Boosting hydropower output of mega cascade reservoirs using an evolutionary algorithm with successive approximation

Abstract

The high complexity of multi-objective joint reservoir operation imposes challenging barriers to the pursuit of optimal hydroelectricity output. Inasmuch as multi-objective evolution optimization algorithms, including the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II), are trapped in the curse of dimensionality which could not be effectively solved the multi-objective operation of more than ten reservoirs. This study proposes a methodology that integrate the NSGA-II with a successive approximation approach to optimize the hydropower output for conquering the curse of dimensionality under the joint operation of 21 mega cascade reservoirs located in the Upper Yangtze River Basin of China. The successive approximation approach could effectively decompose the mutually related M-dimensional problem into M individual one-dimensional problems, which ingeniously overcomes the curse of dimensionality. The proposed model is anchored with strategies of advancing impoundment timings and raising water levels of cascade reservoirs. We show that our methodology, without adding or upgrading hydraulic infrastructures, empowers the joint operation to reach 110.79 billion kW·h/year (9.8% improvement) in hydropower output, which could reduce 86.97 billion kg/year in CO2 emission, and to provide 44.97 billion m3/year in water supply with flood risk less than 0.016. The results suggest that our methodology can spur hydroelectricity output to support China’s tactics in fulfilling the pledge of carbon emission reduction and non-fossil energy expansion to 20% by 2030.