When Green Energy Meets Cloud Radio Access Network: Joint Optimization Towards Brown Energy Minimization



Wireless networks have experienced fast development in the past decades. Various advancing wireless technologies have been proposed. To catch up with the ever-increasing diverse communication needs, cloud-radio access networks (C-RAN), which decouples the baseband processing unit (BBU) from the remote radio head (RRH), has been proposed. On the other hand, it has been widely recognized that huge energy consumption has been raised due to the massive deployment of cellular networks. Lowering the network energy consumption therefore becomes a widely concerned topic. To combat the limitations in traditional power grid, smart grid, with the emphasis on distributed energy resource (DER) and bidirectional energy sharing, is advocated to power the wireless networks. In this paper, we are motivated to investigate a joint RRH-BBU association and energy sharing problem towards brown energy usage minimization in green energy powered C-RAN. The problem is formulated into a mixed integer linear programming (MILP) form. To deal with the computation complexity of solving MILP, a two-phase heuristic polynomial-time algorithm is proposed and evaluated via extensive simulation based studies.


Renewable energy Cloud radio access networks Energy efficiency BBU allocation 

1 Introduction

The past decades have already witnessed a massive pen- etration of wireless devices. It is estimated that there will be 50 billion Internet-of-Things (IoT) devices by 2020 and 7.7 billion mobile subscriptions by 2021 [1]. To catch up with the increasing traffic demands, lots of base stations have been deployed. For example, the number of base stations deployed by China Mobile grew from 200,000 to 500,000 from 2004 to 2009, resulting in a two-fold increase of power consumption, 70% of which came from base stations [2]. Such trend imposes unprecedented challenges not only on the performance but also on the energy efficiency and flexibility to wireless networks. As a result, the radio access networks have evolved towards the 5th Generation (5G) [3], for which many advanced technologies have been proposed, e.g., full duplex radio, small cell, device-to-device (D2D) communications, Cloud Radio Access Network (C-RAN). C-RAN decouples the baseband processing units (BBUs) from the remote radio head (RRH) to form a shared BBU pool, thus allowing scalable and dense deployment of light-weight RRHs. High flexility and efficiency are resulted from such design. C-RAN naturally fits the requirements of the next-generation wireless network development and has attracted much attention from both the academia and industry [4, 5, 6, 7].

Meanwhile, on the energy efficiency, one developing trend is on the revolution of power supply paradigm, i.e., from traditional power grid to smart grid. Smart grid technology, with the possibility of elastic electricity distribution and bidirectional energy sharing, especially with the introduction of the green energy from distributed energy resources (DER), has shown great potential in powering wireless networks to achieve high energy efficiency [8, 9]. Thanks to the introduction of DERs, the carbon emission due to brown energy usage can be significantly reduced. Furthermore, by exploring the energy sharing capability of smart grid, renewable green energy generated at a DER not only benefits its attached local network entity but also the other entities via energy sharing, implying further energy efficiency potential. There are many studies, e.g, [10, 11, 35, 36] aiming at the provision of wireless network service with green energy so as to improve energy efficiency. Although existing studies have proved the potential of applying green energy to wireless networks, none of them considers how to schedule the green energy in C-RAN.

Therefore, we are motivated to investigate the green energy scheduling together with the network resource allocation in C-RAN in this paper. We consider a green energy powered C-RAN where both RRHs and BBUs in C-RAN are powered simultaneously by both brown energy and green energy, as shown in Fig. 1. The green energy can be transferred from one site to another in smart grid. As a result, it is significant to study how to schedule the energy sharing together with the network resource allocation to minimize the brown energy usage. The main contributions of our work exist in the following 3-fold:
  1. 1.

    To our best knowledge, we are the first to investigate the green energy scheduling in C-RAN. Specially, we study on the joint optimization of RRH-BBU association and green energy sharing.

  2. 2.

    We formulate the joint optimization problem aiming at brown energy minimizations as an MILP problem.

  3. 3.

    We propose an efficient two-phase heuristic algorithm, whose high efficiency is extensively validated via simulations.

Fig. 1

Green energy powered C-RAN

The rest of the paper is organized as follows. Section 2 summarize the state-of-art related studies. The system model and the problem formulation are presented in Sections 4 and 5, respectively. We then propose our heuristic algorithm in Section 6 and evaluate its efficiency in Section 7, respectively. The paper is concluded in Section 8.

2 Related work

2.1 Wireless network virtualization

Regarding wireless network virtualization, survey paper [12] overviews its enabling technologies and outlines some recent some recent efforts. Since wireless network virtualization enables abstraction and sharing of infrastructure and radio spectrum resources, resource sharing therefore becomes a hot topic in this area, e.g., [13, 14, 15, 16, 17]. In order to promote the flexibility of the next generation wireless networks, one developing trend is on the wireless network virtualization, as surveyed in [12]. Many recent studies have identified the potential of NFV, SDN and C-RAN in wireless network virtualization and proposed many novel architectures by exploring these newly emerging technologies [18, 19, 20, 21, 22, 23]. Among these wireless network virtualization technologies, C-RAN attracts the most interests recently. For example, Sun et al. [21] integrate NFV, software-defined radio (SDR) and SDN for 4G/5G networks, and identify C-RAN as a core component in future 5G networks. As a result, much recent effort has been devoted to this area [4, 24, 25, 26]. By centralizing the BBU and allowing dense RRH deployment, C-RAN exhibits high performance efficiency. He et al. [4] derive the upper and lower bounds of the downlink ergodic capacity that can be achieved in C-RAN. On the other hand, pioneering researchers have also studied how to explore the flexility introduced by the separation of BBUs and RRHs. Liu et al. [24] discuss a re-configurable backhaul in C-RAN that allows a one-to-many mapping between RRH and BBU, thereby resulting in dynamic activation of BBUs and energy saving. Such concept is also practically proved by a prototype designed by Sundaresan et al. in [19]. Some recent efforts have mainly devoted to infrastructure sharing and spectrum sharing, as surveyed in [27].

Due to the promising advantage on the network flexility and performance efficiency, some recent studies are also contributed to study the energy efficiency in C-RAN. Sabella et al. [5] theoretically discuss the energy efficiency advantage of C-RAN by introducing a generalized holistic power model with the consideration of power consumption at the RRHs, BBUs and the backhaul connections. Alhumaima et al. [7] captures the power consumption of individual components and investigating the effect of different parameters, such as the number of antennas and the system bandwidth, to the energy efficiency of C-RAN. Suryaprakash et al. [6] show that C-RAN requires approximately 10% to 15% less capital expenditure per square kilometer than traditional LTE networks. To optimize the C-RAN energy efficiency, Tang et al. [28] investigate how to adjust the BBU process rate according to the traffic demand. Kong et al. [25] construct an energy-efficiency BBU cluster reference testbed referred as eBase for C-RAN concept verification and performance evaluation. Pompili et al. [26] propose a demand-aware virtual base stations resizing scheme to adapt to the demand fluctuation of the cellular networks.

2.2 Energy efficiency in wireless networks

On the other hand, observing the ever-increasing wireless network energy consumption, many recent studies have been devoted to lowering the energy consumption in wireless networks, especially with the introduction of green energy [8, 9, 10, 11]. For example, Bu et al. [8] study the dynamic activation of cellular base stations according to the traffic, real-time electricity price, and the pollutant level associated with electricity generation and formulate the problem as a Stackelberg game. Similar problem is also investigated by Niyato et al. [9], who adapt the power management of green energy powered base stations to various uncertainties like renewable power generation, power price, and wireless traffic load. Li et al. [11] propose a noncooperative game model for optimal energy allocation with green energy for wireless network. Piro et al. [10] introduce a HetNets system model powered by green energy sources for the next generation cellular networks.

By our literature survey, we notice that green energy aware network resource management is essential to promote the network energy efficiency. However, how to schedule the green energy in C-RAN is still under-investigated. To this end, we are motivated to study the green energy sharing in conjunction with RRH-BBU association in this paper.

3 Green energy powered cloud radio access networks

In this section, we present an overview of green energy powered C-RAN architecture, as shown Fig. 1, which mainly consists of two parts, i.e., network and energy provision.

The network part inherits from C-RAN, where the BBUs are separated from RRHs and centralized into the cloud. The hardware-only RRHs thus respond for simple signal processing while the entire baseband processing operations, or possibly as well as the MAC layer functions [20], are migrated to the BBUs. The lightweight RRHs then can be scattered in the network and deployed densely to shorten the distance between the user equipments, potentially promoting the quality-of-service (QoS). The BBUs locate on the baseband servers, which could reside in a centralized cloud or distributed in the edge cloud. The latter paradigm was proposed recently and referred as Fog Radio Access Networks (FRAN) [29]. Nevertheless, both CRAN and FRAN share the same philosophy, i.e., decoupling and centralizing the BBUs into a shared pool.

On the energy provision part, smart grid provides efficient control of the delivery and use of electricity to revolutionize the power provision paradigm towards high energy efficiency, In particular, distributed energy resources (DER) that harvest energy from green sources, e.g., solar energy and wind energy, can significantly alleviate the energy poverty and substantially reduce carbon footprints [8, 9, 11, 30, 31, 32]. Therefore, we advocate that all the network entities are equipped with local DERs. By such means, we can see from Fig. 1 that all the entities are simultaneously powered by both brown energy from traditional power grid and green energy from DERs. We believe that such design can substantially reduce carbon footprints resulted from brown energy, and therefore will significantly alleviate the energy poverty and sustain the wireless network better.

4 System model

In this section, we introduce the system model, including both network model and energy model. For the conveniences of the readers, the major notations used in this paper are listed in Table 1.
Table 1



The BBU set




The RRH set



λ r

The communication request arrival rate at RRH rR

p b

Total power consumption at BBU bB

e b

Static power consumption at BBU bB

y b

Whether BBU bB is activated or not

g b

Power generation rate at BBU bB


Power consumption and request processing ratio at BBU

μ b

Total requests at BBU bB

p r

Total power consumption at RRH rR

e r

Static power consumption at RRH rR


Power consumption and request processing ratio at RRH

g r

Power generation rate at RRH rR

η x y

Power transferring ratio between DERs at network entities x and y, x,yRB

x r b

Whether RRH r is associated with BBU bB or not

\(\mu _{\max }\)

The maximum processing capacity at BBU bB

p e y

The brown energy consumed by network entity yRB

4.1 Network model

We consider a general green energy powered C-RAN network as shown in Fig. 1. The main network entities in a C-RAN network are the BBUs and RRHs, which are denoted as set B and R, respectively. The BBUs and RRHs are scattered in the network area. Without loss of generality, we assume that the BBUs and RRHs are all inter-connected. That is, a RRH can be served by any BBU in the network. As the RRHs and BBUs do not need to co-located, their association relationship can be freely defined according to various objectives. The RRHs directly serve the users within its transmission area. Let λ r denote communication request arrival rate on RRH rR. Each BBU can process a limited amount of workload constrained by its processing capacity. We assume that all BBUs are homogeneous with the same processing capacity μmax.

4.2 Energy model

We adopt an energy consumption consumption model given in [33] where each transmission consumes a unit of energy from the base station. Accordingly, we calculate the total energy consumed at any active BBU bB as
$$ p_{b}=e_{b} + \alpha \mu_{b}, $$
where e b is the static energy consumption of BBU b, regardless of its actual usage once activated and αμ b is the dynamic power consumption proportional to the workload μ b allocated to the BBU with ratio α. Similarly, the power consumed by RRH rR can be calculated as
$$ p_{r} = e_{r} + \beta \lambda_{r}, $$
where e r is the static energy consumption of RRH r, and βγ r is the dynamic part proportional to the workload λ r with ratio β.

All the network entities are simultaneously powered by both brown energy from traditional power grid and green energy from DERs. The green energy generated at a DER is typically random with certain rate. Without loss of generality, we assume that the energy generated from DER at network entity bB and rR are with rate g b and g r , respectively. The green energy can be shared among the network entities. As indicated in [34], it is impossible to transfer power between DERs without any loss. The loss may account for 7% of the transferred energy or may even reach 55% in extreme cases. To represent such loss relationship, we use η x y ,x,yRB to denote the efficiently energy transferring ratio between the DERs on different network entities. Specially, we define η x x ≡ 1, ∀xRB to indicate that no energy loss will be experienced at local DER.

5 Problem formulation

5.1 RRH-BBU association constraints

Thanks to the pooling of BBU resources, the association between BBU and RRH becomes flexible and can be adjusted at runtime. We denote the association relationship between a RRH rR and a BBU bB as a binary variable
$$ x_{rb}=\left\{ \begin{aligned} 1, & \text{ if RRH}\, \text{r is associated with BBU}\, b,\\ 0, & \text{ otherwise.} \end{aligned} \right. $$
Each RRH must be attached to one BBU by creating a back-haul link between them. Therefore, we have
$$ \sum_{b\in B}x_{rb} = 1, \forall r\in R. $$
Note that, from the consideration of energy efficiency, not all BBUs must be activated. To denote such status, we define a binary variable
$$ y_{b}=\left\{ \begin{aligned} 1, & \text{ if BBU}\, b\, \text{is active,}\\ 0, & \text{ otherwise.} \end{aligned} \right. $$
Consequently, the value of x r b shall be constrained by y b as
$$ x_{rb} \leq y_{b}, \forall r\in R, b\in B, $$
indicating that a RRH can only be associated with an active BBU, i.e., y b = 1; otherwise, x r b is forced to be 0 if y b = 0.

5.2 Process capacity constraints

According to the C-RAN communication process, the workload on a BBU is related to its associated RRHs. The total workload \(\mu _{b}=\sum _{r\in R}x_{rb}\lambda _{r}\) allocated to BBU bB shall not exceed its processing capacity, i.e.,
$$ \sum\limits_{r\in R}x_{rb}\lambda_{r} \leq y_{b} \mu_{\max}, \forall b\in B. $$

5.3 Power supply constraints

To describe the energy transferring relationship, we define variables p x y ,x,yRB to represent the energy transferring amount from DER at network entity x to y. Specially, note that we allow x = y, e.g., p x x ,xRB, indicating the amount of green energy from the local DER. When the green energy cannot catch up with the energy demand of a network entity, certain brown energy must be retrieved from the power grid. We use p e y to denote the amount of brown energy retrieved from the power grid to the network entity y, which could be either a RRH or a BBU. To ensure the functionality of a network entity, its total energy supply must satisfy its energy requirement. That is,
$$ \sum\limits_{r\in R}p_{rb}\eta_{rb} + \sum\limits_{x\in B} p_{xb}\eta_{xb} + p_{eb} \!\geq\! y_{b}e_{b} + \alpha \sum\limits_{r\in R}x_{rb}\lambda_{r}, \forall b\in B, $$
$$ \sum\limits_{x\in R}p_{xr}\eta_{xr} + \sum\limits_{b\in B} p_{br}\eta_{br} + p_{er} \geq e_{r} + \beta \lambda_{r}, \forall r\in R, $$
where p e b and p e r denote the brown energy retrieved from the power grid to BBU b and RRH r, respectively.
The green energy transferring amount is limited by the green energy generation rate. Therefore, we have
$$ \sum\limits_{x\in R\cup B} \sum\limits_{y\in R\cup B} p_{xy} \leq \sum\limits_{r\in R} g_{r} + \sum\limits_{b\in B}g_{b}. $$
Our objective to maximize the energy efficiency is equi- valent to minimizing the brown energy usage. Taking all the constraints discussed above, we eventually obtain an MILP as
$$ \begin{aligned} \mathbf{MILP:} \\ \min: &\hspace{0.2em} \sum\limits_{r\in R}p_{er} + \sum\limits_{b \in B} p_{eb},\\ \text{s.t.} : &\hspace{0.2em} (4), (6), (7) - (10),\\ &\hspace{0.2em} x_{rb}\in \{0, 1\}, y_{b} \in \{0, 1\}, \forall r\in R, b\in B. \end{aligned} $$

6 A two-phase heuristic algorithm

It would be computationally prohibitive to solve the MILP problem for large-scale networks. To tackle this issue, we propose a two-phase polynomial-time heuristic algorithm, as summarized in Algorithm 1, in this section.
Phase I:

The first phase of our algorithm is to find out a green energy generation aware RRH-BBU association relationship, intending to minimize the number of activated BBUs and maximize the usage of local green energy. By minimizing the number of activated BBUs, the overall energy consumption could be lowered, potentially lowering the brown energy usage. Maximizing the local green energy usage is to lower the energy loss due to inter-DER energy transferring. Based on such principle, we design the first phase as shown in lines 1-13. In order to maximize the usage of local green energy, we try to assign BBUs from the one with the highest green energy generation rate. Therefore, we first sort the BBUs into set sortedBBUs according to their green energy generation rates g b , ∀bB in a descend order in line 2. Thereafter, we try to sequentially assign the RRHs to BBUs until all RRHs are completely assigned in lines 3-13. For each RRH, we assign it to the first BBU with enough processing capacity to satisfy its demand in the sorted BBU set sortedBBUs. For each assignment, we correspondingly set the values of association relationship x r b (line 6) and activation status y b (lines 7-9), respectively. Then, we update the accumulated workload and the remaining processing capacity on the chosen BBU in line 10.

Phase II:
In phase II, we take the green energy generation aware RRH-BBU association relationship x r b , ∀rR,bB and BBU activation decisions y b ,bB as input to plan energy transferring among the DERs. As the association relationship (i.e., the values of x r b , ∀rR,bB) has already been known, it is possible to calculate the energy surplus on each network entity. The energy surplus a b and b r on BBU bB and RRH rR can be calculated as
$$ a_{b} = g_{b} - \alpha \sum\limits_{r\in R}x_{rb}\lambda_{r}, $$
$$ a_{r} = g_{r} - \beta \lambda_{r}, $$
respectively. Note that the energy surplus could be negative if the green energy generation rate fails to catch up with the energy demand. Actually, the negative energy surplus refers to the additional energy demand, which can be satisfied by either the brown energy or green energy from the other DERs with positive energy surplus.
Then, based on the energy surplus information and the energy transferring loss ratio, we can establish a linear programming (LP) model to maximize the usage of the surplus energy for overall energy efficiency as follows:
$$ \begin{aligned} \mathbf{LP:} \\ \min: &\hspace{0.2em} \sum\limits_{r\in R}p_{er} + \sum\limits_{b \in B} p_{eb},\\ \text{s.t.} : &\hspace{0.2em} \sum\limits_{r\in R}p_{rb}\eta_{rb} + \sum\limits_{x\in B} p_{xb}\eta_{xb} + p_{eb} \geq a_{b}, \forall b\in B, \;\text{(c1)}\\ &\hspace{0.2em} \sum\limits_{x\in R}p_{xr}\eta_{xr} + \sum\limits_{b\in B} p_{br}\eta_{br} + p_{er} \geq a_{r}, \forall r\in R, \;\text{(c2)}\\ &\hspace{0.2em} \sum\limits_{x\in R\cup B}\sum\limits_{y\in R\cup B} p_{xy} \leq \sum\limits_{r\in R} a_{r} + \sum\limits_{b\in B}a_{b},\;\;\;\;\;\;\;\;\;\;\;\;\; \text{(c3)} \end{aligned} $$
where constraints (c1) and (c2) are to ensure all the additional energy demands get satisfied, and constraints (c3) is to ensure the green energy to be transferred does not exceed to the total available one. The objective is set the same as in Section 5. Accordingly, we solve the LP in line 17 to obtain the optimal energy sharing for the association results obtained in Phase I.

Remark 1

Our two-phase algorithm is with polynomial computation complexity. Note that Phase I totally requires at most |B|∗|R| number of iterations while the LP involved in Phase II can be solved in polynomial time.

7 Performance evaluation

In this section, we present our performance evaluation results on our proposed algorithm (“Two-Phase”) by comparing it with the optimal solution (“Optimal”) and other two competitors to verify the correctness of our joint optimization design and the efficiency of our proposal. The first competitor (“RRH-BBU Opti”) minimizes the number of activated BBUs and fully explores the green energy generated at each DER, but without energy sharing among DERs. While, the second one (“Energy Sharing”) emphasizes the energy sharing but does not optimize the RRH-BBU association.

The simulated network environment strictly follows the system model presented in Section 4. In order to extensively evaluate the efficiency of our proposal, the request rates on RRHs, the green energy generation rates on RRHs and BBUs are all randomly generated with different seeds in each simulation. The basic network settings are as follows, |R| = 40, |B| = 15, e b = 1.0, e r = 0.1, \(\bar {\lambda _{r}}= 1.0\), \(\bar {g_{b}}= 2.0\), \(\tilde {g_{r}}= 1.0\), α = 4, β = 2, where |⋅| and \(\bar {\cdot }\) are the cardinality function and average function, respectively. Several groups of experiments have been conducted to investigate the effect of different network parameters on the energy efficiency. Each group of experiment includes 50 simulation instances, whose average brown energy consumption and total energy consumption are calculated.

Let us first check the effect of the static energy con- sumption of BBU to the overall energy efficiency. Figure 2 shows the average brown energy consumption under different values of static energy consumption from 0.5 to 5.0. We respectively plot the brown energy consumption and the total one in Fig . 2a and b, respectively, from which we can see that both are increased with the increasing of base energy consumption. Under all cases, we can see that our algorithm exhibits the highest energy efficiency compared with the other two competitors and performs much close to the optimal one. The advantage is even obvious when the base BBU energy consumptione b is large. When e b is small, minimizing the number of activated BBUs does not take much benefit to the overall energy efficiency and therefore “Energy Sharing” is even better than “RRH-BBU Opti” when e b ≤ 0.5. However, when e b becomes dominant, it is significantly to minimize the number of activated BBUs and to maximize the usage of each activated one and therefore “RRH-BBU Opti” becomes advantageous over “Energy Sharing”. Nevertheless, our “Two-Phase” algorithm, thanks to the joint optimization of both factors, always performs the best under any value of e b .
Fig. 2

The average energy consumption under different values of static BBU energy consumption

We then investigate the number of RRHs to the network energy efficiency by varying the number of RRHs from 5 to 20. The performance evaluation results are shown in Fig. 3, where the brown energy consumption and the total one are plotted in in Fig. 3a and b, respectively. We can see that both the brown energy consumption and the total energy consumption increase with the number of RRHs. This is because more RRHs imply higher communication request demand and hence higher workload and energy consumption on the BBUs. It is interesting to notice that “Energy Sharing” is advantageous over “RRH-BBU Opti” on the brown energy but disadvantage on the overall energy when |R|≥ 10. This is because when the number of RRHs is large, carefully activating BBUs is beneficial to minimizing the energy consumption but it is also significant to share the green energy to reduce the brown energy consumption at the same time. More RRHs imply higher surplus energy at RRHs and therefore less brown energy will be used.
Fig. 3

The average energy consumption under different numbers of RRHs

Next, we investigate how the green energy generation rate affects the energy efficiency. Figure 4 shows the average brown energy consumption under different average green energy generation rates \(\bar {g_{r}}\) ranging from 0.5 to 1.0, \(\bar {g_{b}}\) from 1.0 to 1.8. We can see that the brown energy consumption shows as a decreasing function of either \(\bar {g_{r}}\) or \(\bar {g_{b}}\) under any algorithm. This is because higher green energy generation rate implies less brown energy requirement. An interesting phenomenon is that obvious advantage of “RRH-BBU Opti” over “Energy Sharing” can be observed from Fig. 4a when g r is small while it becomes marginal when g r is big. This is because the BBU energy consumption deeply affects the overall energy efficiency and therefore it is significant to minimize the number of BBUs by “RRH-BBU Opti”. However, when there are more excessive green energy, “Energy Sharing” also becomes significant to benefit the overall energy efficiency. As a result, it performs close to “RRH-BBU Opti” when \(\bar {g_{r}}\geq 0.9\). Under any values of \(\bar {g_{r}}\), our algorithm always performs the best. This further validates the efficiency of our proposed algorithm.
Fig. 4

The brown energy consumption under different average green energy generation rates

Finally, let us look at how the number of BBUs affects the brown energy consumption. We vary the number of BBUs from 15 to 40 and plot the average brown energy consumption in Fig. 5, from which we can see that the brown energy consumption shows as a decreasing function of the number of BBUs, for all algorithms. There are two reasons leading to such phenomenon. At first, more BBUs imply more options to find a BBU subset with higher green energy generation rate, potentially lowering the brown energy consumption. Secondly, more BBUs also imply more excessive green energy generation. For example, the inactivated BBUs can fully contribute their green energy to the network, significantly reducing the brown energy consumption. Both reasons results in the substantial advantage of our joint optimization over either “RRH-BBU Opti” or “Energy Sharing” when there are a large number of BBUs. As observed in Fig. 5, the brown energy consumption gap expands with the increasing of BBU number.
Fig. 5

The brown energy consumption under different number of BBUs

8 Conclusion

In this paper, we investigate a brown energy consumption minimization problem by joint optimization of the RRH-BBU association and green energy sharing. We formally describe the problem by MILP and then propose a two-phase heuristic algorithm to obtain a sub-optimal solution. The efficiency of our proposed algorithm is extensively verified by the fact it performs much close to the optimal solution and exhibits much advantage over the other two competitors. Our study identifies the advantage of joint optimization of network resources and energy resource in green energy powered C-RAN and will open up more research directions to pursue the high performance efficiency and high-energy efficiency next-generation wireless networks.



The work was jointly supported by the Chongqing Municipal Basic and Advanced Research project under GRANT cstc2015jcyjBX0009 and CSTCKJCXLJRC20.


  1. 1.
    Swan M (2012) Sensor Mania! The internet of things, wearable computing, objective metrics, and the quantified self 2.0. J Sensor Actuator Netw 1(3):217–253CrossRefGoogle Scholar
  2. 2.
    Institute CMR (2011) C-RAN: the road towards green RAN. China Mobile Research InstituteGoogle Scholar
  3. 3.
    Vision I (2014) Framework and overall objectives of the future development of imt for 2020 and beyond, Working document toward preliminary draft new recommendation ITU-R M.[IMT Vision]Google Scholar
  4. 4.
    He Y, Dutkiewicz E, Fang G, Shi J (2014) Differential capacity bounds for distributed antenna systems under low SNR conditions. In: 2014 IEEE International conference on communications, ICC 2014, pp 5550–5554Google Scholar
  5. 5.
    Sabella D, De Domenico A, Katranaras E, Imran M A, Di Girolamo M, Salim U, Lalam M, Samdanis K, Maeder A (2015) Energy efficiency benefits of RAN-as-a-service concept for a cloud-based 5G mobile network infrastructure. IEEE Access 2:1586–1597CrossRefGoogle Scholar
  6. 6.
    Suryaprakash V, Rost P, Fettweis G (2015) Are heterogeneous cloud-based radio access networks cost effective? IEEE J Selected Areas Commun 33(10):2239–2251CrossRefGoogle Scholar
  7. 7.
    Alhumaima RS, Al-Raweshidy HS (2016) Evaluating the energy efficiency of software defined-based cloud radio access networks. IET Commun 10(8):987–994CrossRefGoogle Scholar
  8. 8.
    Bu S, Yu FR, Cai Y, Liu XP (2012) When the smart grid meets energy-efficient communications: green wireless cellular networks powered by the smart grid. IEEE Trans Wirel Commun, 1–11Google Scholar
  9. 9.
    Niyato D, Lu X, Wang P (2012) Adaptive power management for wireless base stations in a smart grid environment. IEEE Wireless Commun 19(6):44–51CrossRefGoogle Scholar
  10. 10.
    Piro G, Miozzo M, Forte G, Baldo N, Grieco LA, Boggia G, Dini P (Jan 2013) Hetnets powered by renewable energy sources: Sustainable next-generation cellular networks. IEEE Internet Comput 17(1):32–39CrossRefGoogle Scholar
  11. 11.
    Li D, Saad W, Guvenc I, Mehbodniya A, Adachi F (2015) Decentralized energy allocation for wireless networks with renewable energy powered base stations. IEEE Trans Commun 63(6):2126–2142CrossRefGoogle Scholar
  12. 12.
    Liang C, Yu FR, Yu R (2015) Wireless network virtualization: a survey, some research issues and challenges. IEEE Commun Surveys Tutor 17(1):358–380CrossRefGoogle Scholar
  13. 13.
    Park SH, Simeone O, Sahin O, Shamai S (2012) Joint base station selection and distributed compression for cloud radio access networks. In: 2012 IEEE Globecom workshops, GC Wkshps 2012, pp 1134–1138Google Scholar
  14. 14.
    Kamel MI, Le LB, Girard A (2014) LTE wireless network virtualization: dynamic slicing via flexible scheduling. IEEE Veh Technol Conf, 0–4Google Scholar
  15. 15.
    Wang R, Hu H, Yang X (2014) Potentials and challenges of C-RAN supporting multi-RATs toward 5G mobile networks. IEEE Access 2:1200–1208Google Scholar
  16. 16.
    Marotta MA, Kaminski N, Gomez-Miguelez I, Granville LZ, Rochol J, DaSilva L, Both CB (2015) Resource sharing in heterogeneous cloud radio access networks. IEEE Wirel Commun 22(3):74–82CrossRefGoogle Scholar
  17. 17.
    Cai Y, Yu FR, Bu S (2016) Dynamic operations of cloud radio access networks (C-RAN) for mobile cloud computing systems. IEEE Trans Veh Technol 65(3):1536–1548CrossRefGoogle Scholar
  18. 18.
    Costa-Perez X, Swetina J, Guo Tao, Mahindra R, Rangarajan S (2013) Radio access network virtualization for future mobile carrier networks. IEEE Commun Mag 51(7):27–35CrossRefGoogle Scholar
  19. 19.
    Sundaresan K, Arslan MY, Singh S, Rangarajan S, Krishnamurthy SV (2013) FluidNet: a flexible cloud-based radio access network for small cells. In: Proceedings of the 19th annual international conference on mobile computing & networking - MobiCom ’13. ACM Press, New York, p 99Google Scholar
  20. 20.
    Akyildiz IF, Wang P, Lin SC (2015) SoftAir: a software defined networking architecture for 5G wireless systems. Comput Netw 85:1–18CrossRefGoogle Scholar
  21. 21.
    Sun S, Kadoch M, Gong L, Rong B (2015) Integrating network function virtualization with SDR and SDN for 4G/5G networks. IEEE Netw 29(3):54–59CrossRefGoogle Scholar
  22. 22.
    Yang C, Chen Z, Xia B, Wang J (2015) When ICN meets C-RAN for HetNets: an SDN approach. IEEE Commun Mag 53(11):118–125CrossRefGoogle Scholar
  23. 23.
    Chih-Lin I, Han S, Xu Z, Wang S, Sun Q, Chen Y (2016) New paradigm of 5G wireless internet. IEEE J Selected Areas Commun 34(3):474–482CrossRefGoogle Scholar
  24. 24.
    Liu C, Sundaresan K, Jiang M, Rangarajan S, Chang GK (2013) The case for re-configurable backhaul in cloud-RAN based small cell networks. In: Proceedings - IEEE INFOCOM, pp 1124–1132Google Scholar
  25. 25.
    Kong Z, Gong J, Xu CZ, Wang K, Rao J (2013) EBase: a baseband unit cluster testbed to improve energy-efficiency for cloud radio access network. In: IEEE International conference on communications, pp 4222–4227Google Scholar
  26. 26.
    Pompili D, Hajisami A, Viswanathan H (2015) Dynamic provisioning and allocation in cloud radio access networks (C-RANs). Ad Hoc Netw 30:128–143CrossRefGoogle Scholar
  27. 27.
    Baliosian J, Serrat J, Gorricho J-l (2016) Resource slicing in virtual wireless networks. A Survey 4537(c):1–14Google Scholar
  28. 28.
    Tang J, Tay WP, Quek TQS (2015) Cross-layer resource allocation with elastic service scaling in cloud radio access network. IEEE Trans Wirel Commun 14(9):5068–5081CrossRefGoogle Scholar
  29. 29.
    Tandon R, Simeone O (2016) Cloud-aided wireless networks with edge caching: fundamental latency trade-offs in fog radio access networks. In: 2016 IEEE International symposium on information theory (ISIT). IEEE, pp 2029–2033Google Scholar
  30. 30.
    Huang X, Han T, Ansari N (2015) On green-energy-powered cognitive radio networks. IEEE Commun Surv Tutor 17(2):827–842CrossRefGoogle Scholar
  31. 31.
    Huang X, Ansari N (2015) Energy sharing within EH-enabled wireless communication networks. IEEE Wirel Commun 22(3):144–149CrossRefGoogle Scholar
  32. 32.
    Erol-Kantarci M, Mouftah HT (2015) Energy-efficient information and communication infrastructures in the smart grid: a survey on interactions and open issues. IEEE Commun Surv Tutor 17(1):179–197CrossRefGoogle Scholar
  33. 33.
    Li D, Saad W, Guvenc I, Mehbodniya A, Adachi F (2015) Decentralized energy allocation for wireless networks with renewable energy powered base stations. IEEE Trans Commun 63(6):2126–2142CrossRefGoogle Scholar
  34. 34.
    Gkatzikis L, Iosifidis G, Koutsopoulos I, Tassiulas L (2015) Collaborative placement and sharing of storage resources in the Smart Grid. In: 2014 IEEE international conference on Smart Grid communications, SmartGridComm 2014, pp 103–108Google Scholar
  35. 35.
    Luo C, Guo SY, Guo S, Yang LT, Min G, Xie X (2014) Green communication in energy renewable wireless mesh networks: routing, rate control, and power allocation. IEEE Trans Parallel Distrib Syst 25(12):3211–3220CrossRefGoogle Scholar
  36. 36.
    Wang K, Wang Y, Sun Y, Guo S, Wu J (2016) Green industrial internet of things architecture: an energy-efficient perspective. IEEE Commun Mag 54(12):48–54CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Authors and Affiliations

  1. 1.Department of ComputingThe Hong Kong Polytechnic UniversityKowloonHong Kong
  2. 2.Chongqing University of Posts and TelecommunicationsChongqingChina
  3. 3.School of Computer Science, Hubei Key Laboratory of Intelligent Geo-Information ProcessingChina University of GeosciencesWuhanChina
  4. 4.School of Computer Science and TechnologyHuazhong University of Science and TechnologyWuhanChina

Personalised recommendations