Prediction Model of MBR Membrane Flux for Elman Neural Network Based on PSO-GA Hybrid Algorithm

Membrane fouling reduces wastewater treatment efficiency and cause financial and energy costs to some extent. The size of membrane flux reflects the degree of membrane pollution. Make timely cleaning membrane or replacement membrane decision to maintain considerable treatment effect on the basis of the membrane flux. Particle swarm optimization (PSO) algorithm can quickly find the global optimum. Genetic algorithm (GA) has the property of global convergence. The prediction model used in this paper is based on the PSO-GA hybrid algorithm. The combination of these not only improves the convergence speed of genetic algorithm but also reduces the probability of particle swarm optimization algorithm falling into local optimum. Elman neural network acts as the basic network. Compared with Elman neural network and BP neural network, the prediction accuracy of PSO-GA-Elman is improved.