本项目通过随机统计和人工智能等多学科理论和方法的交叉集成，对地下水污染溯源辨识研究前沿中尚待解决的瓶颈问题开展研究。探索应用深度卷积神经网络方法构建地下水污染多相流数值模拟模型的替代模型。分析应用无迹卡尔曼滤波方法辨识地下水污染源的个数。研究应用贝叶斯公式和替代模型，建立搜索点的状态评估函数。应用启发式搜索策略，设计构造启发式搜索迭代过程。探索应用多向量变步长随机游走方法，确定每轮迭代的尝试点。研究构建基于状态评估函数的Metroplis公式，将其计算结果作为状态转移的判别依据。探索在启发式搜索迭代过程中嵌入反馈修正环节，以实现待求变量与其概率分布的互馈更新。结合研究实例，对上述理论和方法进行分析与验证。提高地下水污染源的辨识精度，为地下水污染修复方案的合理设计提供必要前提条件。拓展地下水污染溯源辨识的理论基础，并为修复工程的实际应用提供技术支持。

Some unsolved bottleneck problems in forefront research on inversion identification of groundwater contamination sources will be studied in this project through the cross-integrated multi-disciplinary theories and methods such as stochastic statistics and artificial intelligence. The Deep Convolutional Neural Network (DCNN) method will be utilized to build the surrogate model of the groundwater contamination multiphase flow numerical simulation model. The Unscented Kalman Filter (UKF) method will be adopted to identify the number of groundwater contamination sources. The Bayesian formula and surrogate model will be applied to establish the state evaluation function of the search point. Then, a heuristic search iterative process will be designed using heuristic search strategy. A Multi-vector and Variable-step size Random Walk (MVRW) method will be adopted to determine the candidate point. A Metroplis formula based on the state evaluation function will be constructed, and the calculated result will be used as the judging basis of the state transition. In the end, a feedback correction step will be inserted into the heuristic search iterative process to realize the feedback update of the unknown variables and their probability distribution. The above theories and methods will be analyzed and verified by case study. The identification accuracy of groundwater contamination sources will be improved and it will supply necessary precondition for reasonable design of groundwater contamination remediation plan. The theoretical basis of groundwater contamination identification will be expanded and provide technical support for practical application of remediation project.