A comparative study of GA, PSO and SCE algorithms for estimating kinetics of biomass pyrolysis

Abstract

Optimization performances of three most frequently utilized optimization algorithms, GA (Genetic Algorithm), PSO (Particle Swarm Optimization), and SCE (Shuffled Complex Evolution), are compared to examine their accuracy, computation efficiency, and convergence efficiency. Micro scale TGA (thermogravimetric analysis) experiments of wood were conducted at three heating rates to collect the necessary data for analysis. Gauss multi-peak fitting method was first applied to identify the contribution of each component of wood to the mass loss rate (MLR) curves. Then the Kissinger method and three isoconversional methods, including KAS, Tang, and DAEM methods, were employed to extract kinetics of wood pyrolysis. The average values of the four sets of solutions were adopted to determine the search range in the following optimizations. A thermally thin numerical model was developed to inversely model the collected experimental data combining the three algorithms. The results showed that wood pyrolysis can be described by a four-component parallel reaction scheme. The four sets of kinetic parameters derived using different analytical methods are very close to each other. When extracting kinetics from experimental data using numerical model and optimization algorithms, the accuracies of the three algorithms are ranked as SCE > PSO > GA. While the computation efficiencies and convergency efficiencies are ranked as GA ≈ PSO > SCE and PSO > SCE > GA, indicating each algorithm has its inherent advantages and limits. In most optimization applications, PSO is more favorable considering its better overall performance.

Keywords

• pyrolysis kinetics
• gauss multi-peak fitting method
• isoconversional method
• genetic algorithm (ga)
• particle swarm optimization (pso)
• shuffled complex evolution (sce)