Abstract: The multi-energy complementary system of wind, solar and hydropower has altered the operating conditions of hydropower stations, making the traditional hydropower dispatch rules no longer applicable. We develop a decision tree scheduling rule extraction method based on automatic hyperparameter optimization to examine such a system in the upper reaches of the Yellow River. It extracts a set of scheduling rules for multi-energy complementary systems, while considering new energy consumption and river ecological protection. The results show a competitive relationship exists between the power generation of this system and its river ecological protection goal. And the reservoir inflow is the most important predictor in consideration of its comprehensive power generation benefits. In all three accuracy evaluations, the best is the decision tree scheduling rule extraction model based on automatic hyperparameter optimization, indicating it achieves a significant improvement in accuracy. The research findings help the operation and scheduling of multi-energy complementary systems aimed at river ecological protection and power generation.

Key words: multi-energy complementarity, scheduling rules, new energy consumption, NSGA-III multi-objective optimization algorithm, decision tree