Knowledge discovery and performance-energy optimization in heterogeneous catalytic ozonation via adaptive multi-task learning

Wei Zhuang; Qianqian Luo; Qingyang Jiang; Songhang Du; Xinwei Li; Jun Ma; Xiao Zhao; Minghao Sui

doi:10.1016/j.watres.2025.124422

Knowledge discovery and performance-energy optimization in heterogeneous catalytic ozonation via adaptive multi-task learning

Water Res. 2025 Dec 1;287(Pt A):124422. doi: 10.1016/j.watres.2025.124422. Epub 2025 Aug 18.

Authors

Wei Zhuang¹, Qianqian Luo², Qingyang Jiang², Songhang Du³, Xinwei Li⁴, Jun Ma³, Xiao Zhao⁵, Minghao Sui⁶

Affiliations

¹ State Key Laboratory of Pollution Control and Resource Reuse, College of Environmental Science and Engineering, Tongji University, Shanghai 200092, PR China. Electronic address: trumnkyd@tongji.edu.cn.
² State Key Laboratory of Pollution Control and Resource Reuse, College of Environmental Science and Engineering, Tongji University, Shanghai 200092, PR China.
³ State Key Laboratory of Urban Water Resource and Environment, School of Environment, Harbin Institute of Technology, Harbin, 150090, PR China.
⁴ State Key Laboratory of Marine Geology, School of Ocean and Earth Science, Tongji University, Shanghai 200092, PR China.
⁵ Academy for Engineering and Technology, Fudan University, Shanghai 200000, PR China.
⁶ State Key Laboratory of Pollution Control and Resource Reuse, College of Environmental Science and Engineering, Tongji University, Shanghai 200092, PR China; Shanghai Institute of Pollution Control and Ecological Security, Shanghai 200092, PR China. Electronic address: minghaosui@tongji.edu.cn.

PMID: 40834732
DOI: 10.1016/j.watres.2025.124422

Abstract

Heterogeneous catalytic ozonation (HCO) is a promising technology with significant potential for water treatment, but its application is limited by the optimization of performance and energy consumption. This study proposed an innovative adaptive multi-task learning (MTL) framework for performance optimization (PO), energy consumption (EC), and performance-energy balance (PEB). The PEB-MTL constructed with implicit functions and Pareto optimization achieved a multi-task balance between pseudo-first-order rate constant (k) and Electrical Energy per Order (EE/O), and could simultaneously optimize performance and energy consumption through adaptive task weighting and random perturbation-enhanced PSO algorithm. Feature importance analysis revealed that O₃ dosage is the most critical factor. Catalyst composition plays a decisive role for the degradation performance of O₃-resistant pollutants. Moreover, it exhibited a stronger impact on energy consumption than on degradation performance. Among molecular descriptors, E_gap and ionization potential were foremost for PO and EC, respectively. Reverse experiments demonstrated prediction errors below 10 %. This work provides a robust framework for HCO optimization, advancing sustainable water treatment technologies.

Keywords: Energy consumption; Heterogeneous catalytic ozonation; Machine learning; Multi-task learning; Pareto optimization; Performance.

MeSH terms

Algorithms
Catalysis
Ozone* / chemistry
Water Purification* / methods

Substances

Ozone