基于神经网络的东北地区秸秆焚烧预测

选取农作物秸秆露天燃烧严重的东北地区，采用人工神经网络的方法，结合卫星火点和气象数据，开展秸秆露天燃烧预测研究.结果表明：人工神经网络预测模型成功验证了松嫩平原地区2015年10月25日~11月15日的秸秆露天燃烧情况，其准确度为67.1%，经过多次试验，在神经网络建模与验证数据配比为80：20时，预测准确度最高，可达69.7%，同时该模型的稳定性较好.而对不同区域，不同时间段的预测研究表明，人工神经网络较适用于长时间序列的预测.就影响因素而言，相对湿度是影响秸秆露天燃烧的最重要因素.本研究结果可为空气质量模式提供火点预测数据，提高其预报预警能力，为区域联防联控政策的制定提供科技支持.

Forecasting of agricultural straw burning in the Northeastern China based on neural network

The open agricultural burning produces large amount of air pollutants which contribute significantly to air pollution and climate change. Having vast areas of farmland, the environment of Northeastern China is impacted by the agricultural burning every year, especially during the harvest season. In order to further understand the open field burning activities and improve the prediction ability of it, in this study, the artificial neural network was introduced to conduct the forecasting of crop straw open field burning in Northeastern China. The results showed that the neural network model has a good prediction stability and it successfully reproduced the crop straw burning occurred in the Songnen Plain during the period from October 25 to November 15 2015, with an accuracy of 67.1%. After a series of tests, the forecasting accuracy of the neural network model improved, and could reach up to 69.7% when the ratio of training set to validation set was 80:20. In addition, according to the results of different research time periods, this neural network showed higher performance on the long-term prediction. Furthermore, comparing to with wind speed, precipitation, temperature and pressure, the relative humidity is the most important meteorological factor that affects straw open field burning. This study could not only help to improve the fire emission data used for the air quality forecasting model, but also provided technical support for government departments in controlling the open agriculture burning.