结合气象监测与地基云图的地表太阳辐射预测OACSTPCD
Surficial Solar Irradiation Prediction Based on Meteorological Monitoring Combined with Ground-Based Cloud Images
高精度的太阳辐照度预测是光伏输出功率预测的基础,而云的遮挡是导致太阳辐射波动的主要原因.针对现有技术由于对云图时变特征获取能力不足,导致在复杂天气条件下预测精度显著下降的问题,提出了利用 3D卷积神经网络同时提取单张云图图像特征和云图序列时变特征,建立云图图像特征与云对地表太阳辐射衰减之间的关联,实现太阳辐射高精度预测的方法.实验验证结果表明,较现有方法,所提出的方法在复杂天气条件下的未来 5 min功率预测精度提高 8%以上,具有很高的推广应用价值.
PV output power forecasting is based on high accuracy solar irradiation prediction.Cloud shielding is the primary reason lead-ing to solar irradiation fluctuation.Aiming to solve the problem of prediction accuracy descending in complicated climate conditions due to the insufficient ability of extracting temporal variation feature of Ground-Based Cloud image in related works,a novel solar irradiation prediction method is presented.3D CNN is used to extract features in single GBC image and temporal variation features in GBC image series at the same time and the relationship between GBC image feature and solar irradiance attenuation by cloud is established.The evaluation results of experiments show that,in complicated climate conditions,the RMSE of 5 minutes ahead prediction is improved by 8%for the proposed method compared with existing works.The proposed method has high promotion value.
郁云;王一海;曹潇
南京信息职业技术学院数字商务学院,江苏 南京 210046中国电力科学研究院有限公司新能源研究中心,北京 100192
计算机与自动化
地基云图特征提取3D卷积神经网络太阳辐射衰减光伏功率预测
ground-based cloud imagefeature extraction3D CNNsolar irradiance attenuationPV output power prediction
《电子器件》 2024 (001)
134-139 / 6
南京信息职业技术学院校青蓝工程优秀骨干教师培养对象项目(2020XGG10)
评论