Calibration of Xinanjiang model parameters using hybrid genetic algorithm based fuzzy optimal model

Wen Chuan Wang, Dong Mei Xu, Chun Tian Cheng, Kwok Wing Chau

Research output: Journal article publicationJournal articleAcademic researchpeer-review

46 Citations (Scopus)


Conceptual rainfall-runoff (CRR) model calibration is a global optimization problem with the main objective to find a set of optimal model parameter values that attain a best fit between observed and simulated flow. In this paper, a novel hybrid genetic algorithm (GA), which combines chaos and simulated annealing (SA) method, is proposed to exploit their advantages in a collaborative manner. It takes advantage of the ergodic and stochastic properties of chaotic variables, the global search capability of GA and the local optimal search capability of SA method. First, the single criterion of the mode calibration is employed to compare the performance of the evolutionary process of iteration with GA and chaos genetic algorithm (CGA). Then, the novel method together with fuzzy optimal model (FOM) is investigated for solving the multi-objective Xinanjiang model parameters calibration. Thirty-six historical floods with one-hour routing period for 5 years (2000-2004) in Shuangpai reservoir are employed to calibrate the model parameters whilst 12 floods in two recent years (2005-2006) are utilized to verify these parameters. The performance of the presented algorithm is compared with GA and CGA. The results show that the proposed hybrid algorithm performs better than GA and CGA.
Original languageEnglish
Pages (from-to)784-799
Number of pages16
JournalJournal of Hydroinformatics
Issue number3
Publication statusPublished - 20 Aug 2012


  • Chaos genetic algorithm
  • Flood forecasting
  • Fuzzy multi-objective optimization
  • Simulated annealing
  • Xinanjiang model calibration

ASJC Scopus subject areas

  • Geotechnical Engineering and Engineering Geology
  • Atmospheric Science


Dive into the research topics of 'Calibration of Xinanjiang model parameters using hybrid genetic algorithm based fuzzy optimal model'. Together they form a unique fingerprint.

Cite this