摘要
针对光伏发电预测模型准确率低的问题,构建了基于变分模态分解-反向传播神经网络(VMD-BPNN)的光伏功率预测模型。对光伏发电数据进行变分模态分解得到不同特征数据,解决了数据的随机性和波动性问题。再采用K-means聚类方法对不同特征数据进行聚类,提高模型的泛化能力。通过集成学习bagging的方法对BPNN进行增强,以达到提高光伏功率预测模型整体稳定性的目的。根据RMSE和NRMSE误差标准进行测试,测试结果表明,基于VMD-BPNN预测模型的NRMSE平均值2.77%,RMSE平均值为2.22%。
Aiming at the problem of low accuracy of photovoltaic power generation prediction model,a photovoltaic power prediction model based on variational modal decomposition back propagation neural network(VMD-BPNN)is constructed.Different characteristic data are obtained by variational modal decomposition of photovoltaic power generation data,which solves the problems of randomness and volatility of data.The K-means clustering method is used to cluster different characteristic data to improve the generalization ability of the model.BPNN is enhanced by integrated learning bagging method to improve the overall stability of photovoltaic power prediction model.According to the RMSE and NRMSE error standards,the test results show that the average value of NRMSE based on vmd-bpnn prediction model is 2.77%,and the average value of RMSE is 2.22%.
作者
杨琴
YANG Qin(Nanchang Power Supply Branch of State Grid Jiangxi Electric Power Co.Ltd.,Nanchang 330069,China)
出处
《工业加热》
CAS
2023年第2期27-31,共5页
Industrial Heating
基金
国家自然科学基金(51867010)
江西省自然科学基金重点项目(20202ACBL214021)
江西省重点研发计划(20202BBGL73098)
江西省教育厅科学技术项目(GJJ190311)。
关键词
光伏电站
功率预测
变分模态分解
反向传播神经网络
K-MEANS聚类
photovoltaic power station
power prediction
variational modal decomposition
back propagation neural network
K-means clustering
