved for the MS. MS uses the information sent by the BS for further voice/data communication. The important part of the data sent by the BS to the MS is a secret key. Using the secret key and the unique identication number of the MS as inputs to a hash algorithm, MS generates a number which serves as a starting point of the 242 1 length m-sequence. m-sequence is used for the purpose of spreading. Using the same hashing algorithm and secret key used by the MS, the BS generates the same number to serve as a starting point of the m-sequence for the purpose of despreading. As explained above the entire process of call set up incurs few seconds of delay and unnecessary data overhead due to the MS access request and message exchanged by the BS with a MS. As mentioned earlier a high percentage (> 90%) of the data collected from the smart grid applications is scheduled and a small portion of the data is random. The time scheduling in data collection can be leveraged to obviate the data transfer/call set up process. For example, if a smart meter with unique identication number x is scheduled to send data to the DAP at time y. Both the variables x and y can be used as inputs to a hashing algorithm to generate the starting point of the long m-sequence. Both the smart meter and a demodulator at the DAP can be tuned to the same m-sequence for the purpose of spreading and despreading respectively. In 3G cellular the overhead and the latency associated with the call set up process cannot be avoided because of the randomness in call origination. Whereas in smart grid applications, as explained above, a MAC layer designed to take the advantage

Tune to CDMA carrier

Tune to CDMA carrier

System Initialization State

Mobile acquires Pilot and Synch channel SM acquires Pilot and Synch channel

System Idle State

Mobile acquires Paging channel and monitors for messages SM acquires Paging channel and monitors for messages On demand Data Y%

System Access State

Mobile sends messages on access channel; BS sends messages on paging channel

Scheduled Data X% SM sends messages on access channel; DAP sends messages on paging channel

Traffic Channel State

Speech Communication on forward and reverse link traffic channels, downlink power control

Scheduled data exchange on forward and reverse link traffic channels. No power control.

On demand Data exchange on forward and reverse link traffic channels, downlink power control.

Fig. 2.

Differences between cellular MAC vs AMI MAC

2

Probability of bit error

of time scheduled data collection can reduce the overhead and latency in data transfer. But in order to leverage the time scheduled data collection, DAP and smart meters have to be time synchronised, which is achieved using pilot and synchronization channels. V. S IMULATION
AND

10

10

R ESULTS

In this section, we discuss both PHY and MAC layer simulations. The following assumptions are made for the purpose of simulation: 1) Smart meter radio transmission and reception occurs at different frequency, i.e. Frequency Division Duplexing (FDD) scheme is employed for dividing forward and reverse communication channels. 2) Smart meters employ synchronous frequency hopping. 3) Smart meters are chip synchronous but not bit synchronous with random phases uniformly distributed on (0, 2 ). 4) The system has a perfect power control i.e the received signal power from all the smart meters at the DAP is same. Also the receiver uses a matched lter. 5) Smart meters can transmit and receive data at the same time. A. PHY layer For a given HSS spreading gain, the number of frequencies (q) and direct sequence spreading gain (N) are varied. We assume System I with q = 1 and N = 256, System II with q = 4 and N = 64, System III with q = 8 and N = 32 and System IV with q = 16 and N = 16. System I is a DSSS system and all other systems are HSS systems. The HSS spreading gain (qxN ) of these four systems is 256. The performance of these four systems are compared in AWGN and Jamming channels. 1) AWGN Channel: In this scenario, the performance of all the four systems are compared in AWGN channel without using error correction coding and interleaving. The Number of Bits per Frequency Hop (NBFH) is assumed to be 100. The simulation results are as shown in Fig 3. It is apparent from

10

10

10

20 30 40 Number of users

N=64 q=4 N=32 q=8 N=16 q=16 N =256 q=1 50 60

Fig. 3. Performance comparison of DS-SFH vs. DSSS system in AWGN channel

Fig 3 that the performance of DSSS system is better than all the three HSS systems. 2) Jamming Channel: In this case, the performance of all the four systems are evaluated in the presence of a single band jammer. The Signal to Interference Ratio (SIR) i.e. the ratio of signal power to jammer power is varied from -60 dB to 10 dB. Number of smart meters is xed to be 20. The coding rate and constraint length of the convolutional coder are 1/2 and 7. The simulation results for NBFH = 100 and NBFH = 10 are as shown in Fig 4 and 5 respectively. It is obvious from the simulation results that the performance of all the four systems degrades signicantly in the presence of a high power jammer i.e. the BER is close to 10%. The performance of the Hard Decision Decoded (HDD) System IV with NBFH = 10 is better when compared with all the other systems. The performance of HDD system IV can be further improved by decreasing the convolutional code rate. As shown in Fig 5, using a convolutional coding rate of 1/3, BER is reduces to 10 4, which is more appropriate for data transfer. In conclusion, a low coding rate HDD System IV can be used to combat high power single band jammer.

10

10

10

10 60

Soft N=16 q=16 Hard N=16 q=16 Soft N=32 q=8 Hard N=32 q=8 Soft N=64 q=4 Hard N=64 q=4 50

Signal to Interference Ratio (SIR)

40

30

20

10

10

Fig. 4. Performance comparison of DS-SFH spread spectrum system in Jamming channel with NBFH = 100
10
0

at the PHY layer, facilitate multiple transmission sub-channels (frequencies and/or code distinguished), packet detection and timing acquisition, cumulative interference based Signal to Interference Noise Ratio (SINR) calculations, look up table based Bit Error Rate (BER) and Frame Error Rate (FER) calculations. The proposed architecture in [2] is tested using UWB radios for short range and high bandwidth communications. As mentioned in [2] in order to adapt their architecture, in this paper we have customized the calculation of interference, the modelling of interference from other sub-channels and calculation of FER for DSSS and HSS simulations. The probability of bit error rate for DSSS systems is given as follows: 1 pb = Q ( ) (1) K N0 =2 Pk ) ( 2Eb,1 ) + ( 3k P1 N Where K = the number of smart meters N = Processing gain Pk = Power of kth smart meter Eb,1 /N0 = SNR of desired smart meter In the case of HSS simulations, we assumed that the frequency hopping of smart meters occur synchronously. Therefore there is no scope for partial hits and only full hits occur. In order to compute the pb of the desired smart meter with a given frequency hopping pattern, for each frequency we count the number of smart meters that are transmitting in the same frequency at the same time and based on that the bit error rate pbi for the ith frequency hop is calculated using equation 1. The nal probability of bit error rate for HSS systems is given by L pbi pb = i=1 (2) L where L = length of the frequency hopping pattern. Assuming that the bits are independent and identically distributed and K is the error correction capability, the FER can be obtained from BER as follows:
K

Probability of bit error (BER)

Probability of bit error (BER)

10 10 10 10 10 10

10 60

Soft N=16 q=16 Hard N=16 q=16 Soft N=32 q=8 Hard N=32 q=8 Soft N=64 q=4 Hard N=64 q=4 Soft r=1/3 N=16 q=16 Hard r=1/3 N=16 q=16 50 40 30 20 10 0 10

Signal to Interference Ratio (SIR)

Fig. 5. Performance comparison of DS-SFH spread spectrum system in Jamming channel with NBFH = 10

B. NS-2 Issues and Modications Network simulator NS-2 is used to evaluate the performance of the AMI-MAC layer described in section IV. NS-2 is an invaluable tool for evaluating the performance of new protocols. Unfortunately with the current wireless network simulation platform of NS-2, it is far from easy to implement new wireless PHY layers or to adapt the existing IEEE 802.11 and IEEE 802.15.4 protocols to implement new wireless protocols. In order to overcome this deciency of NS-2, Ruben et al proposed a generalized architecture to implement a new wireless PHY layer [2]. In this subsection, we briey explain the wireless PHY layer issues of NS-2 and also provide a brief summary of the work done in [2]. The following are some of the wireless PHY layer implementation issues in NS-2: a) There exists a strong interdependency between the implementations of wireless PHY and MAC layers of IEEE 802.11 and IEEE 802.15.4. In other words, the packet reception occurs mainly in the MAC layer rather than the PHY layer. b) Lack of support for multiple sub-channels which restricts the simulation of multiple access networks such as CDMA and UWB etc. c) Absence of error model i.e. the current implementation of NS-2 considers that a packet is received if the received signal strength is greater than a threshold. Number of bit errors are not taken into account. d) No model of cumulative interference i.e. the effect of interference from concurrent transmissions is also ignored. In order to solve the above issues of NS-2, the key features of the architecture proposed in [2] are complete packet reception

pf = 1
k=0

n k p (1 pb )nk k b

(3)

C. MAC Layer In order to evaluate the performance of the MAC layer in terms of the throughput and the average packet delay, the following network topology is created in NS-2. Smart meters are randomly placed with in a radius of 10 Kilometres from the DAP. We assume a perfect power control i.e. the received power of all the smart meters at the DAP is same. In NS-2, we achieve perfect power control by adjusting the transmit power of each smart meter. Also, we restrict our simulations to single cell and do not consider the adjacent cell interference. 1) Throughput: We dene throughput as the number of successful frames per frame duration. Fig 6 shows the variation of throughput with the offered load for DSSS and HSS systems. It is obvious from the simulation results that the DSSS system outperforms HSS systems in an AWGN Channel. The simulation results are in conformance with the theory. In the case of DSSS systems, the peak value of throughput (126) occurs when the number of smart meters transmitting at the same time is 142. But this may not be the ideal operating point as 16 packets are dropped. Therefore, in order to nd the optimum operating point we plot the Percentage of Successful Frames (PSF) against the offered load, which is as shown

100

Throughput vs SIR

Throughput vs Offered Load

140

Throughput

120

spreading code = 256 HSS = 128*2 HSS = 64*4 HSS = 32*8

80 60 40 20 0 -70

100

Throughput

80

60

-60

-50

-40

spreading code = 256 HSS = 128*2 HSS = 64*4 HSS = 32*8 -30 -20 -10 0 10 SIR(dB)

20

40

Fig. 8.

Throughput vs SIR in the presence of jammer

20

0 20 40 60 80 100 120 140 160 180 200

Offered Load

Fig. 6.

Throughput vs Offered Load in AWGN channel

Frame Success Rate vs Offered Load

100

90

80

70

60

50

40

spreading code = 256 HSS = 128*2 HSS = 64*4 HSS = 32*8

20 40 60 80 100 120 140 160 180 200

Offered Load

Fig. 7.

PSF vs Offered Load in AWGN channel

in Fig 7. For DSSS systems, when the offered load, i.e. the number of smart meters transmitting simultaneously is less than 100, PSF is close to 100%. But when the offered load is greater than 100, PSF gradually decreases and nally reaches 40% at an offered load of 200. Similar to DSSS, the optimum number of smart meters for HSS systems is also close to 100, i.e. there is very little change in the performance of DSSS and HSS systems in AWGN channel. But in the presence of a single band high power jammer, the HSS system out performs DSSS systems. This is as shown in Fig 8. In Fig 8, number of smart meters is xed at 100 and SIR is varied from -60 dB to 20 dB. Among the four systems compared, the HSS system with q = 8 and N = 32 gives the best performance when SIR is very low i.e. between -60 dB to -20 dB. As SIR is increased above -20 dB, DSSS system gives better throughput than HSS systems. Thus, it can be inferred that, for a given bandwidth HSS system offers higher throughput than DSSS systems in jamming or harsh environments like electric substation. This justies the use of HSS for smart meter radio design. 2) Average Packet Delay: The AMI network is expected to carry both scheduled and random trafc. The average packet delay of scheduled data depends on the size of the data and the frame size. In the case of random data, a smart meter competes

with many other smart meters in the network to send data on the random access channel and it is very likely that a collision occurs. Therefore a smart meter may have to retransmit many times before the data is successfully delivered to the DAP. Hence we intend to evaluate average packet delay of random trafc in AMI networks. The delay of a successfully transmitted packet is dened as the time duration between the time the packet is ready to be transmitted at the head-of-line of the smart meter queue, until an acknowledgement for this packet is received from DAP at the smart meter [20]. If a packet crosses the specied retransmission limit then this packet is dropped and its time delay is not taken into account for the calculation of average packet delay. As discussed in section IV-C the random data transmissions in AMI networks occur similar to that of cellular CDMA. The random access procedure we have implemented in NS2 is given in [21]. The parameters used for the purpose of simulations are as shown in Table II. The number of smart meters that a DAP handles can vary from as few as 500 upto as many as 30000. Generally in cellular CDMA there are 7 paging channels [21]. Therefore we assume that the AMI network also employs same number of paging channels. In the worst case scenario, i.e. when the total number of smart meters is 30000, the maximum number of smart meters that a DAP can handle per a paging channel is 4500. Therefore, in simulations the number of smart meters per paging channel are varied from 1000 to 5000 and average, minimum and maximum packet delay values in seconds are observed. As shown in Table II, ACK TIME OUT and PAGING CYCLE LENGTH are varied in terms of number of frames i.e. 2IN DEX F RAM ESIZE . For the purpose of simulations, INDEX is varied from 0 to 9. In order to evaluate the average time delay, we have simulated a scenario where all the smart meters report power outages after a blackout and expect an ACK from DAP. As per [22], the random access channel data rates are selected based on the payload size. For convenience, these data rates are shown in Table III. We assume the power outage data size to be 25 Bytes (< 232 bits). Therefore we select the random access channel data rate as 9.6 Kbps. For the purpose of convenience in simulations, we used 10 Kbps (close to 9.6 Kbps) data rate and 2.56 Mcps chipping rate which results in a spreading gain of 256. It was observed from the simulations that, when the value of INDEX is less than or equal to 5, a large number of smart meters experience random access transmission failures. Therefore we considered only the INDEX values of 6 to 9. Among these, the INDEX value of 9 gives better results

Frame Success Rate

TABLE II S IMULATION PARAMETERS Parameter values PAM SZ 1 MAX CAP SZ 1 MAX REQ SEQ 15 NUM STEP 15 ACK TIME OUT varied NUM PAGING CHANNELS 1 NUM SM PER PAGING varied MSG PSIST 1 PSIST 0 PD 2 ((P SIST /4) + (M SG P SIST )) FRAMESIZE ((PAM SZ + MAX CAP SZ)*0.0266)s PROBE BACKOFF 15 SEQUENCE BACKOFF 15 NUM RANDOM CHANNELS 32 PAGING CYCLE LENGTH varied TABLE III R ANDOM A CCESS C HANNEL R ATES AND PAYLOAD
Random Traffic (Mbits) Random Traffic (Mbits)

6 5 4 3 2 1 0 0 50

Random Traffic vs Time

No. of SMs = 715(5000) No. of SMs = 572(4000) No. of SMs = 429(3000) No. of SMs = 286(2000) No. of SMs = 143(1000) 100 150 200 Time (sec)

Fig. 9.

Uplink Random Access Trafc vs Time during a blackout

Random Traffic vs Time 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 5 10 No. of SMs = 715(5000) No. of SMs = 572(4000) No. of SMs = 429(3000) No. of SMs = 286(2000) No. of SMs = 143(1000) 15 20 25 30 35 40 Time (sec)

Transmission Rate (kbps) Min Payload (bits) Max Payload (bits) 9.6 1 232 19.2 233 488 38.4 489 1000 76.8 1001 2024

which are give in Table IV. As the number of smart meters increase, the time taken to collect power outage data also increases and this is obvious from simulation results. In the worst case scenario, i.e. when the number of smart meters per paging channel is 5000, the total time taken to collect power outage data using the random access channel is 230 sec (approx 4 minutes). Unlike cellular CDMA which considers average packet delay, in the case of AMI network maximum packet delay is important because power OMS takes into consideration of all the power outage messages to gure out exactly the region that is experiencing the power outage. A large amount of random access trafc is generated during a power outage reporting, the CDF of it is shown in Fig 9. One of the main objectives of the AMI network design is that the last gasps must be delivered in seconds [23]. Therefore we propose a time hierarchical scheme to reduce the trafc volume and last gasp message delay. VI. T IME H IERARCHICAL S CHEME TO R EDUCE R ANDOM T RAFFIC VOLUME A high level overview of power generation, transmission and distribution system is as shown in Fig 11. Based on the operating voltages, distribution system is divided into primary distribution system (2.4 KV to 69 KV) and secondary distribution system (120 V to 600 V). In general, Distribution Substation (DS) transformers step down the transmission system voltage to lower levels. The section of the power network connected between DS and Distribution Transformers (DT) is considered as Primary Distribution System (PDS). PDS consists of primary or distribution feeders that originate from DS. Secondary Distribution System (SDS) connects primary
TABLE IV PACKET DELAYS ( SEC ) OF RANDOM TRAFFIC IN AMI NETWORK No. of Smart Meters 1000 2000 3000 4000 5000 Cellular CDMA Min Avg Max 0.28 31.7 86.66 0.18 53.67 118.18 0.36 75.73 160.26 0.38 96.85 189.1 0.36 118.63 229.84 Min 0.1 0.28 0.18 0.14 0.3 THS Avg 2.91 5.82 7.85 10.18 14.08 Max 8.72 16.68 26.24 33.4 43.98

45

Fig. 10. THS

Uplink Random Access Trafc vs Time during a blackout using

feeders with the industrial and residential customers. SDS consists of step down transformers and secondary circuits at utilization voltage levels. It is obvious from the above description of power distribution system that there exists a hierarchy in its network. This is as shown in Fig 11. We propose a Time Hierarchical Scheme (THS) that leverages the hierarchy in power distribution system to reduce the amount of trafc generated during a blackout. The following assumptions are made about THS: 1) It is assumed that there exists three types of smart meters. The differences among these Smart Meters (SM) exists in the time delays with which they report the power outage after a blackout has occurred. We assume that the smart meters 1, 2 and 3 report respectively with the time delays t1 , t2 and t3 after a blackout. 2) SM1 is installed at DS, SM2 is installed at DT and SM3 is installed at customers house. THS works as follows: Case i): The major blackouts in U.S. and Europe are due to the failures in power transmission system [24]. Therefore in the case of a large scale blackout, it is very likely that along with the DS, all the customers that are fed by the DS are also out of power. Hence, instead of each individual SM3s reporting power outage to a DAP, SM1 present at the DS reports to the DAP that all the customers that are fed by it are out of power. Once DAP receives power outage message from SM1, it sends a broadcast message to all the SM3s under the coverage of SM1 that it is aware of the power outage and it is not necessary for them to report the outage. Thus, with the use of THS, the entire uplink trafc shown in Fig 9 can be suppressed by a single packet transfer on the uplink from SM1 to DAP and a broadcast message on the downlink from DAP to all SM1s and SM2s.

Power Generation Power Transmission Distribution Substation PDS

Smart Meter 1 Smart Meter 2 Smart Meter 3

Sleep Mode Enabled URA_PCH

Via Explicit Signaling inactivity timer expired Via Explicit Signaling Via Explicit Signaling Release DCH timer expired Assign DCH
SDS

CELL_PCH
PCH Paging Channel DCH Dedicated Channel RCH Random Access Channel

Power Distribution

DT

DT

DT

DT

CELL_DCH

CELL_RCH

Fig. 12.

WCDMA RRC States

TABLE V 2011 WCDMA P OWER C ONSUMPTION PARAMETERS Fig. 11. Power Generation, Transmission and Distribution System Parameter values (mW) CELL PCH/URA PCH 30 CELL RCH 200 CELL DCH 400 Power Amplier 93 RF transmit IC 103 RF receive IC 130 Baseband 130

Case ii): Consider the scenario where there is a fault in the primary feeder that is connected between DS and DT. In this case, DS is not out of power but all the DTs and customers are out of power. As per [13], each pad-mount/pole-mount DT can supply power from 7 to 8 houses. Therefore, a single DT can report the power outage to the DAP on behalf of 7 to 8 houses. This can reduce the number of power outage reportings from 5000 to 715, which can signicantly reduce uplink random access trafc during a power outage. Similar to the above case, Once DAP receives power outage message from a SM2 at a DT, DAP sends a message on the paging channel to the SM3s under the coverage of the DT that it is aware of the power outage and it is not necessary for them to report the outage. In this case, the random access trafc generated on the uplink using THS during a power outage is as shown in Fig 10. Comparing Fig 9 with Fig 10, it is obvious that THS can signicantly reduce the random access trafc during an outage. When the number of smart meters is 5000 per paging channel, using cellular random access technique the total power outage reporting data from all the smart meters is 5.86 Mbits whereas the use of THS reduces the data size to 0.71 Mbits. This results in huge bandwidth savings. The reduction in the latency in data transfer due to THS is shown in Table IV. With the use of THS, when the number of smart meters is 5000, the total time in power outage reportings is reduced from 229.84 sec (4 Minutes) to 43.98 sec. Hence the use of THS aids the AMI network to meet the OMS requirements i.e. the outage reporting delay should be of the order of seconds rather than minutes [23]. Case iii): In this case, we assume that a fault exists in secondary feeder or DT. As the number of houses fed by a DT is less than 8, the amount of trafc generated in this case is very less. Therefore SM3s can use the CDMA random access procedure to report the power outages. Whenever there is a power outage, SM3s, do not know the reason for it, i.e. whether the power outage is due to transmission system failure, a fault in primary feeder or secondary feeder. Therefore SM3s cannot decide whether they have to report the power outage or whether DT/DS will report on their behalf. Hence we maintain a time hierarchy (time delays) in power outage reportings of the three types of smart meters and there exists a relationship between these time delays i.e. t1 < t2 < t3 . It is upto the network designer to choose these values.

VII. E NERGY S AVINGS DUE TO THS Generally smart meter draws the power required for its operation from the bus it is monitoring. But during a power outage the bus doesnt supply power. Therefore smart meters are expected to be equipped with a battery that can be used for power outage data reporting. The lower the battery power rating, the lower the hardware costs associated with it. Hence, one of the main objectives of designing an AMI network is to reduce the smart meter energy consumption during the power outage reporting. As discussed in section VI, when compared to cellular CDMA, THS signicantly reduces the total power outage data size and the latency in data transfer. Therefore, in this section we focus on the energy savings due to THS and compare it with cellular CDMA energy consumption. The Radio Resource Connection (RRC) state transition diagram of WCDMA is as shown in Fig 12 [25]. We assume that the RRC state diagram of the HSS based AMI network is similar to that of the WCDMA network. In CELL PCH/URA PCH state, smart meter is in sleep mode and wakes up once in a paging cycle period to monitor for paging messages. When a smart meter is assigned a dedicated channel it will operate in CELL DCH state and it operates in CELL RCH state while transmitting data on the random access channel. There exists slight differences in the RRC states of WCDMA and Cdma2000 but the energy consumption is more or less the same in both. Therefore we just compare the energy consumption of HSS based smart meters employing THS with WCDMA. The WCDMA power consumption of RRC states, RF and baseband ICs is as shown in Table V, as per [25] [26]. Using these values, we calculated the energy consumption of smart meters with and without THS in the three cases discussed in section VI. Similar to [25], the energy consumption of smart meters is calculated based on the amount of time spent by the smart meters in the MAC layer states. case i): A single power outage reporting from SM1 negates a large number of SM3 and SM2 power outage reportings. In this case, SM2s and SM3s will be operating in sleep mode and just wake to monitor paging/broadcast channel. Therefore the average energy consumption of a smart meter as per the

TABLE VI E NERGY C ONSUMPTION (J OULES ) COMPARISON WITH AND WITHOUT THS No. of Smart Meters 1000 2000 3000 4000 5000 Without THS Min Avg Max 0.065 6.34 17.34 0.045 10.73 23.64 0.081 15.15 32.06 0.085 19.37 37.82 0.0811 23.731 45.977 With THS Min Avg Max 0.033 0.622 1.795 0.065 1.25 3.426 0.045 1.69 5.377 0.037 2.206 6.853 0.069 3.03 9.018

U.S. Department of Energy. Accordingly, the United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. R EFERENCES
[1] http://www.qualcomm.com/media/documents/3g-cellular-technologysmart-grid-communications. [2] R. Merz, J.-Y. Le Boudec, and J. Widmer, An architecture for wireless simulation in ns-2 applied to impulse-radio ultra-wide band networks, in CNS 2007, 2007. [3] SAG (Stake holders advisory group) meeting with southern company. [4] P. Jianping, Research - tcp/ip over air links - cdma/hdr, March 2001. [Online]. Available: http://webhome.cs.uvic.ca/ pan/tcpair/tcpcdma.html [5] J. D. Cavanaugh, Protocol overhead in ip/atm networks, 1994. [6] F. Frank, H. Stefan, S. Patrick, and M. Reisslein. (2003) Header compression schemes for wireless internet access. [Online]. Available: www.3g4g.co.uk/Other/ROHC/Whitepapers/rohc book.pdf [7] O. users group, Sg network system requirements specication v4.0-draft6, jun 2010. [Online]. Available: http://osgug.ucaiug.org/UtiliComm/Shared%20Documents/SGNET%20PAP%20work-in-progress/ [8] C. Sanders, Practical Packet Analysis: Using Wireshark to Solve RealWorld Network Problems, 2nd ed., Jul. 2011. [9] M. Zennaro, A. Bagula, D. Gascon, and A. B. Noveleta, Long distance wireless sensor networks: simulation vs reality, in Proceedings of the 4th ACM Workshop on Networked Systems for Developing Regions, ser. NSDR 10. New York, NY, USA: ACM, 2010, pp. 12:112:2. [Online]. Available: http://doi.acm.org/10.1145/1836001.1836013 [10] http://eprints.jcu.edu.au/2034/2/01front.pdf. [11] http://www.digi.com/technology/wireless/products. [12] G. Thonet, Zigbee faq, Feb 2006. [Online]. Available: wsnet.les.wordpress.com/2006/08/zigbee-faq.pdf [13] J. Brodsky and A. McConnell, Jamming and interference induced denial of service attacks on ieee 802.15.4 based wireless networks, in SCADA Security Scientic Symposium, 2009. [14] H.-H. Chen, The next generation CDMA technolgies. West Sussex, England: John Wiley & Sons Ltd, 2007. [15] E. Geraniotis, Coherent hybrid ds-sfh spread-spectrum multiple-access communications, Selected Areas in Communications, IEEE Journal on, vol. 3, no. 5, pp. 695 705, sep 1985. [16] A. Nesbitt, B. Stewart, S. McMeekin, S. Conner, J. Gamio, K. LiebechLien, H. Kristiansen, and S. Krakenes, A novel approach to high voltage substation surveillance using radio frequency interference measurement, in Electrical Insulation Conference, 2009. EIC 2009. IEEE, 31 2009june 3 2009, pp. 159 163. [17] M. Olama, X. Ma, T. Kuruganti, S. Smith, and S. Djouadi, Hybrid ds/ffh spread-spectrum: A robust, secure transmission technique for communication in harsh environments, in MILITARY COMMUNICATIONS CONFERENCE, 2011 - MILCOM 2011, nov. 2011, pp. 2136 2141. [18] E. Geraniotis, Coherent hybrid ds-sfh spread-spectrum multiple-access communications, Selected Areas in Communications, IEEE Journal on, vol. 3, no. 5, pp. 695 705, sep 1985. [19] , Noncoherent hybrid ds-sfh spread-spectrum multiple-access communications, Communications, IEEE Transactions on, vol. 34, no. 9, pp. 862 872, sep 1986. [20] P. Raptis, V. Vitsas, K. Paparrizos, P. Chatzimisios, A. C. Boucouvalas, and P. Adamidis, Packet delay modeling of ieee 802.11 wireless lans. [21] B. d. S. A. Xavier, Call Processing in CDMA Systems. John Wiley & Sons, Ltd, 2005, pp. 239290. [Online]. Available: http://dx.doi.org/10.1002/0470020121.ch6 [22] 3rd generation partnership project 2, 3gpp2 c.s0024-300-c : Medium access control layer for cdma2000 high rate packet data air interface specication, 2011, version 2.0. [23] C. Montgomery. (2012, Jan.) How to know which ami platform is best for outage management. [Online]. Available: http://www.smartgridnews.com/artman/publish/Technologies Metering/Howto-know-which-AMI-platform-is-best-for-outage-management-4339page2.html [24] S. Garlapati, H. Lin, S. Sambamoorthy, S. Shukla, and J. Thorp, Agent based supervision of zone 3 relays to prevent hidden failure based tripping, in Smart Grid Communications (SmartGridComm), 2010 First IEEE International Conference on, oct. 2010, pp. 256 261. [25] J.-H. Yeh, J.-C. Chen, and C.-C. Lee, Comparative analysis of energysaving techniques in 3gpp and 3gpp2 systems, Vehicular Technology, IEEE Transactions on, vol. 58, no. 1, pp. 432 448, jan. 2009. [26] H. Harri and T. Antti, WCDMA for UMTS, 5th ed. John Wiley & Sons Ltd., 2010, pp. 272,579.

calculations in [25] is very small, i.e. 58 mJ (milli Joules). On the other hand, the average energy consumption of smart meters without THS is 23.73 Joules. case ii): In this case, SM2s report power outage to the DAP on behalf of SM3s. The energy consumption of smart meters during power outage reportings with and without THS are as shown in Table VI. It is obvious from Table VI that the average energy consumption of smart meters with THS is approximately 1/8th of that of the energy consumption without THS. Thus, THS can signicantly reduce the power rating of battery that is associated with the smart meter. case iii): Unlike the above two cases, in this case, the average energy consumption of the smart meters using THS is expected to be higher than the energy consumption of the smart meters without THS. The reason for this is that the smart meters wait and monitor broadcast/paging channels for the time period (t1 + t2) before reporting power outage to the DAP. We assume t1 = 0, and the values of t2 = 11 sec and t3 = 55 sec are obtained from simulation results. The energy consumption of smart meters without THS is 28.92 mJ whereas the use of THS increases the energy consumption to 260 mJ. As the number of smart meters reporting power outage in this case is very less, i.e. 7 or 8, the difference in energy consumption of smart meters with and without THS is expected to be of the order of mJ and this is obvious from the above mentioned values. The increase in energy consumption with the use of THS is not a major issue because the battery power rating is generally specied for worst case scenario, i.e case ii. VIII. C ONCLUSION In conclusion, 3G cellular CDMA technology is not a good choice for distribution side smart grid applications as its usage results in high overhead and latency in data transmission. On other hand, 4G technologies like LTE and LTE-Advanced are tailor made to meet the high data rate requirements of smart phones in the range of 75 Mbps to 300 Mbps. Smart grid applications data transfer rates are likely to be less than 250 kbps. Compared with 4G technologies, spread spectrum technologies excels at these low data rates. Therefore we proposed a custom long distance hybrid spread spectrum based wireless sensor network. We have presented a preliminary PHY and MAC layer design of the HSS based AMI network and performed simulations using matlab and NS-2. Compared to 3G cellular technologies, the proposed design can achieve low overhead, less latency and lower power consumption for smart grid applications. We also proposed a time hierarchical scheme that can signicantly reduce the random access trafc generated during a blackout and also it can reduce the delay and energy consumption of smart meters in power outage reporting. ACKNOWLEDGMENT This paper has been co-authored by employees of UTBattelle, LLC, under contract DE-AC05-00OR22725 with the