基于随机森林算法的日光温室内气温预测模型研究

开展日光温室气温预报,为农业生产提供参考,指导农户采取调控措施,为作物生长提供适宜条件,促进品质和产量提升。研究选取温室外气温、日照等气象因子,建立随机森林算法预测模型,就室内最低、最高气温进行拟合预测分析和预测因子重要性评估。结果表明,温室内最低、最高气温拟合值与观察值的拟合度分别达99.69%和99.85%,温室外最低气温是室内最低气温的重要预测因子,室外日照是室内最高气温的重要预测因子。同时建立支持向量机、神经网络、多元回归、逐步回归模型,通过对各个模型中平均绝对误差、均方根误差等3个指标进行比较,得出随机森林模型的预测精度优于其他模型。基于随机森林算法的气温预测模型精确度较高,可推广应用到后期日光温室气温预测中。

Temperature Prediction Model in Solar Greenhouse Based on Stochastic Forest Algorithm

The study carries out temperature forecast of solar greenhouse, aiming to provide reference for agricultural production, guide greenhouse temperature control to ensure a suitable condition for crop growth, and promote agriculture products’ quality and yield. Meteorological factors such as temperature and sunshine outside the greenhouse were selected to build a prediction model based on random forest algorithm, and then the indoor minimum and maximum temperatures were fitted for prediction analysis and the importance of the prediction factors were evaluated. Results showed that the fitting degree of the fitting value and the observed value of the lowest and the highest air temperature in greenhouse was 99.69% and 99.85%, respectively. The lowest air temperature outside the greenhouse was an important predictor of the indoor minimum air temperature, and the outdoor sunshine was an important predictor of the indoor maximum air temperature. At the same time, the support vector machine, neural network, multiple regression and stepwise regression models were established. By comparing the mean absolute error and root-mean-square error in each model, the prediction accuracy of the random forest model was better than that of other models. The air temperature prediction model based on random forest algorithm is more accurate, which can be popularized in the air temperature prediction of solar greenhouse.