Provisioning Green Energy for Base Stations in Heterogeneous Networks

Tao Han, Nirwan AnsariIEEE Trans. Vehicular Technology 2014 · DOI

TL;DR

This paper tackles the challenge of minimizing the capital expenditure (CAPEX) for deploying green energy systems in cellular base stations (BSs) within heterogeneous networks (HetNets). They propose a novel two-stage solution: a provision cost-aware traffic load balancing algorithm to optimize energy consumption and a binary energy system sizing algorithm to determine optimal solar panel and battery capacities, ultimately reducing overall network costs while maintaining Quality of Service (QoS).

What Problem Does This Paper Address?

Cellular networks are significant energy consumers, and with the explosion of data traffic, this consumption is rapidly increasing. While green energy technologies offer a solution to reduce reliance on the grid and lower carbon footprints, equipping base stations (BSs) with these systems incurs substantial upfront capital expenditure (CAPEX). This CAPEX is influenced by the size of the green energy generator, battery capacity, and installation costs. The core problem is to find a way to minimize this CAPEX for deploying green energy systems in BSs, particularly in heterogeneous networks (HetNets) which combine macro base stations (MBSs) and small cell base stations (SBSs), while still meeting the required Quality of Service (QoS) for users. A key challenge is the complex interplay between optimizing traffic load across multiple BSs and time slots, and then sizing the green energy systems accordingly.

Methods

The authors decompose the complex Green Energy Provisioning (GEP) problem into two main sub-problems:

  1. Weighted Energy Minimization (WEM) Problem: This sub-problem focuses on minimizing the weighted energy cost of MBSs within individual time slots. They propose a provision cost-aware traffic load balancing algorithm to achieve this. This algorithm uses a Lagrangian dual decomposition approach, involving a traffic redirect algorithm (to determine optimal user-BS association based on SINR and costs) and a traffic load update algorithm (to adjust Lagrangian multipliers). The goal is to intelligently offload traffic to SBSs where appropriate, thereby reducing the energy consumption of MBSs.
  2. Green Energy System Sizing (GESS) Problem: Once the energy consumption of MBSs is determined by the WEM problem, the GESS problem aims to find the optimal solar panel size and battery capacity for each MBS. They introduce a binary energy system sizing (BESS) algorithm. This algorithm uses a binary search to efficiently find the optimal solar panel size, and then derives the corresponding battery capacity to meet the MBS's energy demands over multiple time slots, ensuring that the green energy generated and stored is sufficient.

Datasets

  • UCSD Solar Resource Web Application: Used to obtain solar power generation data.
  • Mobile Traffic Pattern Data: Used to generate mobile traffic rates.

Key Insights and Ideas

  • Decomposition is Key: The GEP problem is too complex to solve directly. Decomposing it into WEM and GESS sub-problems significantly simplifies the optimization process.
  • Traffic Load Balancing is Crucial for Cost Reduction: By intelligently balancing traffic load, especially offloading to SBSs, the energy consumption of MBSs can be reduced, directly impacting the CAPEX of their green energy systems.
  • QoS as a Constraint, Not Just an Objective: The proposed solution actively considers QoS constraints (latency) when balancing traffic, ensuring that cost reduction doesn't come at the expense of user experience.
  • Binary Search for Efficient Sizing: The BESS algorithm's use of binary search provides an efficient way to find optimal solar panel and battery sizes, balancing cost and energy sufficiency.
  • Green Power Percentage Trade-off: Increasing the percentage of power drawn from green energy (parameter $\alpha$) leads to higher provision costs, as it necessitates larger and more expensive green energy systems.

Key Results

  • The proposed solution, combining the Provision Cost-Aware (PCA) traffic load balancing and the Binary Energy System Sizing (BESS) algorithm, achieves the smallest total provision cost compared to other schemes (e.g., DRB-BESS, LM-BESS).
  • The provision cost of the network increases as the green energy percentage ($\alpha$) increases, indicating a direct correlation between the reliance on green energy and the system's upfront cost.
  • The PCA traffic load balancing scheme effectively minimizes the weighted power cost of the network by offloading traffic to SBSs while adhering to QoS constraints.
  • The maximum traffic latency ratio is shown to be a convex function of the data rate bias, allowing for optimization.
  • The convergence of the provision cost aware load balancing algorithm is demonstrated, with the dual function converging after approximately fifty iterations.

Limitations and Future Work

  • Predetermined Parameters: The paper assumes that parameters like the green power percentage ($\alpha$) and the QoS threshold ($\zeta$) are predetermined by the mobile network operator. Future work could explore dynamic adaptation of these parameters.
  • Simplified Models: The paper uses simplified models for traffic arrival (Poisson process) and channel gain. More complex and realistic models could be investigated.
  • Focus on MBSs: While HetNets are considered, the primary focus for green energy provisioning is on MBSs, with SBSs assumed to be grid-powered. Extending the green energy provisioning to SBSs could be a future direction.
  • Various Traffic Load Balancing Algorithms: The paper notes that various traffic load balancing algorithms could be adopted, suggesting that exploring other algorithms in conjunction with their green energy provisioning framework could yield further insights.

Summary updated

BibTeX
@article{han2014provisioning,
  author = {Tao Han and Nirwan Ansari},
  title = {{Provisioning Green Energy for Base Stations in Heterogeneous Networks}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2014},
  doi = {10.1109/tvt.2015.2466101},
}
Figure