On the Impact of PRB Load Uncertainty Forecasting for Sustainable Open RAN

TL;DR

This work examines PRB load uncertainty in sustainable Open RAN and evaluates probabilistic forecasting approaches—SFF, DeepAR, and Transformer—against a deterministic LSTM baseline within a realistic power-consumption framework. By modeling the conditional distribution $P(\mathbf{y}_{t_0:T}|\mathbf{y}_{1:t_0-1})$, the study demonstrates that DeepAR yields lower forecast uncertainty and better temporal accuracy, enabling substantial power savings through percentile-based provisioning. The results highlight the trade-off between over-/under-provisioning and energy efficiency: lower percentiles reduce resource use but risk under-provisioning, while higher percentiles secure QoS at the cost of higher power. The findings support integrating probabilistic forecasts as AI-enabled rApps in non-real-time RICs to improve energy-aware resource management in Open RAN, with future work on real-world validation and integration with broader energy optimization strategies.

Abstract

The transition to sustainable Open Radio Access Network (O-RAN) architectures brings new challenges for resource management, especially in predicting the utilization of Physical Resource Block (PRB)s. In this paper, we propose a novel approach to characterize the PRB load using probabilistic forecasting techniques. First, we provide background information on the O-RAN architecture and components and emphasize the importance of energy/power consumption models for sustainable implementations. The problem statement highlights the need for accurate PRB load prediction to optimize resource allocation and power efficiency. We then investigate probabilistic forecasting techniques, including Simple-Feed-Forward (SFF), DeepAR, and Transformers, and discuss their likelihood model assumptions. The simulation results show that DeepAR estimators predict the PRBs with less uncertainty and effectively capture the temporal dependencies in the dataset compared to SFF- and Transformer-based models, leading to power savings. Different percentile selections can also increase power savings, but at the cost of over-/under provisioning. At the same time, the performance of the Long-Short Term Memory (LSTM) is shown to be inferior to the probabilistic estimators with respect to all error metrics. Finally, we outline the importance of probabilistic, prediction-based characterization for sustainable O-RAN implementations and highlight avenues for future research.

Figures (5)

• Figure 1: O-RAN architecture with sustainable radio resource allocation as rApp.
• Figure 2: Percentage of over/under-estimated PRBs.
• Figure 3: Simple-Feed-Forward estimator with power saving.
• Figure 4: DeepAR estimator with power saving.
• Figure 5: Transformer estimator with power saving.