Answers are grounded in this paper and cite their sources.
The present research paper examines the capability of Long-Short Term Memory (LSTM) with Attention Mechanism (AM) (termed as LSAM) and Natural Gradient Boosting (NGB) in predicting monthly inflows of the Sri Ram Sagar Project, India. The parameters of these algorithms are optimized by Particle Swarm Optimization (PSO) and Grid Search (GS), creating four combinations: LSAM-PSO, NGB-PSO, LSAM-GS, and NGB-GS. Three input scenarios, S1, S2, and S3, are considered based on monthly rainfall, maximum temperature, minimum temperature, and inflow. A dataset comprising 42 years (1983–2024) is utilized for training and testing. The Root Mean Square Error, Nash-Sutcliffe Efficiency, Percent Bias, and Willmott Index (RMSE, NSE, PBIAS, WI) evaluated the performance of the algorithms. LSAM-PSO achieved its best performance in S3 with RMSE = 16.8333 Million Cubic Metre (MCM), NSE = 0.9892, PBIAS = −0.6877, WI = 0.9816 in training and RMSE = 36.0273 (MCM), NSE = 0.8462, PBIAS = −2.4603, WI = 0.8951 in testing. Accordingly, it was utilized to predict reservoir inflows from a climate change perspective for Short-Term, Mid-Term, and Long-Term for four Socioeconomic Pathways (SSPs). The predictions indicated a substantial rise in the Long-Term inflows, specifically under SSP 585, with annual inflows increasing up to 247.28% compared to the baseline averages. The frequency of peak inflow events remains higher than the baseline for all SSPs across time segments except SSP 585.
No takes yet. Share an insight, caveat, or question.
Deb et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: