Monthly Pan Evaporation Prediction using Hybrid Extreme Learning Machine with Bio-inspired Optimization Algorithms

Evaporation is a significant component in hydrological processes, and reliable prediction of evaporation is essential for water resources management, irrigation system design and hydrological modeling. This study explored the potential of hybrid extreme learning machine (ELM) model with two novel bio-inspired algorithms, i.e. whale optimization algorithm (WOA) and flower pollination algorithm (FPA) for prediction of monthly pan evaporation (E_p) in the Poyang Lake Basin of Southern China as a case study. The hybrid ELM-WOA and ELM-FPA models were also compared with commonly used artificial neural networks (ANN), M5 model tree (M5Tree) and differential evolution algorithm-optimized ELM (ELM-DEA) models. Monthly meteorological variables during 2001-2015, i.e. maximum and minimum temperature (T_max and T_min), sunshine duration (n), relative humidity (RH), wind speed (U₂) and E_p were collected from four weather stations in the basin, with data from 2001-2010 for model training and those during 2011-2015 for model testing. The obtained results showed that the hybrid ELM-FPA model exhibited the highest prediction accuracy at all stations, followed by the hybrid ELM-WOA model, both of which outperformed the ELM-DEA, ANN and M5Tree models. Compared with the M5Tree model, the prediction accuracy of the hybrid ELM-FPA and ELM-WOA model were increased by 15.9% and 15.2% respectively, based on root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The bio-inspired FPA and WOA algorithms are thus highly recommended for improving the performance of standalone machine learning models.