.png){kind=icon}
.png){kind=icon}
Return
Fault prediction method of electric vehicle charging pile based on hierarchical attention mechanism
.png){kind=icon}
OA
AI
Q
D
J
F
DOI：10.1007/s12083-026-02269-9.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
As the core charging equipment of electric vehicles, the reliable operation of charging piles is directly related to user experience and industrial promotion. Accurate prediction of charging pile failures is crucial to ensure charging safety and build a city-level safety protection system. In this paper, a fault prediction model based on hierarchical modeling of temporal features and hierarchical attention mechanism is proposed to construct a multi-level fault prediction system. Firstly, a time series dataset with hierarchical constraints is generated according to the topological relationship of charging piles, and then a differentiated time series model is designed to realize the basic prediction according to the sparsity characteristics of different levels of time series, and then the cross-level features are fused through the hierarchical attention mechanism to constrain the basic prediction results to meet the hierarchical consistency. Experimental results demonstrate that the proposed model achieves the best overall performance in multi-level prediction, with significantly lower prediction error for outliers compared to the baseline models. At the global level, compared with the Transformer, Bi-LSTM, PROFHiT, and HAILS models, the proposed model reduces the Mean Squared Error (MSE) by 21.53%, 17.52%, 31.52%, and 6.61% respectively, the Root Mean Squared Error (RMSE) by 11.57%, 9.15%, 17.35%, and 3.63% respectively, and the Mean Absolute Error (MAE) by 5.70%, 16.76%, 36.60%, and 25.87% respectively.
Keywords:
Charging pile
Failure prediction
Time series forecasting
Hierarchical attention
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
2.6
Papers:
2.2K
Citations:
2.9K
Organization
Cited Papers