Power marketing data analysis method based on AI large model

CN120634628BActive Publication Date: 2025-11-07SHANGHAI WANGMAI INFORMATION TECH GRP CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511127274.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-13
Publication Date
2025-11-07
Estimated Expiration
2045-08-13

AI Technical Summary

Technical Problem

Existing methods for analyzing electricity marketing data suffer from inconsistent coding of multi-source data, difficulties in processing cross-modal features, and a lack of effective integration of knowledge graphs and expert rules. These issues lead to inconsistent modeling inputs, difficulties in integration, and a lack of semantic relevance and causal interpretability.

Method used

We adopt an AI-based large-scale model for electricity marketing data analysis. We generate unified encoded data through a spatiotemporal alignment module and a pre-trained word embedding model. We use an electricity knowledge graph and graph attention mechanism to decouple and fuse features, and combine a dynamic weighted gating network for multi-task decision-making.

Benefits of technology

It achieves highly consistent coding of multimodal data, improves the generalization ability and robustness of data-driven models, significantly enhances the cognitive ability of strategy adaptability and fault propagation mechanisms, and improves the interpretability of abnormal user identification and strategy recommendation, as well as the user retention ability of the power marketing system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120634628B_ABST
    Figure CN120634628B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of electric power marketing, in particular to an electric power marketing data analysis method based on an AI large model, which comprises the following steps: collecting structured data and unstructured data in an electric power marketing system, generating unified coding data after the structured data and the unstructured data are processed through space-time alignment and a pre-trained word embedding model; inputting the unified coding data into a pre-trained electric power field large model, extracting static, dynamic and semantic features through multi-modal fusion, feature decoupling and semantic anchoring; combining an electric power knowledge graph to construct an entity link feature pair, enhancing and fusing feature expression through a graph attention mechanism; finally, inputting a dynamic weight gate network, outputting an abnormal user identification label, a demand response strategy and a customer loss early warning probability, and executing strategy optimization under specific conditions. The application can be widely applied to risk identification, strategy formulation and user behavior prediction tasks in electric power marketing.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Power supply service risk identification system based on big data analysis

    CN119809319A

  • Abnormality detection method and system for redundancy-eliminating perception multi-dimensional power utilization data fusion

    CN119885013A