Energy-Aware Slice Selection

The central decision logic that balances energy efficiency against SLA compliance. It evaluates candidate slices by computing a weighted utility function that considers:

• Current power consumption per slice (kW/Gbps)
• Requested QoS profile (5QI, GBR, latency budget)
• Slice load and remaining capacity
• Carbon intensity of the underlying power grid The engine enforces hard constraints—a URLLC slice will never be selected for an eMBB session, even if it is idle—while optimizing soft targets like minimizing the aggregate Power Usage Effectiveness (PUE) across the RAN.

Energy-aware selection depends on a continuous feed of granular power metrics from the Network Data Analytics Function (NWDAF) and O-RAN Service Management and Orchestration (SMO) framework. Key inputs include:

• Per-slice Resource Block utilization and muting ratios
• Base station Cell DTX and Sleep Mode Coordination states
• Dynamic Voltage and Frequency Scaling (DVFS) states of virtualized network functions
• Accelerator offloading status for L1 processing This telemetry pipeline enables the selection function to build an accurate, sub-second Slice-Level Energy Model that reflects the true marginal cost of admitting a new PDU session.

Energy-aware selection is not a one-time event at session establishment. The function continuously monitors active sessions and initiates Slice Remapping when:

• A GBR slice enters a low-utilization period, allowing sessions to be consolidated onto fewer, higher-efficiency slices
• The carbon intensity of the local grid spikes, triggering a temporary migration to slices hosted on infrastructure powered by greener energy sources
• A slice's Adaptive Bandwidth Part (BWP) configuration changes, altering its power profile Remapping is executed via standard 5G procedures with session continuity preserved, ensuring no perceptible impact on the user.

Beyond instantaneous power draw, the selection function incorporates the Slice Carbon Footprint as a decision metric. This requires integration with external grid intensity APIs (e.g., Electricity Maps) and internal data center PUE monitoring. The function can:

• Apply a carbon penalty factor to slices running in high-emission regions
• Prioritize Edge Slices powered by on-site renewables during peak grid carbon intensity
• Generate per-tenant carbon accounting reports for ESG compliance This transforms slice selection from a pure energy play into a verifiable sustainability mechanism aligned with corporate net-zero targets.

Energy-aware selection operates within a broader Closed-Loop Slice Optimization framework. The selection decisions feed into an O-RAN Non-Real-Time RIC (Non-RT RIC) that:

• Adjusts slice resource allocations and Resource Overbooking ratios based on long-term selection patterns
• Recommends Slice Decommissioning for chronically underutilized instances
• Tunes the weights of the multi-objective policy engine using reinforcement learning to adapt to seasonal traffic patterns This ensures that the selection function's decisions actively reshape the network's energy posture over hours and days, not just milliseconds.

The process of dynamically reassigning an active user equipment session from one network slice instance to another to optimize for changing service requirements, load conditions, or energy efficiency targets. This is the direct execution mechanism triggered by the energy-aware selection function. Key considerations include:

An automation framework where a policy-driven controller continuously monitors slice KPIs, analyzes deviations from the desired state using AI, and automatically executes corrective reconfiguration actions without human intervention. Energy-aware slice selection operates as a critical decision-making component within this broader closed loop, translating sustainability policies into real-time steering actions that balance performance with power consumption targets.

A 5G core network function that collects and analyzes network data from various sources using machine learning to provide predictive analytics on slice load, performance, and user behavior. The NWDAF supplies the energy-aware slice selection function with predictive insights on future slice utilization and energy consumption trends, enabling proactive rather than reactive steering decisions that preempt congestion and power spikes.

A formal contract between a slice tenant and a network operator that defines the quantifiable performance metrics a network slice instance must deliver. The energy-aware selection function must operate within the strict boundaries of these SLAs, ensuring that steering a user to a more energy-efficient slice does not violate guaranteed bit rate, latency, or reliability commitments. This creates a constrained optimization problem balancing sustainability with contractual obligations.

A centralized control strategy that synchronizes the activation of low-power states across multiple network components within a slice to maximize energy savings. Energy-aware slice selection complements this by consolidating users onto fewer slices, enabling the decommissioned or emptied slices to enter deep sleep modes or cell discontinuous transmission states, amplifying the total network energy reduction.