Professor Ryoichi Shinkuma of the Department of Computer Science, Faculty of Engineering, Shibaura Institute of Technology has devised a system that accurately predicts the communication volume of mobile phone base stations and switches the operation and suspension of the base stations accordingly.
In the world, similar research is underway to reduce the power consumption of global base stations (4), which is estimated to be 5-2016 billion kwh.In Japan, the total power consumption related to base stations of the three major domestic mobile phone companies is about 3 billion kWh (estimated from the latest sustainability reports of each company).On the other hand, 51G base stations, whose services are being developed, have a narrower coverage than 5G, and it is necessary to increase the number of base stations.In addition, although the power supply is being re-energized for decarbonization, it is necessary to reduce the power consumption by turning off the power according to the communication volume because the supply amount is unstable.However, it is not possible to learn the communication record of the base station that is inactive, and the problem is that the prediction accuracy of the communication amount is lowered.
Professor Shinkuma estimates the most important records from the records from the base stations in operation and incorporates them into learning, and combines general machine learning and simple feature selection to achieve high prediction accuracy. Achieved.Each stopped base station can reduce power consumption by about 6%, and control according to communication volume is important for reducing CO2 emissions.It is also possible to supply base station power with renewable energy whose power supply depends on the climate.
In the future, we will distinguish various communication application data such as texts, videos, and Web pages, extract the characteristics of each, create machine learning models, and improve prediction accuracy.For example, online conferences are long, and the communication time differs depending on the application, so control that takes into account its characteristics is also possible.
Reference: [Shibaura Institute of Technology] Shibaura Institute of Technology Highly accurate traffic forecasting by AI makes it possible to replace the power of base stations with renewable energy.