Multi-electrode resistivity method is equipped with various electrode array configurations. Due to their different electrode arrangements, various electrode array configurations tend to exhibit significantly varying detection effects under different survey environments. Hence, to achieve satisfactory detection effects using high-resolution electrical resistivity tomography, it is necessary to explore the strategy for choosing appropriate array configurations targeting different objects in practical work. Given the high applicability of the method, its forward and inversion calculations remain a critical task. Based on the differential equations to be satisfied by the point source potential in a three-dimensional structure, this study derived the variational problem to be satisfied by the 2.5D potential and conducted the unstructured gridding using the Delaunay triangulation algorithm, thus achieving finite-element forward modeling. By combining practical applications, this study designed common geological models and performed forward and inverse calculations using Wenner
\alpha , Wenner
\beta , Schlumberger, and dipole-dipole arrays, analyzing their detection effectiveness in different environments. Key findings are as follows: (1) For the detection of isolated anomalous bodies in an unknown area, the Wenner
\beta and Schlumberger configurations, determined by considering the accuracy and efficiency, can yield better detection effects. (2) The Schlumberger and dipole-dipole configurations exhibit higher horizontal resolution and can distinguish multiple anomalous bodies nearby. (3) For the detection of low-resistivity fractured zones, the Wenner
\beta and dipole-dipole configurations enjoy better performance. (4) For strata with distinct boundaries, the Wenner
\alpha , Wenner
\beta , Schlumberger, and dipole-dipole configurations can all yield encouraging detection results. Therefore, for data collection using high-resolution electrical resistivity tomography, it is necessary to choose multiple array configurations and conduct comprehensive comparisons and interpretations of the forward and inverse modeling results.