Construction and Practice of Power Grid Dispatching Intelligent Defense System Based on Multivariate Data Fusion and Deep Learning

Abstract

In this paper, multivariate data fusion and deep learning artificial intelligence algorithms are introduced into the power grid dispatching business. Combined with dispatching procedures and instruction specifications, the dispatching phone voice is converted into information in text through a voice recognition platform in real time. Key words are extracted through semantic understanding, multivariate data fusion and deep learning for the recognized text to identify and detect business scenarios. The extracted key information is used to verify and prevent errors on the grid operation platform based on the grid real-time status verification and dispatching business scenario rules. Alarms and prompts for irregular and incorrect dispatching instructions is proposed through the voice platform. Through the operation and operation of historical big data, constantly learning and discovering laws, a perfect knowledge map of dispatching business is established, and the accuracy of voice recognition and the accuracy of scene detection is continuously improved, thereby the function of 24-h security supervision of dispatching telephone business are realized. The system realizes full chain intelligent anti-misoperation management and control of the entire process status and power flow, which can solve the problems of wrong dispatching orders and misoperations when the phone is ordered due to inadequate monitoring, irregular orders, and incorrect understanding of scheduling instructions.