Presentation Abstract

AI is expected to change the world and Power Electronics should not be an exception. But in engineering 99% correct statements or calculation is not acceptable; a very small design error can destroy a power converter and the expensive systems they are connected to. We will first show that combining flexible AI tools with strict design rules and systematic check is needed to obtain reliable design tools for converter and their control. On the other hand, hyperscale datacenters with ever increasing power consumption become a major application field for Power Electronics and efforts must be made to limit negative impacts: energy consumption, resources and material needs, heat generation, water consumption, etc. This challenge will be addressed, and it will especially be shown that TransformerLess Partial Power Converters allow dividing by two these side effects for Battery Backup Units and Capacitor Bank Units of the datacenters thus saving hundreds of MWh, tons of water, copper, etc. In the last part we will show that such TLPPCs for datacenters can also be designed using AI-aided tools incorporating engineering check so that optimal solutions can be found and even tracked when the state of the technology is making progress.