Application Study Of Artificial Neural Networks In Air Quality Forecast

With the development of society and economy and the population expansion of the city, air pollution becomes more and more serious, therefore, it is significant to study air quality forecasting so as to understand environmental variation trend, prevent and control air pollution. Jinan city environment monitoring has carried on urban air quality monitoring since 1999 to stand, having accumulated a large amount of historical data of monitoring. The historical data and predict to the quality analysis of whole urban surrounding air has very important meanings. With the setting-up of real-time monitoring system and real-time online monitoring system of pollution sources of air quality, the growth of the data is faster. Generaly the atmospheric environmental system is complex, and now mass historical monitoring data has accumulated, therefore, the traditional forecasting methods have difficulty in excavating useful information from the mass data to forecasting precisely. This paper applied the artificial neural network(ANN) in the air quality forecasting, presented an air quality forecasting method based on artificial neural networks in virtue of the ANN's good nonlinear processing ability and fault-tolerant ability, According to the actual application, improvement commonly used model and application to a specific environment, back propagation neural network(BPNN) air quality forecasting based on samples self-organizing clustering and resource allocating network(RAN) air quality forecasting model based on hidden node correlation pruning were constructed, and then provided a accurate and comprehensive environmental air quality information for making decision.The main work of this paper is listed as follows:(1)Research on the error back propagation neural network and radial basis function neural network(RBFNN) of the structure and training methods, analyze BP network defects, explore a variety of RBF neural network learning method.(2)Train samples to usually have inherent characteristic and regularity according to the neural network in practical application, this paper presents a BP neural network predicting model based on samples self-organizing clustering. Improve the effect of samples training on BP neural network performance with the clustering characteristic of self-organizing competitive network. BP neural network use adaptive learning rate momentum algorithm has fast convergence rate and high error precision. And according to the air quality forecast experiment based on this kind of model, indicate that BP neural network predicting model based on samples self-organizing clustering improve convergence rate at first, secondly will reduce the possibility of get into local minimum, improve the prediction accuracy.(3)Air quality forecast for a variety of nonlinear factors on the impact of prediction accuracy, and the air pollution time-varying property, adopted to resource allocation network algorithm for online learning, combined with hidden layer nodes correlation pruning algorithm and deleting strategy of useless hidden node, simplify the network structure, Improve the generalization ability, establish a simple structure air quality forecast model with online learning. Through training and testing networks model, the results shows that RAN forecasting method based on hidden node pruning algorithm is not only simplifies the network structure of RAN, but also the prediction accuracy is better than RAN.