Fine-tuning inflow prediction models: integrating optimization algorithms and TRMM data for enhanced accuracy

Abstract

This research explores machine learning algorithms for reservoir inflow prediction, including long short-term memory (LSTM), random forest (RF), and metaheuristic-optimized models. The impact of feature engineering techniques such as discrete wavelet transform (DWT) and XGBoost feature selection is investigated. LSTM shows promise, with LSTM-XGBoost exhibiting strong generalization from 179.81 m3/s RMSE (root mean square error) in training to 49.42 m3/s in testing. The RF-XGBoost and models incorporating DWT, like LSTM-DWT and RF-DWT, also perform well, underscoring the significance of feature engineering. Comparisons illustrate enhancements with DWT: LSTM and RF reduce training and testing RMSE substantially when using DWT. Metaheuristic models like MLP-ABC and LSSVR-PSO benefit from DWT as well, with the LSSVR-PSO-DWT model demonstrating excellent predictive accuracy, showing 133.97 m3/s RMSE in training and 47.08 m3/s RMSE in testing. This model synergistically combines LSSVR, PSO, and DWT, emerging as the top performers by effectively capturing intricate reservoir inflow patterns. HIGHLIGHTS Proposed models can provide accurate predictions for daily inflow, rainfall, and inflow rates in similar regions and conditions.; Flood prediction using machine learning algorithms for Três Marias Reservoir (TMR) in Brazil.; Machine learning models combined with DWT can accurately predict daily inflows to the TMR in Brazil.;

Keywords

• data-driven frameworks
• discrete wavelet transform
• inflow prediction
• parameter optimization
• particle swarm optimization
• reservoir management