Multi-objective optimization of empirical hydrological model for streamflow prediction

- The MODE-ACM algorithm is firstly introduced for parameter estimation of hydrologic model.
- We propose an enhanced Pareto multi-objective differential algorithm EPMODE.
- The performance of EPMODE and MODE-ACM is tested on five benchmark problems.
- EPMODE and MODE-ACM are further applied to an artificial neural network model for monthly streamflow forecasting.
- The results of EPMODE and MODE-ACM are compared with those of NSGA-II and SPEA2.

SummaryTraditional calibration of hydrological models is performed with a single objective function. Practical experience with the calibration of hydrologic models reveals that single objective functions are often inadequate to properly measure all of the characteristics of the hydrologic system. To circumvent this problem, in recent years, a lot of studies have looked into the automatic calibration of hydrological models with multi-objective functions. In this paper, the multi-objective evolution algorithm MODE-ACM is introduced to solve the multi-objective optimization of hydrologic models. Moreover, to improve the performance of the MODE-ACM, an Enhanced Pareto Multi-Objective Differential Evolution algorithm named EPMODE is proposed in this research. The efficacy of the MODE-ACM and EPMODE are compared with two state-of-the-art algorithms NSGA-II and SPEA2 on two case studies. Five test problems are used as the first case study to generate the true Pareto front. Then this approach is tested on a typical empirical hydrological model for monthly streamflow forecasting. The results of these case studies show that the EPMODE, as well as MODE-ACM, is effective in solving multi-objective problems and has great potential as an efficient and reliable algorithm for water resources applications.