摘要
针对光伏功率预测中未充分考虑光伏功率季节性与趋势特征的问题,提出一种基于Neural-Prophet(NP)与深度神经网络的光伏功率预测方法。首先,通过互信息法筛选出影响光伏功率的主要因素,利用NP模型对光伏功率建模得到光伏功率的季节性与趋势特征,将季节性与趋势特征及主要影响因素作为模型输入。其次,采用改进残差网络(ResNet)和双向门控循环单元(BiGRU)建立NP-ResNet-BiGRU光伏功率预测模型并完成光伏功率预测。利用春夏秋冬四季的数据进行实验,结果显示相较于其他方法,所提方法的MAE至少提升7.44%,RMSE至少提升4.62%。
A photovoltaic power prediction method based on Neural-Prophet(NP)and deep neural networks is proposed to address the issue of insufficient consideration of seasonal and trend characteristics of photovoltaic power in photovoltaic power prediction.Firstly,the main factors affecting photovoltaic power are selected through the mutual information method.The PV power is then modeled using the NP model to obtain the seasonal and trend characteristics of PV power.These seasonal and trend characteristics,along with the main influencing factors,are used as inputs to the model.Secondly,an NP-ResNet-BiGRU photovoltaic power prediction model was established using an improved residual network(ResNet)and a bidirectional gated loop unit(BiGRU),and the photovoltaic power prediction was completed.Experiments are conducted using data from the four seasons—spring,summer,autumn,and winter.Results show that,compared to other methods,the proposed method improves the MAE metric by at least 7.44%and the RMSE metric by at least 4.62%.
作者
王东风
李青博
张博洋
黄宇
Wang Dongfeng;Li Qingbo;Zhang Boyang;Huang Yu(Department of Automation,North China Electric Power University,Baoding 071003,China)
出处
《太阳能学报》
北大核心
2025年第3期348-356,共9页
Acta Energiae Solaris Sinica
基金
中央高校基本科研业务费专项(2021MS089)。
