A Novel Fast Optimisation Algorithm Using Differential Evolution Algorithm Optimisation and Meta-Modelling Approach
Genetic algorithms (GAs), Particle Swarm Optimisation (PSO) and Differential Evolution (DE) have proven to be successful in engineering model calibration problems. In real-world model calibration problems, each model evaluation usually requires a large amount of computation time. The optimisation process usually needs to run the numerical model and evaluate the objective function thousands of times before converging to global optima. In this study, a computational framework, known as DE-RF, is presented for solving computationally expensive calibration problems. We have proposed a dynamic meta-modelling approach, in which Random Forest (RF) regression model was embedded into a differential evolution optimisation framework to replace time consuming functions or models. We describe the performance of DE and DE-RF when applied to a hard optimisation function and a rainfall-runoff model calibration problem. The simulation results suggest that the proposed optimisation framework is able to achieve good solutions as well as provide considerable savings of the function calls with a very small number of actual evaluations, compared to these traditional optimisation algorithms.
KeywordsDifferential evolution optimisation Meta-modelling Random forest regression Automatic model calibration
The research reported in this paper was conducted as part of the “Developing a Real Time Abstraction & Discharge Permitting Process for Catchment Regulation and Optimised Water Management” project funded by EPSRC (Engineering Physical Sciences Research Council) and TSB (Technology Strategy Board) in the UK as part of the Water Security managed programme (TSB/EPSRC grant reference number: TS/K002805/1). This financial support is gratefully acknowledged.
- 1.Holland, H.J.: Adaptation in Natural and Artificial Systems, An Introductory Analysis with Application to Biology, Control and Artificial Intelligence. The University of Michigan Press, Ann Arbor (1975)Google Scholar
- 2.Storn, R., Price K.: Differential evolution: a simple and efficient adaptive scheme for global optimisation over continuous spaces. Technical Report TR-95-012, International Computer Science Institute, Berkley (1995)Google Scholar
- 3.Kennedy, J., Eberhart, R.: Particle Swarm Optimisation. In: Proceedings of the IEEE International Conference on Neural Networks, pp. 1942–1945 (1995)Google Scholar
- 10.Yan, S., Minsker, B.S.: A dynamic meta-model approach to genetic algorithm solution of a risk-based groundwater remediation design model. In: American Society of Civil Engineers (ASCE) Environmental & Water Resources Institute (EWRI) World Water & Environmental Resources Congress 2003 & Related Symposia, Philadelphia, PA (2003)Google Scholar
- 11.Jin, Y., Olhofer, M., Sendhoff, B.: On evolutionary optimisation with approximate fitness functions. In: Proceedings of the Genetic and Evolutionary Computation Conference (2000)Google Scholar
- 16.Bishop, C.M.: Neural Networks for Pattern Recognition. Oxford University Press (1995)Google Scholar
- 19.Carolin, S., Malley, J., Tutz, G.: Supplement to an introduction to recursive partitioning: rational, application, and characteristics of classification and regression trees, bagging, and random forests (2010). doi: 10.1037/a0016973.supp. Accessed 11 Nov 2010
- 20.Sudhanshu, M.: Some new test functions for global optimization and performance of repulsive particle swarm method. http://www.mpra.ub.uni-muenchen.de/2718/. Accessed 1 Aug 2009