| 摘要: |
| 随着电动汽车的迅速发展，充电基础设施成为新型城市基建的发力方向，而充电桩的选址定容是充电基础设施建设的核心问题之一。在此过程中，电动汽车充电行为的预测尤为重要。但由于用户的充电行为数据具有隐私性，而当前研究构建的预测模型中欠缺对这一重要因素的考虑，导致敏感信息泄露。针对此问题，提出了一种隐私保护的电动汽车充电行为及充电桩需求安全预测方法。首先，通过充电行为拟合充电负荷特性的概率密度函数；其次，结合Paillier同态加密，利用蒙特卡罗算法进行隐私保护的充电负荷预测；最后，根据负荷预测结果，采用粒子群优化算法对区域充电桩布局进行优化遍历，安全地得到充电桩预测结果。仿真实验结果表明，所提方法能够在保障用户隐私的情况下取得良好的负荷预测准确度，其中最大误差为11.9%，最小误差为2.8%；相较于现有方案，安全性提升的同时预测误差降低了32.98%。 |
| 关键词: 车桩网 数据隐私安全 充电负荷预测 同态加密 充电桩布局优化 粒子群优化 |
| DOI：10.20079/j.issn.1001-893x.250202001 |
|
| 基金项目:国家电网有限公司总部管理科技项目（5700-202441247A-1-1-ZN） |
|
| A Secure Charging Behaviour Forecasting Method with Privacy Protection |
| GUO Jing,GU Zhimin,ZHU Daohua,LIANG Wei |
| (State Grid Jiangsu Electric Power Co.,Ltd. Research Institute,Nanjing 211000,China) |
| Abstract: |
| With the rapid development of electric vehicles(EVs),charging infrastructure has become the policy priority of new infrastructure.The location and capacity planning of charging plies is the core issue,where the prediction of charging load is crucial.However,due to the privacy of the user搒 charging behavior data,the prediction models constructed in the current works lack consideration of this important factor,resulting in the leakage of sensitive information.To address this problem,a privacy-preserving method for EV charging load and charging pile number forecasting is proposed.First,the probability density function of charging load characteristics is fit by charging behavior.Second,Monte Carlo algorithm is used for predicting charging load with privacy protection,which combines the algorithm with the Paillier encryption scheme.Finally,according to the load prediction result,particle swarm optimization is used to optimize traversal in the regional charging pile layout and then the charging pile number forecasting result is obtained.The simulation results show that the proposed method can achieve good charging load prediction accuracy while protecting charging data privacy.The maximum error is 11.9%,and the minimum error is 2.8%.Compared with the existing schemes,the proposed solution enhances the security and meanwhile reduces the prediction error by about 32.98%. |
| Key words: car pile net data privacy and security charging load forecasting homomorphic encryption optimization of charging load layout particle swarm optimization |
《电讯技术》编辑部 