Recovering 3D Basin Basement Relief Using High-Precision Magnetic Data Through Particle Swarm Optimization and Back Propagation Algorithm

The Inversion of magnetic basement interfaces in basins is critical for interpreting potential field data and studying geothermal resource distribution as well as basin formation and evolution. This paper proposes a new method for the inversion of magnetic basement interfaces using a particle swarm optimization algorithm that combines potential field processing and machine learning techniques. This method generates magnetic base interface models and the corresponding magnetic anomaly data through the random midpoint displacement method and magnetic interface finite element forward simulation. These anomalies are then handled using techniques, such as directional transformations, analytical continuation, spatial derivatives, and fractional transformations. Feature attributes were extracted, and the Gini importance was used to quantify feature factor contributions, screen out effective features, and improve algorithm efficiency. Validity and practicality were verified through an analysis of the theoretical and noise models. The proposed machine learning-based method is more intelligent, efficient, and accurately reflects the undulations of magnetic-based interfaces. Application to magnetic survey data in the Datong Basin resulted in a reliable basin-based model that matched known structural information, thereby opening a new direction for magnetic interface inversion research.