Inference method for power large-scale model operation based on distributed computing resource scheduling

Electric power computing resources are distributed in the respective regions of provincial data centers and transmission and transformation equipment, and the scheduling efficiency of computing resources is limited by conditions such as long distribution locations. Composite data model inference has been widely used in power equipment maintenance analysis, and large-scale model inference requires the use of distributed computing resources. In order to improve the scheduling and fusion efficiency of distributed computing power resources, this paper adopts a lightweight language model to infer and calculate the scheduling parameters of computing power resources in different regions. At the same time, in order to make more rational use of computing power resources, the intelligent models and algorithms used for power equipment maintenance are optimized. Configure the adaptation parameters of the model according to different application scenarios, and adjust the fusion method of input data for the fault diagnosis model and the classification strategy for the fault prediction model separately. Using a fine-tuning language model to map the predicted results to the semantics of power business, and then enhancing the rationality and interoperability of the answer results based on historical knowledge memory, ultimately pushing the reasoning results of the reasons and the classification results of the equipment defect diagnosis classifier to the equipment diagnosis designer.