适用于新能源汽车动力电池的霍尔阵列电流传感器研究

【作者】 沈悦；

【导师】 唐玥；

【作者基本信息】 南京信息工程大学 ， 电子信息（专业学位）， 2025， 硕士

【摘要】 近年来,为满足新能源汽车主机厂提出的动力电池小型化、轻量化、持续降本的需求,霍尔阵列电流传感器逐渐成为新兴的适用于动力电池的电流测量方案。但是目前霍尔阵列电流传感器仍然存在安装占用面积大、初始精度低、偏心误差和倾斜误差大等问题,因此本论文针对适用于新能源汽车动力电池的霍尔阵列电流传感器进行深入研究,设计了完整的电流测量方案。本论文的主要工作内容如下:首先,针对目前霍尔阵列电流传感器安装占用面积大的问题,结合新能源汽车动力电池中矩形导体的几何特征,提出一种基于均匀曲线段长度法(Uniform Curve Segment Length Method,UCSL)的椭圆阵列参数模型,并通过MATLAB计算不同霍尔元件个数以及椭圆阵列尺寸与电流误差关系,得到鲁棒性更好的霍尔元件椭圆阵列参数。根据实际应用需求选择合适的霍尔芯片并进行最小系统电路、电源模块、DC-DC降压模块、信号采集模块以及通讯模块等模块设计,最终完成霍尔元件椭圆阵列电流传感器硬件电路设计。其次,针对霍尔阵列电流传感器初始误差较大的问题,设计一种全温域标定算法。将接收到的不同环境温度下的电流报文转换为数字信号,计算不同环境温度下的电流测量值和电流偏差,并进行最小二乘拟合,得到对应的电流偏差拟合值,然后计算单位温度变化产生的温漂系数,并结合拟合值和温漂系数对电流测量值进行补偿。通过全温域标定将电流传感器的初始误差控制在3‰以内,进而提高电流传感器的测量精度。然后,针对霍尔阵列电流传感器偏心误差和倾斜误差较大,进而影响电流传感器整体性能的问题,提出一种基于灰狼算法(Grey Wolf Optimizer,GWO)优化BP神经网络的GWO-BP误差优化算法。通过构建二维磁场模型和三维磁场模型,分析误差来源并确定导体偏心和倾斜参数,在此基础上进行积分区域重建和电流反演,得到电流积分模型。同时在导体状态参数估计模型中,引入灰狼算法(GWO)优化BP神经网络的初始权值和阈值,提高其回归能力和预测精度,并结合电流积分模型实现偏心误差与倾斜误差优化,进而提高霍尔阵列电流传感器的抗干扰能力。最后,设计有限元仿真环境和测试平台,验证本论文设计的误差优化算法以及霍尔阵列电流传感器的性能。仿真与实验结果表明,本论文设计的电流传感器具有较好的可靠性,全温域下的测量精度均保持在3‰以内;本论文设计的误差优化方算法具有较好的鲁棒性,导体X向偏心产生的电流误差减小65.07%,导体Y向偏心产生的电流误差减小45.74%,导体偏离Z轴产生的电流误差减小76.15%。更多还原

【Abstract】 In recent years,in order to meet the demand for miniaturization,lightweight,and continuous cost reduction of power batteries put forward by new energy vehicle OEMs,Hall array current sensors have gradually become an emerging current measurement solution applicable to power batteries.However,the current Hall array current sensors still have problems such as large installation area,low initial accuracy,and large eccentricity and tilt errors etc.Therefore,this paper carries out an in-depth study aiming at the Hall array current sensors applicable to the power battery of new energy vehicles,and designs a complete current measurement scheme.The main work of this paper is as follows:Firstly,aiming at the current problem of large area occupied by the installation of Hall array current sensors,combining with the geometrical characteristics of rectangular conductors in power battery for new energy vehicles,an elliptical array parameter model based on the Uniform Curve Segment Length Method(UCSL)is proposed.And the elliptic array parameters of Hall elements with better robustness are obtained by calculating different numbers of Hall elements and the relationship between elliptic array size and current error through MATLAB.According to the actual application requirements to select the appropriate Hall chip and design the minimum system circuit,power supply module,DC-DC step-down module,signal acquisition module and communication module and so on,and finally complete the Hall element elliptic array current sensor hardware circuit design.Secondly,aiming at the problem of large initial error of Hall array current sensor,a full temperature domain calibration algorithm is designed.The current telegrams received at different ambient temperatures are converted into digital signals,the current measurements and current deviations at different ambient temperatures are calculated and fitted by least squares to obtain the corresponding fitted values of current deviations,then the temperature drift coefficients generated by the unit temperature change are calculated,and the combined fitted values and the temperature drift coefficients are used to compensate for the current measurements.The initial error of the current sensor is controlled within 3‰by full temperature domain calibration,which in turn improves the measurement accuracy of the current sensor.Then,aiming at the problem that the Hall array current sensor has large eccentricity error and tilt error,which in turn affects the overall performance of the current sensor,a GWO-BP error optimization algorithm based on the Grey Wolf Optimizer(GWO)optimized BP neural network is proposed.By constructing a two-dimensional magnetic field model and a three-dimensional magnetic field model,the error sources are analyzed and the conductor eccentricity and tilt parameters are determined,based on which the integral region reconstruction and current inversion are performed to obtain the current integral model.Meanwhile,in the conductor state parameter estimation model,the Gray Wolf Optimizer(GWO)is introduced to optimize the initial weights and thresholds of the BP neural network to improve its regression capability and prediction accuracy,and combined with the current integral model to realize the optimization of eccentricity error and tilt error,which in turn improves the anti-interference capability of the Hall array current sensor.Finally,a finite element simulation environment and a test platform are designed to verify the error optimization algorithm designed in this paper and the performance of the Hall array current sensor.Simulation and experimental results show that the current sensor designed in this paper has good reliability,and the measurement accuracy under the full temperature domain is kept within 3‰;the error optimization square algorithm designed in this paper has good robustness,and the current error generated by the conductor’s X-direction eccentricity is reduced by 65.07%,that generated by the conductor’s Y-direction eccentricity is reduced by45.74%,and that generated by the conductor’s deviation from the Z-axis is reduced by 76.15%.更多还原