为解决目前隧道超前预报瞬变电磁常规反演方法对掌子面前方含水体定位存在偏差的问题，提出基于全空间模型正演算法的隧道瞬变电磁粒子群非线性反演方法，对掌子面前方含水体进行预报。建立全空间低阻夹层模型和掌子面前方含低阻体模型进行数值模拟，首先，采用全空间正演算法计算得到正演数据，将正演数据添加高斯噪声后作为观测数据； 然后，采用粒子群反演算法对观测数据进行迭代反演，通过粒子群的更新—评价—择优过程不断调整模型，直到满足拟合误差阈值得到最优解； 最后，将粒子群反演方法应用于工程实例，并与常规烟圈反演结果进行对比分析，验证所提方法预报隧道含水体的有效性。数值模拟结果表明： 1）反演结果的电性特征和层厚与理论模型基本一致，反演方法对低阻夹层界面和低阻体的位置刻画比较准确，对低阻体具有较高的定位精度； 2）算法的计算速度在迭代后期随着高斯噪声的增大而降低，但仍存在下降收敛的趋势，具有较高的全局寻优能力。工程应用结果表明，粒子群反演结果相比常规烟圈反演结果，其低阻异常范围变小、异常轮廓更为清晰、浅部的细节更加丰富、成像精度相对更高，隧道开挖结果验证了粒子群反演方法能准确定位含水体的空间位置。

To minimize or eliminate the detection deviation in positioning water-bearing bodies in front of the tunnel face, which often occurs with conventional transient electromagnetic inversion methods used in tunnel advance prediction, the authors propose a nonlinear particle swarm inversion method for tunnel transient electromagnetics. The proposed method is based on a full-space forward algorithm for predicting water-bearing bodies in front of tunnel faces. For the numerical simulations, a full-space low-resistivity interlayer model and a low-resistivity body model in front of the tunnel face are established. The full-space forward algorithm is used to calculate the forward data, and Gaussian noise is added, which serves as the observation data. The particle swarm inversion algorithm is used to iteratively invert the observation data. The model is continuously adjusted through an update-evaluation-optimization process within the particle swarm until the fitting error threshold is satisfied, thereby yielding the optimal solution. The particle swarm inversion method is applied to an engineering project and compared with conventional smoke ring inversion results to verify its validity in predicting tunnel water-bearing bodies. The numerical simulation results demonstrate the following: (1) The electrical characteristics and layer thickness of the inversion results are consistent with the theoretical model. The method accurately depicts the position of the low-resistance interlayer interface and the low-resistance body, exhibiting higher positioning accuracy for the low-resistance body. (2) The calculation speed of the algorithm decreases with increasing Gaussian noise in later iterations, but it still exhibits a downward convergence trend, indicating high global optimization ability. Furthermore, the engineering application results demonstrate that compared with the conventional smoke ring inversion results, the low-resistance anomaly range of the particle swarm inversion is smaller, the anomaly contour is clearer, shallow details are more abundant, and the imaging accuracy is relatively higher. The tunnel excavation results verify that the particle swarm inversion method can accurately position the spatial location of the water-bearing body.