IEEE Wireless Communications - April 2017 - page 16

IEEE Wireless Communications • April 2017
of smart grid and non-smart grid devices within
the LTE cell, the random access attempts to set up
the individual connections might collide, resulting
in failed random access attempts.
•The bottleneck in the communication phase
(that is, after successfully finishing the random
access phase) when a large number of smart grid
devices would like to push their measurement data
toward smart grid applications. Since many uplink
messages are contending for the limited LTE uplink
resources, the maximum delay that some measure-
ment reports might experience (for example, due
to the waiting time until a device is granted an UL
transmission resource) could exceed the require-
ments of smart grid applications such as DSSE.
In the following section we analyze the achiev-
able LTE uplink delay of SM, PMC, and
devices. Hereafter, we consider specifically the
bottlenecks in the random access phase and
describe a proposal for ensuring reliable random
access. The analyses are based on simulation
models described in [10–12].
In particular, we investigate and quantify the per-
formance of LTE for different possible deployment
scenarios in terms of the number of devices, their
type, and the amount of reserved LTE uplink physi-
cal radio resources, as configured by the LTE cellu-
lar network operator for supporting the smart grid
data traffic. For the following studies, we consider
the ratio R between the number of
PMU devices
over the number of PMC/SM devices. The PMC
and SM devices are considered jointly, since their
uplink traffic patterns are similar. We consider ratios
of R = 1/10 and R = 1/3, where the former rep-
resents a scenario with moderate DER penetration
where not too many
PMU devices are needed,
and the latter represents a heavy DER penetration.
The measurement report sizes from the PMC/SM
PMU devices are assumed to be 70 Bytes and
560 Bytes, respectively, as the
PMU devices report
more detailed power measurements. We assume
that the reporting interval of all types of measure-
ment reports is 1 s, as mentioned earlier.
The analysis in this section is focused on the radio
part of the LTE uplink transmission, i.e., between
the end-node and the LTE base station. This is
because it is assumed that this is the most critical
part of the end-to-end path between the measure-
ment device and the smart-grid publish-subscribe
server and applications, which are typically con-
nected through high-speed network infrastructure,
as indicated in Fig. 2.
In LTE, the uplink radio resources are orga-
nized in time-frequency blocks, also called physi-
cal resource blocks (PRBs), with duration of 0.5 ms
and 12 consecutive OFDMA frequency sub-carriers.
The shortest uplink radio transmission duration is
1 ms, also known as the transmission time interval
(TTI), which consists of two consecutive PRBs in
the time domain. The PRB allocation per individual
device and per TTI (or per block of TTIs) is done
by the scheduler located in the eNB. The assumed
scheduling approach for this analysis is fair fixed
assignment (FFA) [13], where in every TTI a device
is randomly selected from a number of devices will-
ing to transmit and it is allocated a fixed number of
PRBs per device. Depending on the signal-to-inter-
ference-plus-noise ratio (SINR) as experienced by
the device on the allocated PRBs, an appropriate
modulation and coding scheme is selected for the
transmission. This, in turn, determines the amount
of data (in bits) that can be transferred, and finally
the number of TTIs needed to transmit the measure-
ment report by the devices. Additionally, as there
is a limited number of reserved uplink LTE radio
resources (such as a limited total number of PRBs)
for the transmission of the measurement reports,
not all active devices in the LTE cell can begin their
transmission within one TTI. As a consequence, a
number of active devices have to wait until they
receive an uplink transmission grant, resulting in a
certain amount of waiting time. Then the maximum
uplink LTE delay is the sum of the transmission time
and the waiting time. For more details regarding the
analysis of the maximum uplink LTE delay, the read-
er is referred to [10].
In order to quantify the maximum LTE uplink
delay, Monte-Carlo system-level simulations were
performed for an urban environment with an
increased number of total smart grid devices per
LTE cell.
In Fig. 3a and Fig. 3b the maximum LTE delay
results are presented for R = 1/10 and R = 1/3,
respectively, for a fixed number of two PRBs
assigned per device. It can be seen that even for
a very small total number of devices, the maximum
uplink LTE delay is about 20 ms and 200 ms, respec-
tively. This is the intrinsic delay of transmitting the
measurement report (including any re-transmis-
sions), since for a low number of devices the wait-
ing time is practically zero.
As the number of devices increases, the max-
imum uplink delay remains constant with zero
waiting time, only for when the whole bandwidth
(10 MHz or 50 PRBs in this case) is reserved for
the smart grid data. As the number of available
PRBs is decreased to 20 or six, which is more real-
istic in shared networks, the max delay rapidly
increases due to the waiting time incurred at the
devices until they receive a scheduling grant for
uplink transmission. The only exception here is
the max delay for the
PMU in Fig. 3a, where the
max delays for 50 PRBs and 20 PRBs are equal
and stays constant up to 4000 devices per LTE
cell (i.e. these two curves overlap). If the maxi-
mum delay requirement for the real-time applica-
tion is 1 s, for instance, then in order to achieve
this requirement for, e.g., up to 4000 devices per
LTE cell, the operator is required to reserve 50
PRBs (or an entire 10 MHz LTE carrier), which
might not be an economically viable solution. For
a more realistic amount of reserved resources (for
example six PRBs) the achievable maximum delay
is 6s or 3s for R = 1/3 or R = 1/10, respectively.
When a measurement device wants to transmit
a report, it will need to change state from idle to
connected in the LTE network, through the ARP
[14]. This procedure has several steps in which
failures can occur, especially in case of preamble
collisions [12]. The collision probability increas-
es with the number of active devices [11]. This
means that in a traditional LTE network, a large
The existing LTE cellular
networks carry various
types of traffic, e.g.
mobile broadband
traffic, and are expected
to additionally serve
the traffic originated by
many IoT applications
including the smart grid
applications such as
DSSE and demand-re-
1...,6,7,8,9,10,11,12,13,14,15 17,18,19,20,21,22,23,24,25,26,...132
Powered by FlippingBook