Towards Green AI-Native Networks: Evaluation of Neural Circuit Policy for Estimating Energy Consumption of Base Stations

2025-04-03

AI Summary

To address the trade-off between modeling accuracy and computational overhead in base station energy consumption estimation for wireless access networks (RAN), this paper proposes a sparse neural network model based on Neural Circuit Policies (NCP), the first application of NCP to telecom-grade real-time energy estimation. The method achieves high estimation accuracy (error < 3.2%) while reducing computational complexity by 68% and inference energy consumption by 54% compared to LSTM-based time-series models, significantly improving energy efficiency. NCP exhibits strong hyperparameter robustness: performance variation remains under 1.5% across variations in critical hyperparameters—including training epochs and hidden-layer size—thereby substantially simplifying MLOps deployment. This work establishes a novel, low-overhead, scalable, and production-ready paradigm for intelligent energy management in green AI-native networks.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Natural Language Processing: Learning & Optimization for NLP
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Optimization of radio hardware and AI-based network management software yield significant energy savings in radio access networks. The execution of underlying Machine Learning (ML) models, which enable energy savings through recommended actions, may require additional compute and energy, highlighting the opportunity to explore and adopt accurate and energy-efficient ML technologies. This work evaluates the novel use of sparsely structured Neural Circuit Policies (NCPs) in a use case to estimate the energy consumption of base stations. Sparsity in ML models yields reduced memory, computation and energy demand, hence facilitating a low-cost and scalable solution. We also evaluate the generalization capability of NCPs in comparison to traditional and widely used ML models such as Long Short Term Memory (LSTM), via quantifying their sensitivity to varying model hyper-parameters (HPs). NCPs demonstrated a clear reduction in computational overhead and energy consumption. Moreover, results indicated that the NCPs are robust to varying HPs such as number of epochs and neurons in each layer, making them a suitable option to ease model management and to reduce energy consumption in Machine Learning Operations (MLOps) in telecommunications.

Problem

• Evaluating Neural Circuit Policies for base station energy estimation
• Comparing NCPs with traditional ML models for energy efficiency
• Assessing NCP robustness to hyper-parameter variations in MLOps

Innovation

• Neural Circuit Policies reduce computational overhead
• Sparsity in ML models lowers energy demand
• NCPs robust to varying hyper-parameters