Optimizing hydropower scheduling through accurate power load prediction: A practical case study.

Area of Science:

• * Engineering
• * Environmental Science
• * Operations Research

Background:

• * Grid-connected hydropower stations face challenges with uneven power generation and benefit distribution due to delayed load data.
• * Current scheduling models struggle with suboptimal dispatch when real-time load information is unavailable.

Purpose of the Study:

• * To develop a novel scheduling model for hydropower stations that combines power load prediction and dual-objective optimization.
• * To address the issue of delayed load data and improve dispatch efficiency for grid-connected hydropower operations.
• * To achieve a harmonious balance between hydropower generation and navigation requirements.

Main Methods:

• * Assessment of various power load prediction models, identifying the Convolutional Neural Network-Gated Recursive Unit (CNN-GRU) as the most accurate.
• * Integration of predicted power load data into an enhanced elite non-dominated sorting genetic algorithm (GA-NSGA-II).
• * Optimization of hydropower station discharge flow using proposed objective functions.

Main Results:

• * The CNN-GRU model achieved high prediction accuracy with R-squared of 0.991 and RMSE of 0.026.
• * Scheduling based on predicted load values showed minimal variance (within 5%) compared to actual load values, demonstrating practical effectiveness.
• * The optimized scheduling successfully balanced hydropower generation and ship navigation demands in a real-world case study.

Conclusions:

• * The developed model provides an efficient solution for hydropower station scheduling, even with incomplete or delayed load data.
• * The approach effectively addresses practical challenges in grid-connected hydropower operations.
• * Achieved dual benefits of optimized power generation and unimpeded navigation, showcasing the model's practical applicability.