Optimal Monthly Reservoir Operation Rules for Hydropower Generation Derived with SVR-NSGAII

Abstract

A novel tool is proposed that couples the nondominated sorting genetic algorithm (NSGAII) with support vector regression (SVR) and nonlinear programming (NLP) to optimize monthly operation rules for hydropower generation. The SVR-NSGAII is applied to calculate the optimized release for hydropower generation by minimizing (1) the error committed by the SVR in extracting the optimized operation rule, and (2) the number of input variables used as predictors (the parsimony feature) in a regression model. The SVR calculates the optimized reservoir release for hydropower generation based on input variables and parameters values that are found by the NSGAII. An evaluation of results obtained for the Karoon-4 reservoir of Iran indicates that the SVR-NSGAII is well suited to calculate the optimal hydropower reservoir operation rule in real time with approximately 90% accuracy.