Comparative study of two different prediction models for winter AOD

摘要

Aerosol optical depth (AOD), one of the key factors affecting the atmosphere visibility, has great influence on the prediction of radiation intensity and photovoltaic power generation. Considering the problem that AOD is difficult to obtain real-timely and conveniently with high accuracy, in this paper, PM2.5 concentration, PM10 concentration and temperature, wind speed grade and relative humidity in the winter, are collected from the ground air quality monitoring site, correlation analysis between these data and AOD is performed, then BP neural network algorithm and support vector machine (SVM) algorithm are adopted respectively to establish the AOD Prediction model. Comparative study results indicate that both SVM model and BP neural network model have strong nonlinear fitting ability, and the SVM model has higher prediction accuracy than BP neural network model.