PON Energy Saving Proof of Concept and Experiment based on Traffic Prediction using fine granular Telemetry

Electricity costs are a major operational expense for telecom operators, making energy efficiency critical. Strategies like optimizing device power consumption and adjusting operational modes can significantly reduce energy use and expenses. In Passive Optical Networks (PONs), traffic volumes fluctuate rapidly, creating periods of low activity. Leveraging this, PON ports can transit to low-power modes during idle periods to conserve energy. Achieving this requires a real-time traffic load sensing system that dynamically adjusts the Optical Line Terminal (OLT) interface mode based on energy-saving scenarios, enabling a closed-loop mechanism for perception, control, and execution. Accurate traffic prediction is essential for effective energy savings. Traditional SNMP polling (15-minute intervals) is too slow for granular traffic forecasting. China Telecom addressed this by pioneering Technical Specifications for Access Network Telemetry, enabling rapid data collection via active reporting protocols (see ETSI GS F5G 011 V1.1.1 (2022‑11) and ETSI GS F5G 016  V1.1.1 (2023‑06). A Proof of Concept and experiments in a region have been performed. There is no customer complaint claimed when energy-saving is enabled, so services are not affected or the impact is small.

This PoC demonstrates traffic prediction based on telemetry streaming solutions for energy-saving in Passive Optical Network (PON) ports within optical access networks. In essence, the PoC focuses on the energy-saving use-case for PON ports based on traffic load and demonstrates a complete close-loop-action pipeline. This includes telemetry retrieval from data sources (i.e., Optical Line Terminal (OLT)), traffic prediction based on ML models, generation of intelligent energy-saving strategies, and execution of energy-saving actions for PON ports. It is expected to yield tangible results that use the F5G Telemetry Work Item and contribute the AI‑based Prediction and energy-saving (in the future) components of the F5G architecture. 

The figure below shows the implementation of the energy saving solution. Based on traffic data, the PON port with low traffic are selected and on those energy management based on telemetry is implemented.

Finished Energy Saving PoC

Experiment on PON Port Energy Saving Pilot in Jiangxi Province: Around 13 % of the PON port are eligible for energy saving measures due to low traffic. The algorithm is right in 98.7 % of the time that traffic is actually low, based on the prediction. Extrapolating to a whole region with ~141 000 PON ports, on ~19 000 PON ports energy saving management is running. That saves approximately 140 MWh annually, or 7.3 kWh per eligible for energy saving PON port per year.

 

Driving the Optical Networks towards Higher Autono...