Database

To address the significant increase in petroleum-induced soil contamination caused by frequent leaks in buried oil pipelines, a prediction method for oil content in contaminated soil based on Particle Swarm Optimization (PSO)-Backpropagation (BP) neural networks has been proposed. This prediction model has been integrated into the pipeline leak monitoring system independently developed by the research team. It enables real-time online monitoring of petroleum-contaminated soil conditions and quantifies pollutants. Based on hundreds of experimental data sets studying the influence characteristics of resistivity on contaminated soil, moisture content, dry density, resistivity, and temperature were selected as the modeling dataset. Prediction models for oil content in contaminated soil were established using Standard Backpropagation (BP) Neural Network, Genetic Algorithm (GA) Optimized BP Neural Network, Particle Swarm Optimization (PSO) Optimized BP Neural Network, Random Forest Neural Network. The results indicate that the particle swarm optimization (PSO)-enhanced backpropagation neural network prediction model delivers the most accurate performance, achieving a coefficient of determination (R²) of 0.901. This model can reliably predict the oil content in contaminated soil, providing an effective solution for precise future predictions of oil content in contaminated soil.