Energy-Based Cell Association

• Energy-based cell association schemes are dynamic frameworks that adjust user-to-BS associations based on real-time renewable energy metrics and grid availability.
• They employ two-phase algorithms integrating BS state configuration and user association, significantly reducing on-grid power consumption and balancing load.
• Analytical models and simulations demonstrate enhanced throughput, energy efficiency, and reduced outage probabilities in sustainable heterogeneous networks.

Energy-based cell association schemes are a foundational element of modern wireless network resource management, particularly in heterogeneous cellular architectures that increasingly rely on renewable energy sources. At their core, these schemes leverage the real-time, local, or statistical availability of energy at base stations (BSs)—whether from the grid, harvested renewables, or hybrid sources—to guide both user-to-cell associations and BS activation. Their overarching purpose is to minimize on-grid power consumption, optimize energy efficiency, and support network sustainability objectives without compromising user coverage and quality of service.

1. System Models and Core Principles

Energy-based cell association operates in diverse heterogeneous cellular network (HCN) topologies with multiple classes of BSs, each with distinct energy supply characteristics:

• Macro-cell BSs (MBSs): Always active, powered entirely by the electric grid.
• Conventional Small-cell BSs (CSBSs): Smaller, grid-powered, can be switched into sleep.
• Renewable Small-cell BSs (RSBSs): Receive power solely from local energy harvesting (solar, wind, etc.), often modeled without long-term storage; harvested energy must be used immediately or is lost.
• Hybrid Small-cell BSs (HSBSs): Can use both harvested energy and grid supply.

Energy harvesting at BSs is modeled as a Poisson arrival process, and the harvested energy, $E_{EH}$, and consumption, $E_{SBS}$, per time slot determine admissible cell activity. A design margin $C$ ensures a minimal buffer: $\Delta = E_{EH} - E_{SBS} \geq C$ (Rehman et al., 2018).

System-level metrics include instantaneous and long-term energy availability, channel conditions (including interference), and real-time load per BS ($U_b$). The key decision variables capture cell on/off states, transmit powers, user associations, and dynamic spectrum allocation (You et al., 2016, Zhuang et al., 2015).

2. Association and BS Activation Algorithms

The fundamental algorithmic structure of energy-based cell association features a two-phase decision process (Rehman et al., 2018):

1. BS State Configuration:
   • Iteratively update each cell's transmit power to drive its energy balance $\Delta$ within $[0,C]$.
   • For RSBSs: expand or shrink cell by adjusting transmit power; switch off cells if renewable energy is insufficient.
   • For CSBSs: switch to sleep when under minimum load threshold.
   • For HSBSs: adjust transmit power based on available green energy; always in active mode.
2. User Association:
   • Users prioritize association to the nearest active EH-SBS (RSBS first, then HSBS). Fallbacks are to CSBS, then MBS.
   • The selection uses beacon scanning and considers only BSs currently, and likely, able to serve within energy budget constraints.

The interplay between checkpointed BS energy states, local harvested power, user distribution, and channel availability is critical. Constraints ensure all users are covered with load limitations per BS and adherence to harvested energy availability (Rehman et al., 2018, Parzysz et al., 2016).

In distributed IoT and ultra-dense networks, user-side association operates with minimal coordination. Devices apply energy-causal policies to allocate harvested power only if association to a given SBS is feasible within harvested energy constraints, leveraging mean-field multi-armed bandit models for convergence (Maghsudi et al., 2016).

3. Analytical Models, Metrics, and Optimization

Energy-based schemes are underpinned by stochastic and Markovian models that characterize BS battery dynamics, user association probabilities, coverage, and performance metrics:

• Battery evolution: Modeled as finite state Markov chains, updated each slot based on harvested energy and user-driven consumption (Parzysz et al., 2016, Parzysz et al., 2016).
• Coverage/outage probability: Formulated via stochastic geometry, considering the density of scBSs able to serve given real-time battery state and user association (Parzysz et al., 2016).
• Throughput and energy efficiency: Aggregated using

$$R = \sum_{u} B \log_2 \left(1 + \frac{P_b h_{u,b}}{\sum_{z\neq b}P_z h_{u,z}} \right), \qquad EE = \frac{R}{P_{\text{grid}}}$$

(Rehman et al., 2018).

Optimization objectives include:

• On-grid power minimization: By dynamically activating/deactivating BSs and adjusting association to favor renewable-powered cells.
• Energy efficiency maximization: Bits delivered per Joule from non-renewable (grid) energy.
• Carbon efficiency: Through metrics such as the ratio of total throughput to net carbon emissions, optimized via genetic algorithms for bias selection in association rules (Zhao et al., 19 Jan 2026).

Constraints typically ensure user coverage, BS load bounds, frugality in battery use (to prolong service and minimize outages), and convergence to globally feasible operating points (You et al., 2016, Kuang et al., 2016).

4. Bias-Based and Distributed Association Mechanisms

A major class of energy-based association strategies employs biasing, either globally or via adaptive, real-time methods:

• Adaptive biasing (availability-aware): Each user computes the feasibility of renewable versus grid-based supply at neighboring BSs, applying a lower bias ($\beta_A$) to those able to serve with harvested energy, and higher bias ($\beta_G$) to grid-only supply (Parzysz et al., 2016).
• State-aware priority: Association biases for hybrid BSs are dynamically set based on instantaneous battery state, user power requirement, and statistical load estimate. This allows for optimal energy resource sharing among tiers and time-varying demand (Parzysz et al., 2016).
• Periodic battery state broadcasts: Enable users to receive up-to-date availability maps, crucial for minimizing outages and maximizing grid offload (Parzysz et al., 2016, Parzysz et al., 2016).
• Market-equilibrium models: In downlink small-cell EH networks with uncertainty, small cells act as consumers in a competitive market, computing Arrow–Debreu equilibria over user associations and energy states, with prices set via tâtonnement (Maghsudi et al., 2015).

5. Performance Results and Comparative Analysis

Empirical evaluations and simulations across multiple studies decisively show the benefits of energy-based cell association:

• On-grid power reductions: Up to 72% compared to always-on nearest-BS association at low user densities and similar or improved throughput due to reduced inter-cell interference (Rehman et al., 2018).
• Energy and carbon efficiency: Gains up to 2× in energy efficiency and 11.3% in carbon efficiency by adaptive bias optimization, joint user association, and resource assignment (Rehman et al., 2018, Zhao et al., 19 Jan 2026).
• Outage probability and coverage: Availability-aware schemes reduce power outages for edge users by 50–70% over conventional RSS-based association, with comparable SIR coverage (Parzysz et al., 2016).
• Distributed/mechanism design approaches: Mean-field bandit and competitive market models achieve near-centralized optimality in massive IoT deployments without central control or CSI exchange (Maghsudi et al., 2016, Maghsudi et al., 2015).
• Load balancing and fairness: Utility-maximizing schemes integrate per-BS load penalties, yielding high Jain’s index and balancing network resources (Zhou et al., 2015).

6. Advanced Variants, Cross-layer Designs, and Practical Considerations

The landscape of energy-based cell association is rapidly evolving to address multi-tier, multi-objective, and real-world deployment scenarios:

• Joint trajectory and cell association: For mobile users (e.g., UAVs), reinforcement learning jointly optimizes routes and cell handoff to minimize energy, handoffs, and disconnectivity (Cherif et al., 2023).
• Centralized/distributed hybrid resource management: Traffic control-based and regret-learning hybrid schemes for green RANs coordinate within and across local clusters, sustaining high network energy efficiency even under fronthaul and control limitations (Shen et al., 2021).
• Spectrum and interference management: Mixed-integer and convex-relaxation approaches allow global spectrum reallocation, coordinated BS activation, and association, capturing interference coupling via precomputed patterns for large-scale networks (Kuang et al., 2016, Zhuang et al., 2015).
• Implementation and signaling: Battery and bias state broadcasting is periodic and lightweight; policies are robust to estimation errors. Cell-specific bias mapping matches the optimal solution of full resource allocation in practice (Kuang et al., 2016, Parzysz et al., 2016).

---

These energy-based schemes fundamentally shift the user association logic from static signal-based policies to adaptive, resource- and environment-aware frameworks. By exploiting fine-grained BS energy states and embedding these into association decisions, the network dynamically reallocates load, reduces its environmental footprint, and achieves systematic improvements in operational efficiency, scalability, and robustness. This approach underpins the future of sustainable, green wireless access (Rehman et al., 2018, Parzysz et al., 2016, Zhao et al., 19 Jan 2026).