Can AI/ML improve smart grid automation?

Can AI/ML improve smart grid automation?

Berlin, Aug. 11, 2022

Actually, the use of AI and ML should make the automation of the network go smoothly. This has not been confirmed in this way. The experience in a nutshell: the greater the network complexity, the more and more frequent the errors. This has led to the fact that the functions in 3GPP were constantly extended. This is especially the case with Releases 17 and 18. ‚This issue is competently addressed in the latest whitepaper from Ericsson „5G Advanced: Evolution towards 6G“….

https://www.ericsson.com/en/reports-and-papers/white-papers/5g-advanced-evolution-towards-6g

We have summarized the study in a way that makes the key messages clear:

Intelligent network automation: because networks are becoming increasingly complex, traditional solutions such as reconfiguration are no longer a guarantee of success. Moreover, this method also inefficient and costly.

By using AI and ML, appropriate solutions could be found.

An important point is data collection. This is the only way to optimize lifecycle management. With the use of 5G Advanced, we are now also able to optimize the standardized interfaces for data collection.

As part of the Release 17 study, three use cases were identified that relate to improving RAN performance through the use of AI/ML techniques.These use cases are as follows: 1) network energy savings; 2) load balancing; and 3) mobility optimization.

In this regard, Ericsson experts state “ The selected use cases can be supported by enhancements to the current NR interfaces that target performance improvements through AI/ML capabilities in the RAN while maintaining the 5G NR architecture. One of the goals is to incentivize vendors in terms of innovation and competitiveness by keeping the implementation of the AI model specific.“

The 3GPP TSG RAN has selected three use cases to explore the potential performance improvements of the air interface using AI/ML techniques, such as beam management, improving channel state information feedback, and improving positioning accuracy for different scenarios. AI/ML-based methods can provide advantages over traditional methods at the radio interface. The challenge is to define a unified AI/ML framework for the air interface through appropriate AI/ML model characterization at different levels of collaboration between gNB and UE.

AI/ML in the 5G core: 5G Advanced will provide further enhancements to the architecture for analyzing and managing the lifecycle of ML models, for example, to improve model correctness. The advances in the architecture for analytics and data collection serve as a good foundation for AI/ML-based use cases within the various network functions (NFs). Additional use cases are being explored where NFs use analytics to support their decision making, such as generating UE policies for network slicing using Network Data Analytics Functions (NWDAF).

Translated with www.DeepL.com/Translator (free version)