一种基于机理约束与深度学习融合的充电场站损耗预测方法与系统
By integrating mechanistic constraints with deep learning, a charging station loss prediction system was constructed. This system addresses the issues of insufficient granularity and cross-station adaptability in existing loss analysis technologies, enabling refined management and intelligent operation and maintenance, and improving the accuracy and adaptability of predictions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID JIANGSU ELECTRIC VEHICLE SERVICE CO LTD
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for analyzing charging station losses are insufficient to accurately characterize segmented loss features, lack consideration for the coupled effects of multiple factors, have inadequate generalization capabilities across charging stations, lack calibration mechanisms after deployment, struggle to maintain long-term prediction accuracy, and lack online loss prediction and anomaly identification capabilities.
A unified closed-loop technical route based on the fusion of mechanism constraints and deep learning is constructed, which includes multi-site data preprocessing, power flow path mechanism modeling, mechanism prior feature generation, site-level calibration and adaptive updating. Loss prediction is performed through a mechanism-constrained deep learning model, and combined with site-level calibration and drift-triggered adaptive updating mechanism, loss prediction under cross-site and long-term operation conditions is realized.
It improves the accuracy, physical consistency, and cross-site adaptability of loss prediction, enhances the engineering practicality of the model, can accurately characterize the segmented losses of the power supply link, reduces model deployment errors, and supports online loss prediction and anomaly identification.
Smart Images
Figure CN122402292A_ABST