摘要
为提高光伏发电功率短期预测的精度,提出一种结合时变滤波经验模态分解和北方苍鹰优化算法优化长短期记忆神经网络的组合预测方法。首先,利用时变滤波经验模态分解将光伏发电功率分解成多个固有模态函数分量。其次,利用北方苍鹰优化(northern goshawk optimization,NGO)算法优化长短期记忆(long short-term memory,LSTM)神经网络隐含单元的个数、最大训练次数和初始学习率,构建NGO-LSTM预测模型。最后,把每一个固有模态函数分量都输入到预测模型中进行预测,将所有固有模态函数分量的预测结果进行叠加便可得到光伏发电功率短期预测的结果。仿真结果表明,所提的预测模型可以有效提高光伏发电功率的预测精度。
In order to improve the accuracy of short-term prediction of photovoltaic power generation,a combined prediction method combining time-varying filtering based empirical mode decomposition and northern goshawk optimization algorithm to optimize long-term and short-term memory neural network was proposed.Firstly,the time-varying filtering based empirical mode decomposition was used to decompose the photovoltaic power generation into multiple intrinsic mode function components.Then,by utilizing the northern goshawk optimization(NGO)algorithm to optimize the number of hidden units,maximum training times,and initial learning rate of long short term memory(LSTM)neural networks,an NGO-LSTM prediction model was constructed.Finally,each intrinsic mode function component was input into the model for prediction,and the prediction results of all intrinsic mode function components were superimposed to obtain the short-term prediction results of photovoltaic power generation.The simulation results show that the proposed prediction model can effectively improve the prediction accuracy of photovoltaic power generation.
作者
陈晓华
吴杰康
CHEN Xiaohua;WU Jiekang(Zhanjiang Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Zhanjiang 524005,China;School of Automation,Guangdong University of Technology,Guangzhou 510006,China;School of Electrical Engineering&Intelligentization,Dongguan University of Technology,Dongguan 523808,China)
出处
《山东电力技术》
2024年第10期10-17,共8页
Shandong Electric Power
基金
国家自然科学基金项目(50767001)。
关键词
时变滤波经验模态分解
北方苍鹰优化算法
光伏发电功率
短期预测
长短期记忆神经网络
time varying filtering based empirical mode decomposition
northern goshawk optimization algorithm
photovoltaic power
short-term forecast
long short-term memory neural network
