Tang Yue, Lu Jiajia, Shen Yue
School of Internet of Things Engineering, Wuxi University, Wuxi 214105, China.
School of Automation, Nanjing University of Information Science and Technology, Nanjing 210044, China.
Sensors (Basel). 2025 May 15;25(10):3116. doi: 10.3390/s25103116.
Accurate current sensing in rectangular conductors is challenged by mechanical deformations, including eccentricity (X/Y-axis shifts) and inclination (Z-axis tilt), which distort magnetic field distributions and induce measurement errors. To address this, we propose a bio-inspired error compensation strategy integrating an elliptically configured Hall sensor array with a hybrid Grey Wolf Optimizer (GWO)-enhanced backpropagation neural network. The eccentric displacement and tilt angle of the conductor are quantified via a three-dimensional magnetic field reconstruction and current inversion modeling. A dual-stage optimization framework is implemented: first, establishing a BP neural network for real-time conductor state estimations, and second, leveraging the GWO's swarm intelligence to refine network weights and thresholds, thereby avoiding local optima and enhancing the robustness against asymmetric field patterns. The experimental validation under extreme mechanical deformations (X/Y-eccentricity: ±8 mm; Z-tilt: ±15°) demonstrates the strategy's efficacy, achieving a 65.07%, 45.74%, and 76.15% error suppression for X-, Y-, and Z-axis deviations. The elliptical configuration reduces the installation footprint by 72.4% compared with conventional circular sensor arrays while maintaining a robust suppression of eccentricity- and tilt-induced errors, proving critical for space-constrained applications, such as electric vehicle powertrains and miniaturized industrial inverters. This work bridges bio-inspired algorithms and adaptive sensing hardware, offering a systematic solution to mechanical deformation-induced errors in high-density power systems.
矩形导体中的精确电流传感面临着机械变形的挑战，包括偏心（X/Y轴偏移）和倾斜（Z轴倾斜），这些变形会扭曲磁场分布并导致测量误差。为了解决这个问题，我们提出了一种受生物启发的误差补偿策略，该策略将椭圆形配置的霍尔传感器阵列与混合灰狼优化器（GWO）增强的反向传播神经网络相结合。通过三维磁场重建和电流反演建模来量化导体的偏心位移和倾斜角度。实施了一个双阶段优化框架：首先，建立一个BP神经网络用于实时导体状态估计，其次，利用GWO的群体智能来优化网络权重和阈值，从而避免局部最优并增强对非对称场模式的鲁棒性。在极端机械变形（X/Y偏心：±8毫米；Z倾斜：±15°）下的实验验证证明了该策略的有效性，对于X、Y和Z轴偏差分别实现了65.07%、45.74%和76.15%的误差抑制。与传统圆形传感器阵列相比，椭圆形配置将安装占地面积减少了72.4%，同时保持对偏心和倾斜引起的误差的强大抑制能力，这对于空间受限的应用（如电动汽车动力系统和小型化工业逆变器）至关重要。这项工作将受生物启发的算法与自适应传感硬件联系起来，为高密度电力系统中机械变形引起的误差提供了一种系统解决方案。
