基于随机森林模型的太阳辐射中长期变化分析
ANALYSIS OF MEDIUM-AND LONG-TERM CHANGES IN SOLAR RADIATION BASED ON RANDOM FOREST MODEL
贾兴斌 1汪国菊 1王仁政 2宫响1
作者信息
- 1. 青岛科技大学数理学院,青岛 266061;青岛市人工智能海洋技术创新中心,青岛 266061
- 2. 中国海洋大学环境科学与工程学院,青岛 266100
- 折叠
摘要
该文基于多源辐射观测资料,采用随机森林(RF)算法、季节差分自回归移动平均(SARIMA)模型及特征重要性等方法,对山东省济南市太阳辐射长期变化趋势和影响因素进行综合分析.结果显示:RF模型拟合月太阳辐射效果较好,决定系数和平均绝对百分比误差分别为0.92和9%,优于SARIMA模型;济南市及周边地区月太阳辐射1980-2020年经历"变暗"到"变亮"的过程,空间呈现西北高东南低的特点;最高温度和日照时数是影响太阳辐射月变化拟合准确度的主要因素,降雨量是导致月太阳辐射总量突变的重要原因,大气污染物中SO2和O3与太阳辐射的相关性最大.
Abstract
This paper presents a comprehensive analysis of long-term solar radiation trends and influencing factors in Ji'nan City,Shandong Province,based on multi-source radiation observations,using the random forest(RF)algorithm,seasonal autoregressive integrated moving average(SARIMA)model and characteristic importance.The results show that the RF model fits the monthly solar radiation better,with the coefficient of determination and the mean absolute percentage error of 0.92 and 9%,respectively,which are better than the SARIMA model.The monthly solar radiation in Ji'nan City and the surrounding area experiences the process of"darkening"to"brightening"from 1980 to 2020.The maximum temperature and sunshine hours are the main factors affecting the accuracy of the monthly variation of solar radiation,rainfall is an important cause of sudden changes in total monthly solar radiation,and SO2 and O3 among atmospheric pollutants have the greatest correlation with solar radiation.This paper is important for guiding the development of the solar energy industry and environmental management in Ji'nan.
关键词
太阳辐射/随机森林/特征选择/SARIMA模型/拟合分析Key words
solar radiation/random forest/feature selection/SARIMA model/fitting analysis引用本文复制引用
基金项目
国家自然科学基金-山东省联合基金(U1906215)
青岛科技大学研究生自主创新研究专项(S2022KY029)
出版年
2024
NETL
NSTL
万方数据