Abstract:Due to the impact of excessive influent water quality and water quantity fluctuations, current industrial effluent treatment plants (IETPs) cannot maintain stable decontamination capabilities, posing serious threats to the ecological environment. Therefore, predicting the influent quality of IETPs to facilitate predictive maintenance is imperative. However, because of the highly non-linear nature and variability of industrial wastewater, single models often have limited prediction accuracy. Therefore, this study innovatively constructs a comprehensive and highly adaptable hybrid deep learning prediction framework called GCN optimised CNN-GRU-attention model (GCN-CNNGA). By integrating a graph convolutional network (GCN), convolutional neural network (CNN), gated recurrent unit (GRU), and an attention mechanism, the proposed framework effectively captures the high-dimensional features of influent data and complex non-linear spatiotemporal patterns among variables. Specifically, a CNN is used to accurately extract key local features from the IETP time series. Building on this, a GCN is utilised to further capture the topological information among key influent parameters, revealing their interconnections and deeply mining the complex dependencies among pollutants. The GRU effectively captures long-term dependencies not considered in the above processes, whereas the attention mechanism optimises the weight distribution, thereby enhancing the precision of predictions. During testing using real data from an IETP in an industrial park in East China, the GCN-CNNGA model outperformed the mainstream prediction methods in terms of prediction accuracy and robustness. This study introduces a high-accuracy prediction model that is expected to significantly improve the predictive maintenance levels of IETPs, provide strong technical support for maintaining optimal operational conditions, and have significant implications regarding protecting the ecological environment and promoting sustainable development.

A hybrid spatial-temporal deep learning prediction model of industrial methanol-to-olefins process

Spatial-temporal Self-attention Network Based on Bayesian Optimization for Light Olefins Yields Prediction in Methanol-to-olefins Process

Spatio-temporal attention-based hybrid deep network for time series prediction of industrial process

Industrial Soft Sensor Prediction Based on Multi-model Integrated Method

Hybrid Modeling of Methanol to Olefin Reaction Kinetics Based on the Artificial Neural Network

Multivariate Deep Reconstruction Neural Network for Multi-step-ahead Prediction of Industrial Process Quality Variables

Novel Hybrid Data-Driven Modeling Integrating Variational Modal Decomposition and Dual-Stage Self-Attention Model: Applied to Industrial Petrochemical Process

Multi-modal Hybrid Modeling Strategy Based on Gaussian Mixture Variational Autoencoder and Spatial–temporal Attention: Application to Industrial Process Prediction

Forecasting Methanol-to-olefins Product Yields Based on Relevance Vector Machine with Hybrid Kernel and Rolling-windows

A Novel Methanol Futures Price Prediction Method Based on Multicycle CNN-GRU and Attention Mechanism

Intelligent Hybrid Modeling of Complex Leaching System Based on LSTM Neural Network

Learning Deep Multimanifold Structure Feature Representation for Quality Prediction with an Industrial Application

A Deep Learning Hybrid Framework Combining an Efficient Evolutionary Algorithm for Complex Many-Objective Optimization of Sustainable Triple CO2 Feed Methanol Production

Prediction of Steelmaking Process Variables Using K-medoids and a Time-Aware LSTM Network

A Hybrid Oil Production Prediction Model Based on Artificial Intelligence Technology

A Hybrid Deep Learning Framework for Predicting Industrial Wastewater Influent Quality Based on Graph Optimisation

Online Support Vector Machine Based on Mixed Kernel Function and Its Application in Methanol Aynthesis

Efficient Prediction Framework for Large-Scale Nonlinear Petrochemical Process Based on Feature Selection and Temporal-Attention LSTM: Applied to Fluid Catalytic Cracking

A Robust Hybrid Predictive Model of Mixed Oil Length with Deep Integration of Mechanism and Data

A Hybrid Deep Learning Framework Driven by Data and Reaction Mechanism for Predicting Sustainable Glycolic Acid Production Performance

Interpretable Machine Learning for Predicting Biomethane Production in Industrial-Scale Anaerobic Co-Digestion