Research Of The Intelligence Method Of Atmosphere Quality Evaluation And Prediction

The World Climate Conference in Cancun are followed, Continued after the climate summit in Copenhagen. The climate issue has increasingly become the object of attention, including air quality assessment and prediction of atmospheric science is a hot issue, and is now the most intelligent method is efficient, the most widely used analytical techniques has important theoretical value and Practical significance. Methods of computational intelligence, intelligent thought from the bionics. Inspired by the rise and development of bionics and driven with intelligent algorithms to flourish. Intelligent technology is modeled to be solved, through a specific mathematical model to describe the problem itself, in order to achieve the purpose of solving the problem. Now, the air quality assessment and forecasting of intelligent application of the method is still in the initial stage, the more widely used of the neural network. In this context, of Changchun city with air quality testing centers for major air pollutants measured concentration data as a data sample of application of intelligent methods of evaluation and forecasting of air do research.Based on the existing number of intelligent methods, such as particle swarm optimization, neural network, fuzzy system theory, the atmosphere and atmospheric prediction model evaluation study of the theory, references and citations in a large number of literature based on research and Improved algorithm based on particle swarm optimization and fuzzy neural network (WFNN) used in air quality forecasting, through experiment to obtain a better prediction.Firstly, given the PSO algorithm has a nonlinear global optimization ability of the relevant papers at home and abroad through research and experimental comparison of advantages and disadvantages of various improved algorithms. Since the decision of PSO only three factors, namely, inertia weight wt, acceleration factor ac1, ac2. wt for the particle size of their inheritance, ac1 their best to follow the particle acceleration, ac2 best individual for the particles to follow the trend of population. Further analysis of the various factors that influence the particle swarm algorithm, the final analysis the swarm intelligence particle swarm optimization and characteristics of the field of application, in particular, the superiority of the global optimization.Secondly, contrast the traditional model of air quality evaluation of common and mainstream Evaluation of API. our current evaluation method using the API, the method for different kinds of pollutants has a different formula, the formula has characteristics associated with the contaminants Parameters, arithmetic complexity of trouble. In this paper, large number of experiments the parameters of the traditional method of the situation of individual pollutants, through the comparative analysis of the experimental results, researchers found an alternative way to improve the evaluation parameters of the dependence of pollutant characteristics.Again, an improved particle swarm optimization MOVPSO, improved particle swarm optimization process in the selection of the best programs, and through the geometric approach to visual proof of its correctness. Algorithm not only improves the selection of the best strategy, but also through the adaptive inertia weight Wt, to ensure no premature convergence of PSO and to avoid trapped in local optimum. In this paper, comparative analysis of the experiment with the traditional PSO algorithm in global optimization performance. On air quality assessment and prediction of the application requirements and the characteristics of the experimental data analysis, research evaluation of existing models of atmospheric and improved MOVPSO algorithm based on the formula selected to optimize the parameters. In the optimization process the parameters of the formula factors affecting a lot of research demonstration, select the appropriate parameters, such as the biggest change in speed, acceleration factor and so on. According to the results the most suitable parameters selected to obtain for the universal formula for all pollutants model can be applied to a variety of pollutants in air quality assessment work, without considering the relevant parameters and pollutants. In this paper, environmental quality monitoring stations used in Changchun City, the measured data available for evaluating the samples, using MOVPSO air quality assessment model to evaluate the quality of the atmosphere, the results obtained compared with the actual pollution model to verify the feasibility and practicality.Finally, in-depth study of atmospheric prediction model and fuzzy neural network development process, given the neural network has a wide range of adaptability, self-learning, as well as the rules of fuzzy inference system has the ability to in-depth analysis of the fuzzy neural network applied to the atmosphere Predict the feasibility of, and through experiment to verify. This paper uses neural network based on fuzzy weighted (WFNN) prediction model, with its main pollutants in the atmosphere to predict. Fuzzy neural network is constructed WFNN good structure, the use of genetic algorithms to the fuzzy neural network learning and training, when the network error and stability to meet the request as soon as the atmosphere used for forecasting, this paper used by Changchun Municipal Environmental Monitoring Center Station Measured data on the fuzzy neural network training and prediction. Prediction results were compared with the actual situation, from the result than the standard model prediction accuracy, with a considerable prospect.In summary, this study of particle swarm optimization, and improve and propose a new MOVPSO algorithm, the algorithm is applied to atmospheric pollution damage rate equation parameters optimization, have been consolidated based on the quality of air MOVPSO Evaluation model. While the generalized weighted fuzzy neural network is introduced into the air quality forecast WFNN work, has been based on WFNN weighted fuzzy neural network prediction model. Finally, experimental verification by measured data and the practical feasibility of the two models.