Shale wastewater treatment policies recommended by integrated knowledge graph and probability-based algorithms

Area of Science:

• Environmental Science
• Artificial Intelligence
• Chemical Engineering

Background:

• Shale wastewater (SWW) poses significant environmental and health risks.
• Selecting appropriate SWW treatment technologies is challenging without expert knowledge or data.
• Existing methods often rely on complex physical models and extensive quantitative data.

Purpose of the Study:

• To develop a data-driven modeling framework for recommending shale wastewater treatment policies.
• To integrate knowledge graphs (KG) with probability-based algorithms for policy generation.
• To assess the stability and robustness of the recommended policies using the Monte-Carlo (MC) technique.

Main Methods:

• Developed a modeling framework combining Knowledge Graph Convolutional Networks (KGCN) and RippleNet with KG.
• Applied the model to shale regions in China, including Chongqing, Sichuan, Yunnan, Guizhou, Shaanxi, and Inner Mongolia.
• Utilized the Monte-Carlo (MC) technique to evaluate the stability of the recommended treatment policies.

Main Results:

• The framework successfully recommended specific SWW treatment technologies, including chemical precipitation, ultrafiltration, membrane bioreactor, electrodialysis, electrocoagulation, reverse osmosis, and electrocatalytic oxidation.
• Recommended policies demonstrated high stability, with MC-based probabilities exceeding 0.86.
• The model proved robust, offering reliable policy recommendations without conventional physically based models.

Conclusions:

• The integrated KG and AI framework offers a viable alternative for developing shale wastewater treatment strategies.
• The model's robustness and stability are confirmed by MC simulations.
• Future work will focus on enhancing the KG, algorithms, and providing policy explanation mechanisms.