基于企业维度的电动汽车充电桩网络异常识别数据

本数据在充电桩网络故障诊断与设备运维领域具有以下应用价值：
1.对平台的价值（自动化运维决策）：支撑实时识别网络异常桩，当同一桩企-桩型设备连续多次输出“1”时，自动触发维修工单；结合场站ID维度优先调度维修资源至问题桩，提升运维效率。
2.对桩企商家的价值（设备质量改进）：本数据可支撑对企业维度异常率排名，为定位硬件设计缺陷，驱动产品迭代，联动供应链追溯问题批次设备提供数据支持。
3.对场站商家的价值（故障精准归因）：区分单桩硬件异常与场站级网络故障（如同一场站＞30%桩异常且城市分布集中时，预警SIM卡集体欠费），缩短故障恢复时长，减少充电订单损失。1.数据采集。原始数据经授权合法获取并使用，采集字段包括桩企名称、桩型、充电桩ID、场站ID。
2.特征加工。以1小时为间隔，滚动截取每台桩前30小时数据形成滑动窗口。加工字段包括：①离线时长占比：窗口内离线总时长/30小时；②离线次数：窗口内离线事件累计次数；③单次最大离线时长：窗口内最长单次离线持续时间。
3.样本构建。按桩企-桩型维度划分正负样本：正样本为网络故障桩，负样本为同维度内随机抽取的正常桩，比例1:3。样本特征包含加工字段及桩企、桩型，输出目标为网络状态标签（0-正常，1-异常）。
4.模型训练。采用XGBoost算法构建二分类模型，输入加工后的特征字段，输出网络状态结论。训练时通过桩企维度交叉验证增强对特定桩型硬件故障的敏感性，利用SHAP值剔除场站ID等非企业维度的干扰因素。
5.异常诊断。实时滑动窗口数据输入模型，输出0/1异常结论（0-正常，1-异常）。

This dataset holds the following application values in the domain of charging pile network fault diagnosis and equipment operation & maintenance:
1. Value for the platform (automated operation and maintenance decision-making): Support real-time identification of network-abnormal charging piles on the platform. When the same pile enterprise-pile type device outputs "1" consecutively for multiple times, automatically trigger a maintenance work order; prioritize dispatching maintenance resources to problematic charging piles based on the station ID dimension, thereby improving operation and maintenance efficiency.
2. Value for pile enterprise merchants (equipment quality improvement): This dataset can support the ranking of abnormal rates by enterprise dimension, providing data support for locating hardware design defects, driving product iteration, and tracing problematic batch equipment in conjunction with the supply chain.
3. Value for station merchants (accurate fault attribution): Distinguish between single-pile hardware abnormalities and station-level network faults (for example, issue an early warning for collective SIM card arrears when more than 30% of piles in the same station are abnormal and the urban distribution is concentrated). Shorten fault recovery time and mitigate losses incurred from charging orders.

1. Data Collection. The original data is legally obtained and used with authorization. The collected fields include pile enterprise name, pile type, charging pile ID, and station ID.
2. Feature Engineering. Use a 1-hour interval to rollingly intercept the first 30 hours of historical data for each individual charging pile to construct a sliding window. The processed fields include: ① Offline duration ratio: total offline duration within the window / 30 hours; ② Offline times: cumulative number of offline events within the window; ③ Maximum single offline duration: the longest single offline duration within the window.
3. Sample Construction. Partition positive and negative samples according to the pile enterprise-pile type dimension: positive samples are network fault charging piles, and negative samples are normal piles randomly selected within the same dimension, with a ratio of 1:3. The sample features include the processed fields, pile enterprise and pile type, and the output target is the network status label (0 - normal, 1 - abnormal).
4. Model Training. Adopt the XGBoost algorithm to build a binary classification model, input the processed feature fields, and output the network status conclusion. During training, use pile enterprise-level cross-validation to enhance the model's sensitivity to hardware faults of specific pile types, and utilize SHAP values to eliminate interference factors that are not at the enterprise level, such as station ID.
5. Anomaly Diagnosis. Feed real-time sliding window data into the trained model and output 0/1 anomaly conclusions (0 - normal, 1 - abnormal).

搜集汇总

背景与挑战

背景概述

该数据集是一个基于企业维度的电动汽车充电桩网络异常识别数据，包含1501条实时更新的企业数据，主要用于充电桩网络故障诊断与设备运维。数据集通过XGBoost算法构建二分类模型，输出网络状态结论，对平台、桩企商家和场站商家具有不同的应用价值。