Differential Evolution for Assisted History Matching Process: SAGD Case Study

Arash Mirzabozorg

Supervisor: Dr. Long Nghiem
Co-supervisor: Dr. Zhangxing(John) Chen


Differential evolution as a population-based optimization algorithm is an excellent candidate for tackling history matching problems. It is robust, easy to use and can easily be parallelized. It requires few control parameters. Comparison PSO and DE in this study showed that DE performed much better in finding multiple history matched models with a much lower number of simulation runs than PSO. Ensemble of multiple history matched models yields a more realistic uncertainty assessment in the forecast stage.