Prediction Of Some Meteorological Elements Based On Improved Genetic Algorithm

Choose BP network to predict that having the parallel processing capabilities, based on the complex characteristics of meteorological elements, In order to void setting sensitive network and network generalization, some effective improvement strategies are put forward, which can optimized BP neural network model in this paper. These models have satisfied results that are applied to the predict temperature, seal level pressure, precipitation and so on. Specifically, the work of this paper in the following aspects:(1)Through analyzing the algorithm principle of BP neural network, in order to void setting sensitive network, niche genetic algorithm is introduced in this paper. As the classic niche genetic algorithm belongs to the scope of genetic algorithm, which has the disadvantage of slow convergence, the niche genetic algorithm based on the average method is proposed in this paper. Select complex functions to verify the correctness, reliability and superiority of the improved niche genetic algorithm.(2) In order to void genetic algorithm easy to fall the local minimum, quantum genetic algorithm is introduced in this paper. As using quantum bit coding, the population has diversity comparison and the performance of algorithm is higher than the classical genetic algorithm. To enhance the superiority of the algorithm, a quantum genetic algorithm based on adjusting rotation angle adaptively is proposed in this paper. Select complex binary function to verity the improved quantum genetic algorithm, the simulation shows that the improved quantum genetic algorithm has higher precision, smaller error, and can search the global optimal solution quickly.(3) Use these two improved genetic algorithm to optimize BP neural network weight threshold and create two BP network models to predict meteorological elements. Use the two optimization models to forecast the temperature, seal level pressure, precipitation of Nanjing station, and compare with standard niche GA-BP model, BP model. The simulation results show that the two optimization models have higher prediction accuracy and stability of predictive ability.