Docstoc

Control-Aware Wireless Sensor Network Platform for the Smart Electric Grid

Document Sample
Control-Aware Wireless Sensor Network Platform for the Smart Electric Grid Powered By Docstoc
					16

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

Control-Aware Wireless Sensor Network Platform for the Smart Electric Grid
James Gadze
Tuskegee University, Tuskegee, AL, USA Summary
The communications needs of monitoring and control of the electric grid is traditionally catered for by wired communication systems. These technologies ensured high reliability and bandwidth but are however very expensive, inflexible and do not support mobility and pervasive monitoring. The communication protocols are Ethernet-based that used contention access protocols which result in high unsuccessful transmission and delay. The use of embedded wireless sensor and actuator networks for monitoring and control of the electric grid requires secure, reliable and timely exchange of information among controllers, distributed sensors and actuators. The exchange of information is over a shared wireless medium. However, wireless media is highly unpredictable due to path loss, shadow fading and ambient noise. Monitoring and control applications have stringent requirements on reliability, delay and security. The primary issue addressed in this paper is the impact of harsh power system environment on reliable and timely information exchange in wireless sensor and actuator networks. A combined networking and information theoretic approach was adopted to determine the transmit power required to maintain a minimum wireless channel capacity for reliable data transmission. We also develop a channel-aware optimal slot allocation scheme that ensures efficient utilization of the wireless link and guarantee delay. Various analytical evaluations and simulations are used to evaluate and validate the feasibility of the methodologies and demonstrate that the protocols achieved reliable and real-time data delivery in wireless industrial sensor networks.

information about the physical process. The prevention and control of such catastrophic events require the availability of sufficient and accurate real time information, which can only be achieved by pervasive sensing and decentralized control of the electric infrastructure. Traditionally, the communications network used in network control systems is wired communications networks. Wired communication technologies although matured and ensure high reliability and bandwidth are very costly, inflexible and do not support mobility and pervasive sensing. The communication protocols are Ethernet-based that used contention access protocols which results in high unsuccessful transmission and delay. What is missing from current approaches is a robust architecture that harnesses the power of wireless networking and communications to support unremitting real-time monitoring, and control in harsh power system operational environments. To attain this goal, it is imperative that the research community places more focus on using recent advances in wireless sensor and actuator networks to develop new sustained autonomic, decentralized, secured, and reliable control-aware networking protocols. An emerging class of wireless networks, embedded wireless sensor and actuator networks has potential benefits for real-time monitoring and control of the electric grid. Although these networks are resource constrained, they provide remote interaction with the physical world and therefore offer great promise for process information capture. The use of wireless technologies in electric infrastructure control seems very attractive because of the potential advantages of wireless solutions. The advantages of using wireless communications include lower hardware cost, lower installation cost, less disruptive installation, mobility, and ad hoc deployment. The use of embedded wireless networks for monitoring and control of the electric grid requires reliable and timely exchange of information among controllers, distributed sensors and actuators. The exchange of information in these networks is over a shared wireless medium. The physical characteristics of the

Key words: Embedded wireless sensor network, real-time monitoring and control, reservation- TDMA, cross-layer optimization
.

1. Introduction
In a deregulated electric power market and a post 9/11 world, the frequency of occurrence of high impact system wide disturbances is likely to increase due to the treat of terrorism. The frequent spread of local disturbances into disastrous system wide disturbances poses a serious problem. The disruption in the operation of the electric infrastructure is often due to the lack of

Manuscript received January 5, 2009 Manuscript revised January 20, 2009

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

17

power system environment, however, present a number of challenges to wireless network deployment. The first challenge is the error properties of unreliable wireless channels. These wireless channel errors result from phenomena such as multi-path, attenuation by obstacles and noise. The unreliable wireless channel has significant implications for the design of networking protocols and applications. There is the need for careful choice of physical layer technologies and appropriate Medium Access Control (MAC) protocol. Such a choice needs to be based on the knowledge of propagation and noise patterns within power system sites and the delay requirements of monitoring and control applications. The second challenge is that wireless medium is an open medium and presents network security issues. Also the use of wireless devices presents power consumption issues. Monitoring and control applications have stringent requirements on reliability, delay and security. The primary issue addressed in this work is the impact of the harsh power system environment on the reliable and timely delivery of critical operational data. The contribution of this work includes the characterization of the power system environment for wireless data communication, the use of combined networking and information theoretic approach to develop a procedure to determine the transmit power required to maintain a minimum wireless capacity for reliable data transmission, the development of a method to estimate the wireless channel quality in terms of probability of connectivity and the development of a QoS and channel-aware scheduling scheme. We propose the use of embedded wireless sensor networks to develop Local Smart Monitoring, Protection and Control (LSMPC) platform at the .substation and generation station levels. The LSMPCs are integrated into a multi-level hierarchical real-time monitoring and control platform as shown in fig. 1.

2. Related Work
Although a number of popular wireless technologies exist, few are adequately characterized for power system environment applications [10, 11, 12]. The existing IEEE 802.11/a/b/g/x family of standards offers high-speeds and interoperability. However, many of them were designed for office settings and consumer use [1, 12]. They are not designed for harsh industrial environments with high temperatures, and electromagnetic interference. They cannot therefore meet the delay and reliability requirements of control applications [3, 9]. The impairment supported by these wireless technologies is AWGN [10, 11]. The media access protocol used in IEEE 802.11b/g is Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) [10, 12]. IEEE 802.11b/g can support a delay of 3sec and transmit power at 40/50mW [10, 12]. The use of a contention-based medium access control protocol results in a large number of unsuccessful transmissions. The IEEE 802.15.1 Bluetooth operates at about 1mW and can support data rates between 1 to 3Mbps and a delay of about 10 sec [13, 14]. The range of operation is between 10 to 100 meters. The IEEE 802.15.4 standard is optimized for low power and low data rate communication among wireless devices within a short range where the delay and data rate are not critical issues [15, 16]. It can support a data rate in the range of 20kbps to 250kbps [16]. The transmission distance is expected to range from 10 to 75 m depending on the power output and the environmental characteristics. Bluetooth and ZigBee cannot support industrial multimedia traffic and lack realtime medium access control. The measure of success of a wireless network with regards to it impact on control systems is how the wireless network ensures secure, timely and reliable flow of control information [4, 5, 6, 7]. Existing wireless technologies do not meet all the communication needs of networked control systems. It is imperative to develop new wireless sensor protocols that can meet the reliability, security and delay requirements of networked control systems.

3. Power System Environment Model
The physical environment within power system sites is characterized by high electromagnetic radiation. The radiated electromagnetic fields from power switching devices and power lines results in impulsive noise. The layout and the components in transmission substations obstruct the path of Radio Frequency (RF) propagation. This leads to a path loss caused by a phenomenon called shadow fading. The loss in signal strength is modeled as log-normally distributed random variable. The quality of the wireless channel in power system environment is

Fig. 1 Multi-level decentralized control platform

18

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

impaired by shadow fading and joint AWGN and impulsive noise. The possible use of wireless data communication in shadow fading and noisy electrical environment would depend on the assessment of the power consumption of the wireless devices and the development of efficient multiple access algorithms.

communication between the embedded sensors and the intelligent controller is modeled as a multi-access wireless communication system
Queue

Data Source

Encoder/ Transmitter

Wireless Channel

Decoder/ Receiver

Data Sink

3.1. Joint AWGN and Impulsive Noise Model
The quality of the wireless channel in power system environment is impaired by shadow fading and noise. AWGN, a widely used model to study the performance of communication system, alone cannot accurately represent the entire noise which affects wireless data communication in a power system environment. This is because in a power system environment radiated electromagnetic fields from power switching devices and power lines result in impulsive noise. Impulsive noise has a negative impact on data transmission and therefore do not allow favorable wireless high speed communication at all times. In power system environment data transmission in data networks is impacted by joint AWGN and impulse noise. We adopt a simple model of the joint AWGN impulsive noise. The impulsive noise samples are with power β (k ) times that of AWGN and the probability of appearance of the impulsive noise samples is μ . The effect of impulsive noise on data transmission depends on how often the impulses occur and how strong they are.
Channel Quality Information

Fig. 3 Joint queuing and information theoretic system model

Impulsive Noise

The operation of the embedded wireless networked sensors is such that data in bits arrive and queued at the transmitter before being transmitted. The transmitter is assumed to have knowledge of the wireless channel conditions and the sensor node rate requirements. The data has a deadline within which it must be delivered to the intelligent controller. The physical environment within power system sites is characterized by high electromagnetic radiation. The radiated electromagnetic fields from power switching devices and power lines results in impulsive noise. The layout and the components in transmission substations obstruct the path of Radio Frequency (RF) propagation. This leads to a path loss caused by a phenomenon called shadow fading. The loss in signal strength is modeled as log-normally distributed random variable. The quality of the wireless channel in power system environment is impaired by shadow fading and joint AWGN and impulsive noise. The possible use of wireless data communication in shadow fading and noisy electrical environment would depend on the assessment of the power consumption of the wireless devices and the development of efficient multiple access algorithms.

AWGN

Fig. 2 Joint AWGN and impulsive noise model

3.3. Power and Delay Constrained Wireless Data transmission
This section assesses the delay, rate and transmitpower issues in joint AWGN and impulsive noise wireless channel. The wireless sensors used in developing LSMPCs are limited in power and bandwidth and operate in joint AWGN and impulsive noise environment. In a given time interval the wireless sensor nodes, based on the delay constraint imposed make rate or capacity demand on the multi-access wireless system based on their buffer content. During the same time interval a set of wireless channel conditions prevailed which determine the wireless capacity that can be supported. The transmission delay is reduced by transmitting at higher rates. However, transmitting at higher rates results in higher error probability and transmit-power. The delay, rate and transmit-power requirements are in conflict. It is imperative to analyze the delay, rate and transmit-power issues of wireless networked embedded

The time varying joint noise power two parameters is given by

p(k ) in terms of the

p(k ) = [(1 − μ ) + μβ (k )]N 0

(1)

where N 0 is the spectral noise density. The parameters determine the quality of the power system wireless channel. The quality of the power system wireless channel randomly changes with time and is different for different users.

μ and β (k )

3.2. System Model
We consider set of wireless sensor nodes communicating over a wireless medium and assumed the wireless sensor nodes and actuators are within a one-hop communication reach of the intelligent controller. The

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

19

systems. We adopt a combined networking and information theoretic approach to develop an algorithm to determine the transmit power required to maintain a minimum wireless channel capacity for reliable data transmission. We extend Shannon’s point to point theorem 1 to multi-user systems and assumed that whenever the multi-user wireless capacity is greater than the sum rate required by users then there is reliable transmission. The sum rate that can be supported simultaneously with arbitrary small error of probability is limited by the zero-outage and outage wireless capacities in joint AWGN and impulsive noise environment. We defined two notions of multi-user wireless channel capacity for set of channel conditions in joint AWGN and impulsive noise environment. The zero outage capacity is the minimum wireless capacity that is greater than the sum rate demand over all noise conditions. The zero outage capacity satisfies ⎡ ⎤ hi pi (2) ∑ Ri ≤ B log2 ⎢1 + ∑ [(1− μ ) + μ β (k )]N B ⎥ i∈S 0 ⎦ i i i ⎣ i∈S The second notion of wireless capacity is the outage capacity. The outage capacity is the rate vector which satisfies: ⎧ ⎫ ⎡ ⎤⎪ hi pi ⎪ (3) Pr⎨∑ Ri > B log2 ⎢1 + ∑ ⎥⎬ ≤ ε ⎪ i∈ ⎣ i∈S [(1 − μi ) + μi βi (k )]N0 B ⎦ ⎪ ⎩ ⎭ where N 0 is the noise spectral density,

γ i (k ) =

hi p i [(1 − μ i ) + μ i β i ( k ) ]N 0
α

(5)

⎛d ⎞ where hi = G⎜ 0 ⎟ ⎝ d ⎠

and G a constant that depends on

the antenna characteristics, d 0 is a reference distance, d is the transmission distance and exponent.
For active users

α

is the path loss

i = 1 to M
Ri Bavi
estimate the starting transmit power

Compute the wireless link reliability threshold, α i , given the bandwidth and the required rate

Given the average noise value

Psi

by equating the signal-to-noise ratio to the wireless link reliability threshold For

Pi
For

such that

Psi ≤ Pi ≤ Ppeak

k = 1 to T
Compute the signal-to-noise ratio, Condition If

γ i [k ] for the prevailing noise

γ i [k ] ≥ α i
C[k ]

The BS computes the zero outage capacity

If the zero-outage capacity is greater or equal to the sum rate

C[k ] ≥
Then

∑R
i =1

M

i

μ

is the

frequency of appearance of impulsive noise and β is the magnitude of the impulsive noise. The outage probability ε is defined as the fraction of time that the transmission sum rate is higher than the instantaneous mutual information.

Pi

is the transmit power required to maintain minimum wireless

capacity reliable data transmission Else increase If no

Pi

but within the specified range until

C [k ] ≥

∑R
i =1
M i =1

M

i

is true

Pi

within the range

Psi ≤ Pi ≤ Ppeak
Pi

results in C [ k ] ≥

∑R

i

3.4. Transmit Power Determination Algorithm
We developed a procedure to determine the transmit power of users at which the multi-user system can maintain enough capacity to meet the rate requirement using the zero-outage and outage capacities given in (2) and (3). For each individual node depending on its required rate and the prevailing wireless channel a wireless link reliability threshold is given by

Then use the outage capacity to determine

Pr { ∑ R i ≥ C [ k ]} = ε
i =1

M

α i = ⎜10 ⎜
⎝

⎛

0.3 Ri B

⎞ − 1⎟ ⎟ ⎠

(4)

4. Propose Reservation TDMA MAC Protocol
This section investigates the impact of the power system environment on wireless link allocation and access coordination. The problem is given the hostile nature of the power system environment to wireless communication and networking which media access scheme to choose to ensure efficient utilization of the wireless link and provide delay guarantees to time-sensitive users sharing an uplink

The signal-to-noise ratio is given by
Shannon theorem states that given a noisy channel with channel capacity C and information transmitted at a rate R , then if
1

R < C then there exists a code that allows the probability of error be made arbitrarily small

20

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

channel. The uplink channel is a time slotted joint AWGN and impulsive noise channel in a shadow fading environment. Guaranteeing delay in shadow fading and noisy environment requires the MAC layer to be aware of the application’s delay requirements and the physical layer characteristics of the wireless channel.

4.2 Slot Allocation Decision Factors
This section presents the decision factors used by the optimal slot allocation module in its decision making. One of the important components of the base station MAC is the slot allocation module. The function of this module is to allot slots to users based on some decision factors. 4.2.1. Wireless Channel Quality Factor The quality of a wireless link depends on the radio propagation parameters such as path loss, multi-path fading, shadow fading, interference and noise. In this work we assume the random fluctuation of the signal strength is an indication of the wireless channel quality. The fluctuation in signal strength is made up of 2 parts the deterministic and the random parts. The deterministic part given by

4.1 Overview of Reservation TDMA MAC Protocol
In a TDMA scheme, a user shares a multi-access communication channel and transmits its message in slots allotted to it by the BS. The communication resource of the channel is time slotted into a contiguous sequence of fixed-length frames. A frame is divided into request reservation and information transmission intervals. Each interval consists of number of slots where a slot has the same length as a data packet. Slots in the request reservation part are further divided into mini-slots. Reservation TDMA is a variation of the TDMA protocol that provides slots on demand type of service. A user which has data packets to send but has no assigned data slots sends a reservation packet during the reservation period to the BS to request data slots. The base station uses the reservation information as input to the slot allocation scheme to allocate number of slots to the user. The reservation packet specifies the user’s traffic characteristics, channel quality information and source buffer status. Access to the reservation mini-slots at the beginning of the uplink channel is either by contention or fixed assignment protocols. The frame structure used in the study is as shown in fig. 4

χ 1 = α10 log10 ⎜ ⎜

⎛ d ⎝ d0

⎞ ⎟ dB ⎟ ⎠

(6)

distance and α is the path loss exponent [][]. The random part, due to shadow fading, is chosen from a log-normal probability distribution function [10]

where d 0 is a reference distance, d is the transmission

χ2 = p(ψ dB ) =

1 2πσψ dB

⎡ ψ dB − μψ dB exp⎢− 2 2σψ dB ⎢ ⎣

(

)⎤
2

⎥ ⎥ ⎦

(7)

The distribution of ψ dB is Gaussian with mean standard deviation σ ψ dB . signal strength is given by

μψ

dB

and

The total fluctuation in the

Frame

χ = χ1 + χ 2

Modem preamble

Reservation interval

Data transmission interval

Fig. 4 Reservation TDMA frame structure

We defined threshold attenuation as the mean attenuation value at which very good reception of data is possible. The probability that the signal fluctuation exceeds the threshold is derived. This quantity is a measure of the wireless channel quality and therefore used as slot allocation decision making factor. The proper reception of data packet is impossible when the attenuation is above the threshold value. Thus when The threshold

In many networking situations each user is allotted a number of time slots during each time frame. An important function of the Base Station (BS) MAC is to allot slots to the users based on their demands. This work focuses on the development of a slot allocation scheme that bases its allocation decision on the queue state, traffic priority factor and the channel quality.

χ > χ th attenuation χ th

is a wireless system

parameter. The quality of the wireless channel of a user depends on the extent to which its signal attenuation departs from the threshold value. The probability that the attenuation is greater than or equal to the threshold value is denoted as

P( χ ≥ χ th )

Using (8) we can write

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

21

⎡ ( χ − μ )2 ⎤ (8) exp ⎢ − ⎥ dχ ∫ 2σ 2 ⎦ χ 2 σ 2π ⎣ The integral in (8) cannot be evaluated in a closed form and thus computed numerically using the Q function. Let y = χ − μ σ P( X ≥ χ 2 ) =
∞

1

class number is the higher its priority weight factor. In this work we considered three traffic classes. The arrival of each class of message is according to a Poisson process with rate λ k . The inter-arrival time between messages is considered to be exponentially distributed. The traffic class information is contained in the reservation request message send to the BS. The traffic class factor is one of the information used by the BS in deciding the number of data slots allocated to a traffic class.

χ2 − μ ⎞ ⎛ P⎜ y > ⎟ σ ⎝ ⎠

∫μ χ
2

∞

−

σ

⎡ y2 ⎤ exp ⎢ − ⎥ dy 2 ⎦ 2π ⎣ 1

χ −μ ⎞ ⎛χ − μ ⎞ ⎛ P⎜ y > 2 ⎟ = Q⎜ 2 ⎟ σ σ ⎝ ⎠ ⎝ ⎠

4.3 Slot Allocation Schemes
Existing slot allocation modules use linear proportional schemes based on the queue state and priority factor. The effect of the wireless channel on the slot allocation decision is neglected. This work considers the inclusion of the wireless channel quality in the slot allocation decision making. This section presents two slot allocation schemes. The two schemes include our proposed optimal slot allocation scheme and the linear prioritized proportional scheme that we compared our scheme to. 4.3.1. Optimal Slot Allocation Module The aim of the proposed scheme is to include the quality of the wireless channel in the slot allocation decision making so that fewer slots can be allocated during bad link states. The function of the optimal slot allocation scheme is to maximize the allocation of slots to each traffic class so as to minimize the queuing delay. The inclusion of the wireless channel quality in the slot allocation decision making results in efficient utilization of the wireless link. In wireless systems allocating time slots to meet user’s delay requirement and taking into consideration the underlying physical impairments is a complex problem. The slot allocation problem is seen as a single server serving k traffic classes as shown in fig. 5. In this work we consider the slot allocation for only uplink transmissions. The slot allocation module uses the buffer content, the traffic class factor and the channel quality information to allot slots to (serve) 3 different queues on frame by frame basis. The channel quality is specified in terms of the probability of connectivity.

But

χ 2 = χ th − χ 1
⎛ d = χ th − 10 α log 10 ⎜ ⎜d ⎝ 0
⎛ ⎜ χ th − 10 α log ⎛ d ⎞⎞ ⎟ ⎜ ⎜d ⎟⎟ ⎟ ⎝ 0 ⎠ ⎟ ⎟ ⎟ ⎠
z ⎞⎤ ⎟⎥ 2 ⎠⎦

⎞ ⎟ ⎟ ⎠

⎜ χ2 ⎞ ⎛ P⎜ y > ⎟ = Q⎜ σ ⎠ ⎝ ⎜ ⎜ ⎝

10

(9)

σ

Also
Q (z ) = 1 ⎡ ⎛ ⎢ 1 − erf ⎜ 2 ⎣ ⎝

Therefore the probability that the attenuation is greater than the threshold attenuation is given by
⎡ ⎛ ⎜ χ ⎢ ⎜ 1 ⎢ 1 − erf ⎜ = 2 ⎢ ⎜ ⎢ ⎜ ⎢ ⎝ ⎣ − 10 α log ⎛ d ⎞ ⎞⎤ ⎟ ⎜ ⎟ ⎜ d ⎟ ⎟⎥ ⎝ 0 ⎠ ⎥ ⎟⎥ ⎟⎥ ⎟⎥ ⎠⎦

th

10

(10)

P out

σ

2

We use Pout as a measure of connectivity between the wireless sensor nodes and the BS. The connectivity measure is used to decide the number of data slots allocated to a node. Instead of absolute no transmission during bad states we develop an allocation scheme that allocates fewer slots during the bad state. 4.2.2. Traffic Class Factor Multi-priority data transmission can be achieved by assigning different weight factor wk to the different priority classes. We assumed that all components in the power system have remote communication capability and can be used as a smart monitoring node. Each smart monitoring node generates messages which can be classified into n different traffic priorities. The different traffic classes have different delay requirements and are queued in separate buffer. The messages are transmitted over a multi-access wireless channel. The lower a traffic

x1i +1
i x 2+1

s1i +1
i s 2+1

N i +1

i x 3+ 1

i s3+1

Fig. 5 Slot allocation model

22

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

Due to propagation and processing delays we assume that data slot requests made in a given frame are not granted in the same frame period. The data slots requested in frame i are used data transmission in frame i + 1

We solve (12) using the Lagrange method for constrained optimization. ⎜ ⎟ {xi+1 − si+1}2w cQ + λ ⎛ Ni+1 − si+1 ⎞ (13) L(si+1, λ) =
jk

∑∑
j k

jk

jk

k

jk

⎜ ⎝

∑∑
j k

jk

⎟ ⎠

Frame i − 1

i

(i + 1)

+ ∂ L s ijk 1 , λ

(

∂s

Fig. 6 Frame cycles

+ ∂ L s ijk 1 , λ

(

i +1 jk

) = 2 (x
)=
N

i +1 jk

+ − s ijk 1 w k cQ jk − λ = 0

)

(14) (15)

i +1

∂λ

−

∑∑s
j k

i +1 jk

= 0

Let x jk denote the estimated number of node j ’s traffic class k packets awaiting transmission at the beginning of frame i . Also let r jk denote the expected number of
i

i

+ + s ijk 1 = x ijk 1 +

λ
2 w k cQ jk

(16)

i and s ijk the number of slots allotted to node j ’s traffic class k in i frame i . The expected number of packet arrival r jk is the
node j ’s packets that arrive during frame product of the mean arrival rate and the frame duration.

Substituting (16) in (15) we have

∑ ∑ ⎜x ⎜
j k

⎛ ⎝

i +1 jk

+

λ
2 w k cQ
jk

⎞ ⎟ = N ⎟ ⎠

i +1

N i is the available number of slots for use in frame i . The estimated number of packets awaiting transmission at the beginning of frame i + 1 is
given by

⎛ ⎜N ⎜ λ = 2⎝

i +1

−

∑∑
j k

⎞ + x ijk 1 ⎟ ⎟ ⎠
jk

∑∑
j k
i +1

1 w k cQ
⎞ ⎟ ⎟ ⎠ ⎞ ⎟ ⎟ ⎠

x

i +1 jk

= x +r −s
i jk i jk i +1 jk

i jk

(11)
+ + s ijk 1 = x ijk 1 −

For given x

and N

i +1

⎛ ⎜ (∑ ⎜ ⎝ j w k cQ

∑
k jk

+ x ijk 1 ) − N

(17)

the goal is to find an optimum

slot allocation strategy based on the traffic class, and channel quality so that the quantity

(x

i +1 jk

−s

i +1 jk

) is

⎛ ⎜∑ ⎜ j ⎝

∑
k

1 w k cQ

jk

minimized. We are of the view that reducing the expected number of packets awaiting transmission translates into reducing the queuing delay. The optimal slot allocation problem is posed as a constrained optimization problem. For node j the objective function for traffic class k is given by
min i +1
s jk

It is seen from (17) that there would be no need for the optimization if the total number of packets awaiting transmission at the beginning of a frame is equal to the total number of slots available. This is because
⎛ ⎜ (∑ ⎜ ⎝ j

∑
k

+ x ijk 1 ) − N

i +1

⎞ ⎟ = 0 and all the requested slots ⎟ ⎠

can be honored. Also the higher the priority factor wk and the wireless channel quality

∑ ∑ {x
j k

i +1 jk

+ − s ijk 1 w k cQ

}

2

cQ jk the higher the allocated

jk

s.t.

∑∑s
j k

(12)
i +1 jk

number of slots s jk . The wireless channel quality factor is specified in terms of the link outage probability and is given by (18) cQ jk = 1 − Poutjk Based on the path exponent and the degree of shadowing Poutjk in (18) can be obtained from fig. 10 and fig.11 4.3.2. Prioritized Proportional Allocation Scheme Linear proportional scheme and its variations are used by many researchers because of their simplicity. However, these proportional schemes cannot be used in all

i +1

= N

i +1

where N is the number of available data transmission slots for use in frame i + 1 , wk is the traffic class priority factor and cQ jk is the channel quality factor of node j ’s traffic class k .

i +1

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

23

T ran sm it p o wer (m W )

operational environments such as the power system. This is because these schemes do not take the channel quality into consideration in slot allocation decision making. The linear prioritized proportional slot allocation is based on + w k x ijk 1 (19) + s ijk 1 = N i +1 ∑ ∑ w k x ijk+1
j k

18 16 14 12 10 8 6 4 2 0 0 0.2 0.4 0.6 0.8 1 1.2 minimum wireless channel capacity (Mbps)
d10 d30 d50 d70 d90

+ x ijk1 is the estimated number of node j ’s traffic stream

k packets at the beginning of frame (i + 1) [8]. The
mean delay is used to compare the optimal slot allocation scheme and the linear proportional schemes.

5.0 Numerical Results
A 4-user system is used to determine the transmit power required to maintain enough wireless system capacity for reliable data transmission in a joint AWGN and impulsive noise environment. The joint AWGN and impulsive noise parameters used are μ = 0.7 and β max = 10 N 0 . The different noise conditions were randomly generated using MATLAB. The following scenarios were considered in the analysis of the 4-user system: a. Sensor nodes transmitting at 16ms average delay constrained, at system bandwidth of 400KHz and at distances of 30m, 50m, 70m and 90m from the intelligent controller; b. Sensor nodes transmitting at 4ms average delay constrained, at system bandwidth of 400KHz and 1000KHz and at distances of 30m, 50m, 70m and 90m from the intelligent controller, For an average delay constraint of 16ms and system bandwidth of 400 KHz the respective transmit power required to maintain a minimum capacity of 1Mbps for a reliable data transmission is 8.2mW at 30m, 11mW at 50m, 13mW at 70m, and about 14mW at 90m.
Fig. 7 Transmit power for 16ms delay constraint at 400KHz and varying distances

For an average transmission delay constraint of 4ms and system bandwidth of 400 KHz, the respective transmit power required to maintain a minimum capacity of 900Kbps for a reliable data transmission is 250mW at 30m, 325mW at 50m, 445mW at 70m and 500mW at 90m.
600 500 Transmit power (mW) 400 300 200 100 0 0 0.2 0.4 0.6 0.8 1 minimum wireless capacity (Mbps)
d10 d30 d50 d70 d90

Fig. 8 Transmit power for 4ms delay constraint at 400 KHz and varying distances

For an average delay constraint of 4ms and system bandwidth of 1 MHz, the respective transmit power required to maintain a minimum capacity of 2.5Mbps for a reliable data transmission is 22mW at 30m, 30mW at 50m, 37mW at 70m and 45mW at 90m.

24

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

60 50
Transmit power (mW)

40 30 20 10 0 0 0.5 1 1.5 2 2.5 3 3.5 minimum wireless capacity (Mbps)

d10 d30 d50 d70 d90

Fig. 9 Transmit power for 4ms delay constraint at 1MHz and varying distances

The outage probability is estimated for attenuation scenarios of path loss exponents of 2.5 and 3.5. The shadowing cases considered are standard deviation of 6, 8 and 10. The results using (4-5) are shown in figures 17 and 18

The performance metric used to compare the optimal slot allocation scheme to the linear proportional schemes is the mean delay. The traffic-class factors used in this work are 0.6, 0.3, and 0.1 for classes 1, 2, and 3 respectively. Assuming perfect channel conditions during each frame, when there are traffic class k packets to send, class k is guaranteed to receive a fraction of the available slots based on its weight. he OPNET setup comprises a traffic source, a buffer, and a server. Three traffic classes were generated using standard OPNET source generators. The traffic parameters configured are the average data rate, the packet size, and the function of the inter-arrival distribution. The buffer is modeled as a passive queue and forwards packets only upon receiving a request from the server. The slot allocation module which acts as the server allots packets to the respective classes as described in section IV. The setup is used to compare the performance of the optimal slot allocation scheme to the linear proportional scheme. It is not used to evaluate the performance of reservation TDMA protocol. For all traffic classes the mean delay performance of the linear proportional scheme is better than that of the optimal slot allocation scheme as seen in figures 21 to 23. This in- my view- is due to the fact the linear prioritized proportional scheme does not consider the wireless channel quality in its decision making. The optimal slot allocation scheme therefore gives a more realistic result.

Fig. 1 Outage probability for path loss exponent of 2.5

Fig. 13 Mean delay for traffic class

Fig. 2 Outage probability for path loss exponent of 3.5

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

25

allocation scheme therefore gives a more realistic result. The results from this work show that wireless data networks can be used for control functions in a harsh power system environment. However, the use of wireless data networks for adaptive protection requires more work to improve the delay performance.

References
[1]. Egea-Lopez, E.; Martinez-Sala, A. and Vales-Alonso, J, “Wireless Communications Deployment in Industry: A Review of Issues, Options and Technologies,” Computers in Industry (Elsevier), 56(1): p. 29 - 53. 2005. [2]. Gungor, V.C. and Lambert, F. “A Survey on Communication Networks for Electric System Automation,” Computer Networks Journal (Elsevier), Vol.50: p. 877-897. 2006. [3]. Pellegrini, F.D.; Miorandi, D.; Vitturi, S. and Zanella, A., “On the use of Wireless Networks at Low Level of Factory Automation System,” In IEEE Transaction on Industrial Electronics, Vol.2: p. 129 - 143. 2006. [4]. Willig, A.; Matheus, K. and Wolisz, A. “Wireless Technology in Industrial Networks,” In IEEE Proceedings of Technology of Networked Control Systems. 93(6): p.1130-1151, 2005. [5]. Caro, D. “Wireless Networks for Industrial Automation,” ed. 2. 2005, ISA press, [6]. Koumpis, K.; Hanna, L.; Andersson, M. and Johansson, M. “Wireless Industrial Control and Monitoring beyond Cable Replacement,” In the proceeding of PROFIBUS International Conference. UK. June, 2005. [7]. Dzung, D.; Apneseth, C.; Scheible, G. and Zimmermann, W. “Wireless Sensor Communication and Powering System for Real-Time Industrial Applications,” In WIP Proceeding of the 4th IEEE Workshop on Factory Communication Systems. (WFCS 2002). Sweden. August, 2002. [8]. Bertsekas, D.P. and Gallager, R., “Data Networks,” ed. 2. 1992, New Jersey: Prentice-Hall, Inc. [9]. Ericsson, G.N., “Classification of Power Systems Communications Needs and Requirements: Experiences from Case Studies at Swedish National Grid,” In IEEE Journal on Power Delivery, 17(2): p. 345 – 347, April 2002. [10]. Goldsmith A, “Wireless Communication,” New York: Cambridge University Press. 2005. [11]. Rappaport, T.S. “Wireless Communications: Principles & Practice,” Prentice Hall. 2002. [12]. Brevi, D.; Mazzocchi, D.; Scopigno, R.; Bonivento, A.; Calcagno, R. and Rusina, F. “A Methodology for the Analysis of 802.11a Links in Industrial Environments,” In Proceedings of IEEE International Workshop on Factory Communication Systems, p. 165-174, Italy, 2006. [13]. Bisdikian, C. “An Overview of the Bluetooth Wireless Technology,” In IEEE Communications Magazine, p. 8694. 2001. [14]. Haartsen, J.C. “The Bluetooth Radio System,” In IEEE Personal Communications, p. 28-36. February, 2000. [15]. Lu, G.; Krishnamachari, B. and Raghavendra, C. “Performance Evaluation of the IEEE 802.15.4 MAC for Low-Rate, Low-Power Wireless Networks,” In IEEE

1
Fig. 13 Mean delay for traffic class 2

Fig. 14 Mean delay for traffic class 3

6. Conclusion
A combined networking and information theoretic approach was adopted to determine the transmit power required to maintain enough wireless channel capacity for a reliable data transmission in a shadow fading and joint AWGN and impulsive noise environment. The analytical tools developed in this paper can help designers determine the transmit power levels required of embedded wireless devices to support a given bit rate in a harsh outdoor industrial environment. This is critical since the transmit power have impact on the power consumption of wireless devices. An optimal slot allocation scheme that bases its allocation decision making on the buffer content, the traffic class factor, and the wireless channel quality is developed. A method to estimate the wireless channel quality in terms of the probability of connectivity between users and the base station is developed. The optimal slot allocation scheme allots fewer slots during bad link states, instead of allotting no slots in bad wireless link states. The optimal slot

26

IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.1, January 2009

International Performance Computing and Communications Conference, p.701-706. April, 2004. [16]. Callaway, E.; Gorday, P.; Hester, L.; Gutierrez, J.A.; Naeve, M.; Heile, B. and Bahl, V. “Home Networking with IEEE 802.15.4: Developing Standards for Low-Rate Wireless Personal Area Networks,” In IEEE Communications Magazine, p. 70 - 77. 2002

James Gadze received the B.S. degree from the University of Science & Technology, Kumasi, Ghana in 1991 and the M.S. degrees in Electrical Engineering from Tuskegee University, Tuskegee, AL in 2003, and the PhD degree from Florida International University, Miami, FL in 2007. He is currently an Assistant Professor in the Electrical Engineering Department, at Tuskegee University. His research interests include Wireless Networking and Communications, modeling and performance evaluation of communication networks, security in computer and wireless industrial networks, medical Information Technology (telemedicine) and application of wireless data networks in electric distribution automation and control. He is a student member of IEEE.


				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:45
posted:7/1/2009
language:English
pages:11
Description: The communications needs of monitoring and control of the electric grid is traditionally catered for by wired communication systems. These technologies ensured high reliability and bandwidth but are however very expensive, inflexible and do not support mobility and pervasive monitoring. The communication protocols are Ethernet-based that used contention access protocols which result in high unsuccessful transmission and delay. The use of embedded wireless sensor and actuator networks for monitoring and control of the electric grid requires secure, reliable and timely exchange of information among controllers, distributed sensors and actuators. The exchange of information is over a shared wireless medium. However, wireless media is highly unpredictable due to path loss, shadow fading and ambient noise. Monitoring and control applications have stringent requirements on reliability, delay and security. The primary issue addressed in this paper is the impact of harsh power system environment on reliable and timely information exchange in wireless sensor and actuator networks. A combined networking and information theoretic approach was adopted to determine the transmit power required to maintain a minimum wireless channel capacity for reliable data transmission. We also develop a channel-aware optimal slot allocation scheme that ensures efficient utilization of the wireless link and guarantee delay. Various analytical evaluations and simulations are used to evaluate and validate the feasibility of the methodologies and demonstrate that the protocols achieved reliable and real-time data delivery in wireless industrial sensor networks.