Content uploaded by Ashutosh Balakrishnan
Author content
All content in this area was uploaded by Ashutosh Balakrishnan on Feb 07, 2022
Content may be subject to copyright.
Energy Sharing based Cooperative Dual-powered
Green Cellular Networks
Ashutosh Balakrishnan1, Swades De1, and Li-Chun Wang2
1Department of Electrical Engineering and Bharti School of Telecommunication, IIT Delhi, New Delhi, India
2Department of Electrical and Computer Engineering, National Chiao-Tung University, Hsinchu, Taiwan, ROC
Abstract—Solar enabled and grid connected “dual-powered”
base stations (BSs) have developed as a cost effective solution
to network operators. While these networks prevent energy
outages and are effective in providing seamless operation for
catering to the user quality of service (QoS), they still rely
significantly on the power-grid, generating carbon footprint.
In this paper, we propose a profitable energy sharing based
cooperative framework to reduce the grid energy consumption
and to facilitate better utilization of network energy. Traffic-
energy imbalances in a dual-powered network often result in
some BSs being energy deficient due to spatio-temporal variation
of traffic. In such an event, it is proposed that the energy deficient
BS, through the proposed energy cooperation (EC) framework,
interact with the other networked BSs and requests them to share
the deficient energy or a portion of it at a much lower price than
the grid price via grid itself. We compare the proposed EC model
with a without energy cooperation (WEC) model, where the BSs
do not interact with one another and release energy above their
storage capacity only to the power-grid for selling. Our results
demonstrate that for a two BS network, the EC model provides
a significant reduction in grid consumption up to 50% with a
38% gain in operator’s revenue at 80% traffic skewness.
Index Terms—Dual-powered base stations, energy harvesting,
green communication, energy cooperation, revenue analysis
I. INTRODUCTION
The past decade has witnessed pertinent efforts towards
developing off-grid base stations (BSs) that are effective in
reducing the energy consumption [1], [2]. While they are
effective in mitigating the carbon footprint generated by the
erstwhile diesel operated BSs, they are not cost effective
from the operator’s perspective [3], [4]. This is mainly due
to the additional costs involved in solar dimensioning a BS
to avoid sporadic energy outages. This has resulted in efforts
towards development of dual-powered BSs [5], [6] as a cost
effective alternative. While the dual-powered BSs have been
successful in reducing the carbon footprint generated by the
BSs, they still rely on power-grid for providing seamless
operation. In this work, we propose a profitable innovative net-
work operation framework to reduce the grid consumption of
dual-powered networks through proper utilization of network
energy, without increasing the capital expenditure (CAPEX).
A. Related works
The advent of the Internet of Things (IoT) has led to a
surge in user traffic and their corresponding data requirements
[7]. The proliferation of the IoT alongside the roll out of 5G
communications is expected to increase the number of BSs [8]
in a network to cater to the user Quality of Service (QoS), thus
increasing the network energy consumption. These BSs being
energy intensive devices, are estimated to generate about 60%
of the carbon footprint in the Information and Communication
Technology sector [9]. Recent studies focusing on reducing
the carbon emissions due to the BSs involve optimal resource
allocation [10], [11] and traffic management techniques, like
dynamic BS ON/OFF [12] or cell zooming methods [13].
Powering the BSs with renewable energy sources like solar has
gained traction in recent studies [1] - [5]. While purely solar-
enabled BSs have been very effective towards envisioning
green communication, these frameworks have been estimated
to be incurring significant CAPEX to the operator [3].
Lately, there have been efforts to design dual-powered BSs
[5], [6], [14] which not only ensures a fair reduction in carbon
footprint, but also provides a cost effective framework for
seamless operation of the BSs without energy outages. The
framework in [5] presents a coverage adjustment based frame-
work to deal with the impact of traffic-energy imbalances
in a network and study the network operator’s revenue. The
studies in [6] and [14] deal with energy cooperation among
BSs in dual-powered networks via additional infrastructure
like power lines among the BSs. These power lines result in
additional CAPEX and maintenance costs to the operator.
In this paper, to deal with traffic-energy imbalance in
cellular networks, we propose a novel energy cooperation
scheme without the need for additional inter-BS power lines
infrastructure and also without any computationally-intensive
coverage adjustment requirement of the BSs, and analyze the
impact of energy cooperation on the operator’s revenue.
B. Motivation and contributions
Designing dual-powered networks are challenging due to
the stochastic nature of solar-harvest and user mobility in the
network. This spatio-temporal variation in user traffic density
and green energy leads to traffic-energy imbalances across the
network [15]. These traffic-energy imbalances result in some
BSs having more energy than required to serve its load or may
result in some BSs being energy deficient. In this paper, our
primary motivation is to propose profitable strategies to reduce
grid consumption by improving the network energy utilization
in dual-powered networks. This is expected to be achieved
via an energy sharing based cooperative framework, without
increasing the operator’s CAPEX. In this process, we also
look to explore new cost metrics in addition to the traditional
metrics and analyze their impact on the operator’s revenue.
(a) (b)
0
β(t)
βcr
βmax
Solar Panels
Base station
EMU
Battery storage
Eb(t)
Hb(t)
BSb
(c)
Figure 1: Representation of (a) dual-powered wireless communication network, (b) normalized network traffic, (c) energy dynamics in a solar-powered BS.
The main contributions in this work are as follows: (i) To
reduce the grid energy consumption and to facilitate better
utilization of network energy, we propose an energy sharing
based cooperative cellular network in which the networked
BSs can share energy. (ii) It is proposed that the BSs will share
energy amongst one another if needed via the grid itself. The
cost of sharing is proposed to be much lower than the cost
of selling or purchasing energy from the grid. This is done as
an incentive for the operator to share energy rather than sell
the energy and earn revenue, thus reducing the operational
expenditure (OPEX). (iii) Individual user power requirements
have been computed through channel inversion technique to
meet the user QoS. (iv) The proposed energy cooperation
(EC) model is compared with a without energy cooperation
(WEC) model. Both the models are compared in terms of
the annual network energy consumption and profitability to
the operator. (v) It is observed that the proposed EC model
provides significant grid energy reduction of about 50% with
a profit gain of around 38% at 80% traffic skewness.
C. Organization
The layout of this paper is as follows. Section II introduces
the system model in detail. Section III deals with the physical
resource allocation while Section IV presents the proposed
network operation strategies. In Section V, we compare the
proposed network operation strategies and discuss the obser-
vations and inferences. Section VI concludes the paper.
II. SY ST EM MO DE L
In this work, we consider the downlink of a dual-powered
wireless communication network subjected to skewed user
traffic, as shown in Fig. 1(a). We consider a set of active
users Ufollowing a Poisson point process (PPP) of density
λin a fixed area A, being served by two BSs. Let these
solar-enabled and power-grid connected BSs be represented as
B={BSA, B SB}. The users are assumed to displace within
this fixed area and not move out of it. It is also assumed that
each user is being served by only one BS, even if the user falls
under the coverage of both the BSs. The user-BS association
is performed on the basis of maximum received power level
criterion, with the initial BS downlink powers assumed equal
and represented as P={Pb} ∀ b∈ B s.t. 0≤Pb≤Pmax .
5 10 15 20
Time (Hour)
0
0.2
0.4
0.6
0.8
1
Normalized Traffic Intensity
BSA
BSB
Figure 2: Traffic distribution among BSs at 60% skewness.
Table I: Traffic skewness levels
Skewness levels: 0%20%40%60%80%
ρA(t)50% 60% 70% 80% 90%
ρB(t)50% 40% 30% 20% 10%
K-Means clustering is used to compute the optimum BS
coordinates using the user locations generated via the PPP.
The number of users associated with each BS is represented
as U={Ub} ∀ b∈ B, with the initial cell radius given as
R={RA, RB}. We discuss the skewed traffic profile and the
energy harvesting profile in the upcoming sub-sections.
A. Traffic profile
The BSs are assumed to be subjected to skewed user traffic,
in the sense that the user intensity varies spatio-temporally
throughout the day. The normalized net traffic intensity ρ(t)
over the area Ais given in Fig. 1(b). Traffic skewness is
defined as the amount of traffic imbalance in a BS. Mathemat-
ically, traffic skewness γ=|ρA(t)−ρB(t)|/(ρA(t) + ρB(t)),
where ρA(t)and ρB(t)are the traffic intensities at BSAand
BSBrespectively, with ρ(t) = ρA(t) + ρB(t). To induce
spatio-temporally varying skewed traffic across the network,
we break the 24 hour day duration into six windows of 4hours
each. The skewness levels along with their traffic distributions
considered for the current two BS network is shown in Table
I. For instance, in an hour t, at 60% skewness, either of the
BS will be subjected to 80% of the net traffic given in Fig.
1(b), while the other BS will be subjected to the remaining
20% of the net traffic. An illustration of the skewed traffic
profile generation is shown in Fig. 2.
B. Energy harvesting model
The BSs being dual-powered are solar-enabled in addition
to being connected with the power-grid. The energy dynamics
of a solar-enabled BS, showcasing the direction of energy
flow in a solar-BS is illustrated in Fig. 1(c). It is assumed
that each BS is equipped with κnumber of batteries, each
having capacity Bcap and NSunit rated solar panels. Thus,
the maximum storage capacity with each BS can be given
as βmax =κBcap. Further, to prevent energy outage, it is
constrained that the battery level cannot go below a critical
level denoted as βcr =δκBcap, with δdenoting the depth of
discharge of the battery storage. Thus, the battery level of any
BS is given as βb(t)∈[βcr, βmax ]∀b∈ B. Mathematically
βb(t) = βb(t−1) + Hb(t)−Eb(t),∀b∈ B (1)
where, Hb(t)denotes the hourly energy harvested by NSunit
rated solar panels enabled at BS band Eb(t)denotes the
hourly energy consumption by BS b. We have obtained the
hourly solar-harvest data by feeding the annual solar radiation
data provided by National Renewable Energy Lab, in the
System Advisor Model [16]. A BS is termed to be energy
surplus at hour t, if βb(t)≥βmax and energy deficient at
hour t, if βb(t)≤βcr as depicted through Fig. 1(c). Following
section describes the resource allocation strategy in detail.
III. RESOURCE ALL OC ATIO N
We assume that all BSs in the network operate with full
frequency reuse, i.e., each BS has full access to the entire
bandwidth spectrum BW . It is further assumed that a user
gets a bandwidth share of BWub =BW/Ub∀b∈ B for the
entire duration of connectivity, without any sharing with any
other active user. Let the data rate achievable by a user uat a
distance dub when associated with BS bbe given as rub(t) =
BWub log2(1 + SINRub (t)). Here, SINRub(t)represents the
signal-interference-to-noise ratio (SINR) observed by user u
when associated with BS band is given as
SINRub(t) = Pub(t)gub
(σ2d2
ub +Pb′∈BPu′b′(t)gu′b′d−2
u′b′).(2)
Pub(t)refers to the power allocated to a individual user u
by BS bat hour t,gub refers to the channel gain between
user uand BS b,b′∈ B − {b}, and σ2refers to the
additive white Gaussian noise power spectral density. The term
Pb′∈BPu′b′(t)gu′b′d−2
u′b′∀b′∈ B − {b}is called inter
cell interference (ICI). In this work, we have assumed that
all the user traffic belongs to a single service class requiring
a minimum rate requirement r0. Since each user mandates a
minimum data rate requirement of r0, it becomes imperative
for the service provider to meet the user QoS demand through
the BSs. The user QoS requirements can be expressed as,
P(rub(t)≥r0)≥p0(3)
where, p0represents the probability of meeting the user QoS.
In order to obtain the individual power requirement for each
user, we solve (3) as follows.
P gub ≥exp (r0ln 2/BWub −1) d2
ubσ2+ICI
Pub(t)!≥p0
or, exp −exp (r0ln 2/BWub −1) d2
ubσ2+ICI
Pub(t)!≥p0
(4)
giving, Pub(t)≥exp (r0ln 2/BWub −1) d2
ubσ2+ICI
ln(1/p0).
(5)
In the above analysis, we have assumed the channel to
be Rayleigh distributed with the corresponding channel gain
being exponentially distributed with unit mean. Thus, the BS
downlink transmit power level can be calculated as,
Pb(t) =
Ub(t)
X
u=1
Pub(t), s.t. 0≤Pb≤Pmax .(6)
The hourly BS energy consumption is then computed as
Eb(t) = NT RX (Pb(t) + P0)∀b∈ B [17] where, NT RX
denotes the number of transceivers per BS, Pb(t)denotes the
hourly varying BS load, and P0denotes the static BS power
requirement. After computing the BS downlink transmit power
level, we further analyze the proposed network operation and
revenue computation strategies in the upcoming section.
IV. NET WO RK OP ER ATIO N STR ATEG IE S
In this section, the network is operated through two network
operation strategies namely, the without energy cooperation
(WEC) model and the energy cooperation (EC) model for a
two BS network. Further, the revenue earned by a network
operator is computed considering all the cost metrics.
A. Without energy cooperation model
The WEC model involves the BSs trying to maintain the
QoS of maximum possible users associated without any flex-
ibility in cell coverage. Being subjected to spatio-temporally
varying user intensities, some BSs in the network may be
energy surplus while some BSs may get energy deficient.
From the network operator’s perspective, in this strategy,
revenue can be earned by selling the hourly excess energy
εback to the grid in addition to serving users. The operator
also incurs the cost of buying energy from the power-grid
(referred as OPEX) in the scenario where the BS gets energy
deficient and there is a need to prevent an energy outage. The
WEC model is presented in Part I of Algorithm 1.
For each BS b, depending on the initial battery level βini
b,
energy budget Ωbud
bfor each BS is computed in Step 4. If a
BS bis energy surplus at a given hour t, then the operator
can sell εb(t)=(βb(t)−βmax)energy to the power-grid and
earn revenue, as shown in Step 9. On the contrary, if a BS
bgets energy deficient at hour t, then the operator will have
to buy ∆b(t)=(βcr −βb(t)) energy from the power-grid to
prevent energy outage, as shown in Step 12 of Algorithm 1.
0
βmax
βcr
βcr
0
βmax
Power grid
BSABSB
Battery A Battery B
PV Panels PV Panels
(a)
Figure 3: Energy flow representation in a two BS network.
B. Energy cooperation model
The proposed EC model for network operation also involves
the BSs maintaining the QoS of maximum associated users
without any flexibility in coverage adjustment. But unlike the
WEC model, it provides flexibility to individual BSs such
that they can share the deficient energy or some portion of it,
from other networked BSs via power-grid itself in the event
of an energy outage scenario. It is proposed that this energy
can be purchased at a much lower price than the grid price
thus reducing the OPEX incurred by the operator. The system
energy flow for the proposed framework has been represented
in Fig. 3 and is presented in Part II of Algorithm 1.
The proposed EC model differs from the WEC model as
to how the energy trade between the BSs and the grid gets
triggered. The WEC model involves buying and selling energy
in extreme cases, i.e., to prevent an energy outage or selling
energy that can’t be stored by the battery storage. On the
contrary, the EC model gets triggered when a BS becomes
energy deficient. It is notable that both the models, WEC and
EC, have similar network QoS performance which is measured
in terms of the number of users served by either of the models.
It has been computed considering that the BSs do not have
flexibility of coverage adjustment [15]. In the current two-BS
network, the following scenarios may arise at any hour t,
1) Both the BSs are not in energy outage, i.e., βb(t)>
βcr ∀b∈ B.
In this case, since both the BSs are not being in outage,
they can sell the surplus energy ε(t) = P2
b=1(βb(t)−
βmax)at hour tto the power-grid and earn revenue.
There is no grid consumption in this scenario.
2) Both the BSs are energy deficient.
In this scenario, both the BSs being energy deficient
rely on power-grid to avoid energy outage and maintain
the user QoS. So, the operator is mandated to purchase
∆(t) = P2
b=1(βcr −βb(t)) energy at hour tto prevent
an energy outage.
3) Either of the BS is energy deficient, while the other BS
is not in outage, i.e., βb(t)< βcr &βb′(t)≥βcr.
This scenario is visualized from the perspective of the
energy deficient BS b, as it triggers request for energy
sharing with the other networked BS b′which is not in
Algorithm 1: Network operation strategies
Result: ∆b, εb, Es
b′b
1Input: Hb(t), βcr, βmax , Eb(t)
2Initialize: βini
b, εb(t)=0,∆b(t) = 0, βb(t)=0
3Part - I: WEC strategy
4Ωbud
b=βini
b−βcr +P24
t=1 Hb(t)
5for t={1,2,· · · ,24}do
6for {b∈ B} do
7βb(t) = βb(t−1) + Hb(t)−Eb(t)
8if (βb(t)≥βmax)then
9εb(t) = εb(t) + βb(t)−βmax
10 βb(t) = βmax
11 else if (βb(t)≤βcr )then
12 ∆b(t) = ∆b(t) + βcr −βb(t)
13 βb(t) = βcr
14 else
15 βb(t) = βb(t)
16 end
17 end
18 Part - II: EC strategy
19 for {b∈ B, b′̸=b}do
20 Sb′b(t) = βb′(t)−βcr +εb′(t)
21 if ∆b(t)̸= 0 then
22 if (Sb′b(t)̸= 0 & Sb′b(t)≥∆b(t)) then
23 Es
b′b(t) = ∆b(t)
Sb′b(t) = Sb′b(t)−∆b(t)
24 ∆b(t)=0
25 else if (Sb′b(t)̸= 0 & Sb′b(t)<∆b(t))
then
26 Es
b′b(t) = Sb′b(t)
∆b(t) = ∆b(t)−Sb′b(t)
27 Sb′b(t)=0
28 buy remaining ∆b(t)from the grid.
29 else
30 (i.e., Sb′b(t) = 0 = Es
b′b(t))
31 buy ∆b(t)from the grid
32 end
33 else
34 i.e., ∆b(t) = 0.
35 No change in parameters required.
36 end
37 end
38 Buy, ∆(t) = Pb∆b(t)
39 Sell, ε(t) = Pbεb(t)
40 Share, ES(t) = Pb′Es
b′b(t)
41 end
energy-outage. If BS bis energy deficient at the tth hour,
i.e., βb(t)< βcr then, BS brequires ∆b(t) = (βcr −
βb(t)) amount of energy to avoid an energy outage. Let
∆b(t)be the amount of grid energy required by BS b
at hour t,Sb′b(t)be the amount of energy that can be
shared by BS b′to BS b, and Es
b′bbe the amount of
energy shared by BS b′to BS bat hour t.Sb′b(t)also
includes the surplus energy εb′(t)and is obtained from
Table II: Network performance trends and analysis
Traffic skewness levels →0%20%40%60%80%
Percent decrease in grid
consumption with EC over
WEC
−42.07% −42.46% −47.45% −47.53% −49.60%
Percent decrease in energy
sold back to grid with EC
over WEC
−12.41% −8.9% −6.12% −4.70% −3.76%
Decrease in revenue earned
by serving users 0% −0.92% −2.80% −3.84% −11.01%
Percent increase in Net
profit with EC over WEC 38.15% 40.60% 46.86% 44.78% 37.20%
Step 20 of Algorithm 1. The energy deficient BS bcan
get the deficient energy as explained below
a) The BS b′shares the entire deficient energy ∆b(t).
In this case since Sb′b(t)≥∆b(t), there is no
requirement of grid energy purchase at BS b. The
steps have been detailed in Steps 22 - 24 of
Algorithm 1.
b) In case, Sb′b(t)<∆b(t), then the energy deficient
BS bstill requires ∆b(t)−Sb′b(t)energy from
the grid to avoid an outage. This case has been
detailed in Steps 25 - 28 of Algorithm 1.
C. Revenue analysis
The following cost metrics are significant from the opera-
tor’s perspective in this work.
1) CAPEX: It refers to the cost incurred upon the opera-
tor during installation of BSs and solar dimensioning.
Taking cost of PV panels to be CS=USD 1300$
and storage batteries CB=USD 216$, CAPEX =
B × (CSNS+CBκ).
2) OPEX: It refers to the cost incurred upon the oper-
ator during BS operations. For a day it is computed
as OPEX =P24
t=1 P2
b=1 cbuy∆b(t),with cbuy =
USD 0.079$ being the cost to buy unit energy from
the power-grid.
3) Energy sharing cost: (Cshare)It can be computed
as Cshare =P24
t=1 P2
b′=1 Es
b′b(t)ζ, where ζ=
USD 0.015$ is the grid maintenance cost borne by the
energy-deficient BS.
4) Revenue by selling energy: (Rsell)It can be com-
puted as Rsell =P24
t=1 P2
b=1 csell εb(t), with csell =
USD 0.057$ being the cost of selling unit energy back
to grid.
5) Revenue by serving users: (Rserv)In this paper,
we calculate Rserv with the BSs having prede-
fined cellular coverage. It is computed as Rserv =
P24
t=1 P2
b=1 cserv Ub(t), with cserv =USD 1.31$ being
the revenue earned by serving a user for 1 month.
Operator’s net profit for the WEC model is calculated as
Profit =Rserv +Rsell −CAPEX −OPEX. For the EC model,
the operator’s profit is calculated as Profit =Rserv +Rsell −
Cshare −CAPEX −OPEX.
V. RE SU LTS A ND DISCUSSION
For simulation and result generation, the area under ob-
servation A= 1km2covered by two single operator BSs is
assumed to have a user density λ= 1500. The performance
of the framework has been analyzed using solar data of New
Delhi city. The dual powered BSs are enabled with 6 unit rated
solar panels and 6 storage batteries in total. The corresponding
CAPEX for the assumed solar dimensioning is computed to
be 9096 USD. The system bandwidth BW is assumed to
be 20 MHz. The values of parameters used in our analysis
are as follows: AWGN noise variance σ2=−150 dBm/Hz,
Pmax = 40W, r0= 800 ×103,p0= 0.9, and δ= 0.3.
Our results show the variation in energy and cost metrics at
varying traffic skewness. The annual grid consumption with
the WEC and the proposed EC model has been depicted
in Fig. 4(a). It is observed that the WEC model results in
a significantly higher grid consumption than the EC model
at all skewness levels. The percentage of grid consumption
that is met by energy sharing through the proposed EC
framework is given in Fig. 4(b). This percent decrease in
grid consumption with the proposed EC model over the WEC
model is showcased in the first row of Table II. We observe
that the grid consumption reduces from about 42% at 0%
skewness to about 50% at 80% skewness, thus inferring that
the grid consumption decreases with an increase in traffic
skewness.
Monthly variation of grid consumption with the WEC and
EC model is shown in Fig. 4(c). We can observe from Fig. 4(c)
that the grid consumption with both the models is lowest in
the summer months of May and June. The grid consumption
then peaks in the monsoon season till August, then again
decreases slightly in the winter months before peaking in the
foggy climate of December - January.
The annual energy sold back to the grid with the WEC and
EC models is given in Fig. 4(d). It can be inferred that the
net energy being sold back to the grid reduces significantly in
the EC model. This trend is shown in second row of Table II,
where we can observe that the negative gain in energy selling
with the EC model reduces from about 12% at 0% skewness
to about 3.7% at 80% skewness. This is intuitively justified
as the EC model involves sharing of the salable energy also
to the energy deficient BS. So, increasing skewness results in
an increase in the energy being sold back to the grid, leading
to decreasing negative gains for the EC model.
Revenue earned by serving users at various skewness levels
is plotted in Fig. 4(e). It is to be noted that both WEC and
EC model, maintain similar QoS guarantees in the network.
It is observed that the decrease in revenue earned by serving
users reduces by up to 11% at 80% traffic skewness, as
given in the third row of Table II. Thus, inferring that the
revenue earned by serving users reduces with increasing traffic
skewness levels.
Finally, the net profit earned by the operator with the WEC
and EC models has been illustrated in Fig. 4(f). We observe
that the profit obtained through the EC model is significantly
higher than the profit obtained through the WEC model. This
trend is shown in the last row of Table II, with the EC
model obtaining a gain up to 47% at 40% traffic skewness
over the WEC model. The observation that the profit gain at
higher skewness levels incur slight reduction can be attributed
0% 20% 40% 60% 80%
Traffic skewness levels
4
6
8
10
12
14
Grid consumption (W)
105
EC
WEC
(a)
0% 20% 40% 60% 80%
Traffic skewness levels
42
43
44
45
46
47
48
49
50
Energy shared (%)
(b)
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
5
6
7
8
9
10
11
12
13
14
Monthly variation of
grid consumption (W)
104
EC
WEC
(c)
0% 20% 40% 60% 80%
Traffic skewness levels
1
1.2
1.4
1.6
1.8
2
Energy sold back to grid (W)
106
EC
WEC
(d)
0% 20% 40% 60% 80%
Traffic skewness levels
6
7
8
9
10
11
12
Revenue earned
by serving users ($)
104
(e)
0% 20% 40% 60% 80%
Traffic skewness levels
4
6
8
10
12
14
Net profit ($)
104
EC
WEC
(f)
Figure 4: Variation of (a) annual grid energy consumption, (b) percentage of
grid consumption met by energy sharing among the networked BSs through
EC, (c) monthly grid consumption with WEC and EC at 0% skewness, (d)
annual energy which is sold back to the power-grid, (e) revenue earned by
serving users at various traffic skewness for both WEC and EC, (f) net profit
earned by the network operator.
to the decrease in revenue earned by selling energy as well
as revenue earned by serving users. Thus, we infer that the
EC model results in providing a significant reduction in grid
consumption with consistently higher profit margins over the
WEC model with increasing skewness.
VI. CONCLUSION
The paper has presented an energy sharing based cooper-
ative framework for reducing the reliance of dual-powered
networks on grid energy consumption and to improve the uti-
lization of network energy. The framework has been designed
to deal with the traffic-energy imbalances occurring in a dual-
powered network, which may result in some BSs becoming
energy deficient. The developed EC framework proposes in-
teraction of the energy deficient BS with other networked
BSs via the grid itself. The proposed energy cooperation
occurs such that a BS not in energy outage, can share the
deficient energy or a portion of deficient energy at a relatively
lesser price. The cooperative energy sharing method has been
compared with the WEC model in terms of annual network
energy consumption and profitability towards the operator.
The proposed energy cooperation strategy is expected to pave
way towards making the network less reliant on power-grid
and achieving self sustainable wireless networks.
ACKNOWLEDGMENT
This work was supported in parts by the Science and
Engineering Research Board, Department of Science and
Technology, Government of India, under the Grant No.
CRG/2019/002293, the Ministry of Science and Technology
under the Grants MOST 110-2634-F-009-021 through Per-
vasive Artificial Intelligence Research (PAIR) Labs, Taiwan,
and the Higher Education Sprout Project of the National Yang
Ming Chiao Tung University and Ministry of Education.
REFERENCES
[1] M. A. Marsan, G. Bucalo, A. Di Caro, M. Meo, and Y. Zhang, “Towards
zero grid electricity networking: Powering BSs with renewable energy
sources,” in Proc. IEEE ICC Wksp., Budapest, Hungary, June 2013.
[2] V. Chamola, B. Krishnamachari, and B. Sikdar, “Green energy and delay
aware downlink power control and user association for off-grid solar-
powered base stations,” IEEE Syst. J., vol. 12, no. 3, pp. 2622–2633,
2018.
[3] S. Suman and S. De, “Low complexity dimensioning of sustainable
solar-enabled systems: A case of base station,” IEEE Trans. Sustain.
Comput., pp. 1–1, 2019.
[4] M. Meo, Y. Zhang, R. Gerboni, and M. A. Marsan, “Dimensioning the
power supply of a lte macro bs connected to a pv panel and the power
grid,” in Proc. IEEE ICC, 2015, pp. 178–184.
[5] A. Balakrishnan, S. De, and L.-C. Wang, “Traffic skewness-aware
performance analysis of dual-powered green cellular networks,” in Proc.
IEEE GLOBECOM, Taipei, Taiwan, Dec. 2020, pp. 1–6.
[6] M. J. Farooq, H. Ghazzai, A. Kadri, H. ElSawy, and M.-S. Alouini, “A
hybrid energy sharing framework for green cellular networks,” IEEE
Trans. Commun., vol. 65, no. 2, pp. 918–934, 2017.
[7] B. K. J. Al-Shammari, N. Al-Aboody, and H. S. Al-Raweshidy, “IoT
traffic management and integration in the QoS supported network,”
IEEE Internet Things J., vol. 5, no. 1, pp. 352–370, 2018.
[8] J. G. Andrews, S. Buzzi, W. Choi, S. V. Hanly, A. Lozano, A. C. K.
Soong, and J. C. Zhang, “What Will 5G Be?” IEEE J. Sel. Areas
Commun., vol. 32, no. 6, pp. 1065–1082, 2014.
[9] E. Oh, B. Krishnamachari, X. Liu, and Z. Niu, “Toward dynamic energy-
efficient operation of cellular network infrastructure,” IEEE Commun.
Mag., vol. 49, no. 6, pp. 56–61, 2011.
[10] H. Jo, Y. J. Sang, P. Xia, and J. G. Andrews, “Heterogeneous cellular
networks with flexible cell association: A comprehensive downlink
SINR analysis,” IEEE Trans. Wireless Commun., vol. 11, no. 10, pp.
3484–3495, 2012.
[11] H. Wu, C. Qiao, S. De, and O. Tonguz, “Integrated cellular and ad hoc
relay systems: iCAR,” IEEE J. Sel. Areas in Commun., vol. 19, no. 10,
pp. 2105–2115, Oct. 2001.
[12] D. Renga, H. Al Haj Hassan, M. Meo, and L. Nuaymi, “Energy
management and base station on/off switching in green mobile networks
for offering ancillary services,” IEEE Trans. Green Commun. Netw.,
vol. 2, no. 3, pp. 868–880, 2018.
[13] Z. Niu, Y. Wu, J. Gong, and Z. Yang, “Cell zooming for cost-efficient
green cellular networks,” IEEE Commun. Mag., vol. 48, no. 11, pp.
74–79, 2010.
[14] Y. Chia, S. Sun, and R. Zhang, “Energy cooperation in cellular networks
with renewable powered base stations,” IEEE Trans. Wireless Commun.,
vol. 13, no. 12, pp. 6996–7010, 2014.
[15] A. Balakrishnan, S. De, and L.-C. Wang, “Network operator revenue
maximization in dual powered green cellular networks,” IEEE Trans.
Green Commun. Netw., pp. 1–1, 2021.
[16] System Advisor Model: National Renewable Energy Laboratory.
[Online]. Available: https://www.sam.nrel.gov
[17] G. Auer, V. Giannini, C. Desset, I. Godor, P. Skillermark, M. Olsson,
M. A. Imran, D. Sabella, M. J. Gonzalez, O. Blume, and A. Fehske,
“How much energy is needed to run a wireless network?” IEEE Wireless
Commun., vol. 18, no. 5, pp. 40–49, 2011.