大气污染物浓度的投影寻踪自回归预测

Prediction of the Concentration of Atmospheric Pollution by Projection Pursuit Auto-regression

环境污染预测是环境科学研究的重要课题之一。线性回归预测、时序预测和灰色GM预测是常用的大气污染预测方法,但始终未能摆脱“假定—模拟—预报”的模式束缚,难以适应非正态、非线性分布的客观内在规律。投影寻踪高新技术,对数据结构或特征不作任何假定,而是通过直接审视和分析数据,应用计算机软件程序设计,为时序分析开辟了一条“审视数据—模拟—预报”的数据分析新途径。该文采用投影寻踪自回归新技术,建立了大气污染物浓度的预测模型,并将该模型用于成都市1982～1987年期间的大气SO_2和NO_x浓度季节均值的建模和预测,获得较好的拟合和预测效果。

Environmental pollution forecast is one of the important themes of environmental sciences study. Linear regression forecast, time order forecast and grey GM forecast are ordinary methods for forecasting atmospheric pollution, but all along could not get rid of the bind of 'assumption——simulation——forecast' model, this model is hard to fit for abnormal or non-linear distribution objective internal law. The highly new technique of projection pursuiting does not take any assumption for its data structure or characteristics, but through direct observation and data analysis, applying computer soft-ware design to develop a new analytical way by observation-simulation-forecast for time order analysis. The study adopting projection pursuit self-regression new technique to establish atmospheric pollution concentration forecast model and applied to 1982-1987, the seasonal average value of SO2 and NOx in Chendu for modelling and forecasting and obtained good result.