Electric power Internet-of-Things dynamic baseline alarm system and method based on machine learning

The invention discloses an electric power Internet-of-Things dynamic baseline alarm system based on machine learning, and the system comprises: a data collection module which is used for collecting preset M-day electric power Internet-of-Things platform database CPU utilization rate data in an electric power Internet-of-Things platform database, and randomly selecting N days of CPU utilization rate data of the electric power Internet-of-Things platform database from the M days of CPU utilization rate data of the electric power Internet-of-Things platform database to form a test data set D; a data processing module which splits the CPU utilization rate data of the power Internet-of-Things platform database in N days into N test data subsets by taking the day as a unit; a CPU utilization rate data baseline function calculation module which obtains CPU utilization rate data dynamic upper and lower baseline functions; and a threshold warning boundary determination module which obtains a threshold warning boundary