T/CES 230-2023 Electricity artificial intelligence algorithm heterogeneous hardware acceleration technology specification

Scope

Scope: The standard specifies the technical requirements and evaluation methods for heterogeneous hardware acceleration of artificial intelligence algorithms, providing a technical reference and evaluation basis for accelerating line inspection and monitoring algorithms in the power sector. Applicable to the evaluation of hardware acceleration technologies for artificial intelligence frameworks supporting training and inference in the power sector; Main technical content: In recent years, breakthrough progress in artificial intelligence algorithms such as deep learning and reinforcement learning has integrated the theories, technologies, and methods of artificial intelligence with the physical laws and technologies of the power system, forming power artificial intelligence, which has provided significant opportunities for the development of smart grids. With the increasing complexity of artificial intelligence algorithms and the growing number of parameters, the requirements for hardware have also increased. Various artificial intelligence algorithms and heterogeneous hardware acceleration technologies have been incubated in various scenarios of the power sector. By combining software and hardware such as artificial intelligence modules, the computing power has been improved, and a large amount of practical experience has been accumulated in various scenarios. With the rapid increase in the number and types of power artificial intelligence algorithms and heterogeneous hardware acceleration technologies, the definitions of heterogeneous hardware acceleration technologies and the computing power adopted by various manufacturers follow their own internal standards, resulting in uneven control of heterogeneous hardware acceleration technologies by different manufacturers, leading to differences in heterogeneous hardware acceleration technologies, computing power, and interfaces, increasing the difficulty of model application and access, and hindering the standardizati

Development Information

Related Standards