Cell Site Sleep-Wake Control for Throughput-Safe Energy Savings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cellular networks face inefficiencies in energy consumption due to static sleep and wakeup parameters that fail to optimize energy savings while maintaining customer experience, as usage patterns vary significantly across different times and locations.
Innovation Solution
A machine learning-based system that utilizes a throughput prediction model, such as a deep neural network, to forecast optimal sleep-wake configurations for cellular antennas, adjusting parameters on a per-cell, per-day basis to maximize sleeping time without impacting user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If static sleep and wakeup parameters are used for cellular antennas, then device complexity is reduced and ease of operation is improved, but energy savings are insufficient and adaptability to varying usage patterns deteriorates
Solution Approach 1:
The patent implements dynamic sleep and wakeup parameters that adapt to varying usage patterns across different times and locations. The system continuously monitors traffic conditions and adjusts antenna sleep-wake configurations in real-time, transitioning from static to dynamic parameter management to optimize energy savings while maintaining service quality.
Solution Approach 2:
The system employs machine learning models that automatically analyze historical and real-time usage data to predict optimal sleep-wake configurations without manual intervention. The antennas self-adjust their operational states based on predicted traffic patterns, eliminating the need for complex manual configuration while achieving adaptive energy optimization.
2Use of energy by moving object
If manual sleep-wake parameter settings are used, then device complexity is minimized, but energy savings are limited and adaptability to different times and locations deteriorates
Solution Approach 1:
The system uses machine learning models to predict future traffic patterns and proactively configures sleep-wake parameters before peak usage periods occur. By analyzing historical data and forecasting future conditions, the system prepares optimal configurations in advance, enabling energy savings during low-traffic periods while ensuring service availability when demand increases.
Solution Approach 2:
The patent implements dynamic adjustment of sleep-wake parameters based on predicted usage patterns, location-specific characteristics, and time-of-day variations. The system modifies operational parameters such as sleep duration, wakeup timing, and antenna activation states to adapt to different conditions, achieving both energy efficiency and adaptability simultaneously.
3Use of energy by moving object
If cells are placed in inactive state to save energy, then energy consumption is reduced, but user throughput may be impacted
Solution Approach 1:
The system continuously monitors actual user throughput and energy consumption, using this feedback to refine machine learning predictions and adjust sleep-wake configurations. By implementing closed-loop control, the system ensures that energy-saving actions do not degrade service quality beyond acceptable thresholds, dynamically balancing energy savings with throughput maintenance.
Solution Approach 2:
The system places only certain cells in inactive state during specific time periods rather than uniformly inactivating all cells, based on predicted local usage patterns. This partial action approach allows energy savings in low-traffic areas while maintaining service in high-traffic areas, achieving energy optimization without significantly impacting overall user throughput.
Data Source
AI summary
A processing system may apply a data set comprising utilization metrics of a cells of a cell sector to a throughput prediction model to obtain a first predicted throughput for the cell sector for a designated future time period and for all cells being in an active state. The processing system may generate a first modified data set simulating a first cell being placed in an inactive state, by distributing utilization metrics of the first cell over at least one additional cell, and may apply the first modified data set to the throughput prediction model to obtain a second predicted throughput. The processing system may then determine that the second predicted throughput meets a threshold throughput that is based on the first predicted throughput, and transmit at least one instruction to place the first cell in the inactive state for the designated future time period in response to the determining.