Application Of Hybrid Optimization Algorithm Of Artificial Fish Swarm And Differential Evolution In The Water Quality Prediction

Water quality prediction plays an increasingly important role in the field of water resource utilization, water environment management and water scheduling. Although parameter estimation of mechanism water quality model is critical and important step to simulation and prediction of water quality, the parameter estimation is often difficult. Hybrid optimization algorithm of Artificial Fish Swarm and Differential Evolution is used for parameter estimation and high accuracy is achieved. As for the non-mechanism models, water quality "black box" model of LS-SVM optimized by hybrid optimization algorithm.of Artificial Fish Swarm and Differential Evolution is built and a smaller error is achieved in the water quality prediction.Main works and innovation points are listed as follows:(1) A new hybrid optimization algorithm, based on Artificial Fish Swarm Algorithm and Differential Evolutionare Algorithm, is proposed. Artificial Fish Swarm Algorithm is for the global optimization and Differential Evolution Algorithm is for the local optimization.(2) The new hybrid optimization algorithm of Artificial Fish Swarm and Differential Evolution is applied in parameter estimation of mechanism water quality model. The algorithm is used for parameter estimation of one-dimension water quality model of uniform river and Thomas BOD-DO model respectively. It has the advantage of simple calculation, high precision and good robustness. Test results show that the method is feasible and effective.(3) Least Squares Support Vector Machine (LS-SVM) is used for building the water quality "black box" model. At zhe same time, hybrid optimization algorithm of Artificial Fish Swarm and Differential Evolution is applied to parameters optimization of the LS-SVM model. With some water quality data, the optimized model is built and the result is acceptable. The root mean square error and mean relative error of the water quality model are smaller which shows that the optimized model is effective and practical.