Improved Alopex-based evolutionary algorithm (AEA) by quadratic interpolation and its application to kinetic parameter estimations

Abstract

The Algorithm of Pattern Extraction (Alopex)-based evolutionary algorithm (AEA) is an intelligent optimization algorithm that combines the Alopex with the evolution algorithm. It has high exploratory ability due to its especial searching mechanism. However, the AEA is poor at exploitation to some extent. To improve the performance of AEA, quadratic interpolation (QI) method is introduced to the AEA in this paper. The improved algorithm chooses three individuals to fit a quadratic function to approximate the objective function in each iterative of the AEA and uses the extreme point of the quadratic function to generate new individuals. The QI method can help the algorithm to converge rapidly near optimal solutions, which can greatly improve the exploitation ability of AEA. The traditional and CEC 2013 benchmark functions are both used to test the performance of the improved algorithm (QIAEA); the experimental results show that the QIAEA has better convergent speed and higher accuracy than the AEA. The QIAEA is also compared with several state-of-the-art algorithms and the results show that the QIAEA is an effective and efficient algorithm. Furthermore, its applications to parameter estimation for a model of sulfur dioxide (SO2) oxidation with a cesium–rubidium–vanadium (Cs–Rb–V) sulfuric acid catalyst and a heavy oil thermal cracking three-lump model demonstrate that QIAEA can forecast more accurately than other methods tested.