Deep learning chips: challenges and opportunities for ubiquitous power internet of things

Abstract

Tasks of Ubiquitous Power Internet of Things run through the power generation, transmission, transformation, distribution, electricity use and other links, requiring advanced communication, artificial intelligence, big data and other technologies. Deep learning chips provide computational power for algorithm execution and data processing, which are indispensable foundations and basic components for intelligent terminals. Therefore, this paper summarizes the challenges and opportunities faced by deep learning chips in the construction of ubiquitous power Internet of Things. Firstly, the four parts of ubiquitous power Internet of Things including terminal layer, network layer, platform layer and application layer are described. Secondly, the key technologies of deep learning technology and deep neural network accelerator involved in deep learning chips are summarized. Finally, the research work of deep learning chips for ubiquitous power Internet of Things is surveyed. The main functions and existing problems are discussed, and the future research work is proposed.