Abstract:Water quality prediction plays a crucial role in pollution treatment. However, inaccurate long-term prediction resulting from complex information patterns and insufficient feature extraction may lead to unnecessary environmental costs. In this study, a novel Interval Decomposition Correlation Particle Swarm Optimization-Extreme Learning Machine (IDCPSO-ELM) model is proposed to improve long-term water quality prediction ability. We employ Multivariate Variational Mode Decomposition (MVMD), Variational Mode Decomposition (VMD), Sliding Correlation and Permutation Entropy to reconstruct features to reduce complexity. Seasonal and Trend decomposition using Loess-Empirical Wavelet Transform decomposition (STL-EWT) is applied to target variables to improve feature extraction ability. Particle Swarm Optimization-Extreme Learning Machine (PSO-ELM) is used to predict high-frequency components based on the reconstructive features, and Back Propagation Neural Network-Extreme Learning Machine (BPNN-ELM) is employed to predict the trend and lowest-frequency components. Grey Wolf Optimization Algorithm (GWO) is used to combine these results. Six stations with high potential pollution threats in the Haihe River basin of Beijing from 2021 to 2022 are taken into model competition. The results show that: (1) In short-term prediction, average MAPE, RMSE and NSE of IDCPSO-ELM model reach 0.0934, 0.0410 and 0.8061, respectively, which are better than Genetic Algorithm-Elman Network (GA-ELMAN), Cuckoo Search Algorithm-Back Propagation Neural Network (CS-BPNN), PSO-ELM and Sparrow Search Algorithm-Kernel Extreme Learning Machine (SSA-KELM) short-term prediction models. (2) In long-term prediction, average MAPE, RMSE and NSE of IDCPSO-ELM model reach 0.1114, 0.0420 and 0.8268, respectively, which are better than the Radial Basis Function Neural Network (RBFNN), Grey Wolf Optimization-Support Vector Machine Regression (GWO-SVR), Whale Optimization Algorithm-Deep Extreme Learning Machine (WOA-DELM), Wavelet Transform decomposition-GWO-SVR-BPNN (WT-GWO-SVR-BPNN) and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise-PSO-ELM-Long Short-Term Memory (CEEMDAN-PSO-ELM-LSTM) long-term prediction models. IDCPSO-ELM model is considered competitive and promising for long-term water quality prediction, particularly in areas with high potential pollution threats.

Forecasting Daily Runoff by Extreme Learning Machine Based on Quantum-Behaved Particle Swarm Optimization

Flood forecasting methods with precipitation prediction and multi-objective parameter optimization

An enhanced monthly runoff time series prediction using extreme learning machine optimized by salp swarm algorithm based on time varying filtering based empirical mode decomposition

Runoff Prediction and Analysis Based on Improved CEEMDAN-OS-QR-ELM

Evolutionary artificial intelligence model via cooperation search algorithm and extreme learning machine for multiple scales nonstationary hydrological time series prediction

A novel interval decomposition correlation particle swarm optimization-extreme learning machine model for short-term and long-term water quality prediction

Improving streamflow prediction using a new hybrid ELM model combined with hybrid particle swarm optimization and grey wolf optimization

Short-term streamflow time series prediction model by machine learning tool based on data preprocessing technique and swarm intelligence algorithm

Research on particle swarm optimization in LSTM neural networks for rainfall-runoff simulation

Prediction of soil humidity based on random weight Particle Swarm Optimized Extreme Learning Machine

A runoff prediction method based on hyperparameter optimisation of a kernel extreme learning machine with multi-step decomposition

An adaptive daily runoff forecast model using VMD-LSTM-PSO hybrid approach

Prediction Based on Online Extreme Learning Machine in WWTP Application

A new hybrid model for monthly runoff prediction using ELMAN neural network based on decomposition-integration structure with local error correction method

Research on Precipitation Prediction Model Based on Extreme Learning Machine Ensemble

Monthly and seasonal hydrological drought forecasting using multiple extreme learning machine models

Parallel cooperation search algorithm and artificial intelligence method for streamflow time series forecasting

Optimizing Extreme Learning Machine for Drought Forecasting: Water Cycle vs. Bacterial Foraging

Improving multi-month hydrological drought forecasting in a tropical region using hybridized extreme learning machine model with Beluga Whale Optimization algorithm

A Hybrid Model of Ensemble Empirical Mode Decomposition and Sparrow Search Algorithm-Based Long Short-Term Memory Neural Networks for Monthly Runoff Forecasting

Hydrological time series forecasting via signal decomposition and twin support vector machine using cooperation search algorithm for parameter identification