摘要
光伏功率时间序列受到多种因素影响,呈现出高度的随机性和波动性。针对光伏功率时间序列可预测性低的问题,提出了一种结合变分模态分解(variationalmodal decomposition,VMD)、深度回声状态网络(deepechostate network,DESN)和稀疏高斯混合过程专家模型(mixtureof sparse gaussian process experts model,MSGP)的超短期光伏功率预测方法。首先采用VMD将光伏功率时间序列分解为不同的模态,降低数据的非平稳性;为提高模型在超短尺度时序的预测能力,对各模态分别建立DESN预测模型,将各模态预测结果进行求和重构;为进一步提高模型预测精度,对误差的特性进行分析,采用MSGP对预测误差进行补偿;最后将误差的预测值与原功率的预测值相叠加作为最终预测结果。仿真结果表明,该方法在光伏功率时序预测中的效果比传统预测模型更好,有效提高了超短期光伏功率时间序列预测的准确性。
Photovoltaic power time series is affected by many factors,showing a high degree of randomness and volatility.Aiming at the problem of low predictability of photovoltaic power time series,an ultra-short-term photovoltaic power prediction method combining variational mode decomposition(VMD),deep echo state network(DESN)and mixture of sparse gaussian process experts model(MSGP)is proposed.Firstly,the VMD is used to decompose the photovoltaic power time series into different modes to reduce non-stationarity of the data.To improve prediction ability of the model in ultra-short-scale time series,a DESN prediction model is established for each mode,and modal prediction results are obtained.In order to further improve prediction accuracy of the model,the error characteristics are analyzed,and the prediction error is compensated with MSGP.Finally,the predicted value of the error is superimposed on the predicted value of original power to obtain final prediction result.Simulation results show that the proposed method is better than traditional prediction model in the prediction of photovoltaic power time series,effectively improving accuracy of ultra-short-term photovoltaic power time series prediction.
作者
王粟
江鑫
曾亮
常雨芳
WANG Su;JIANG Xin;ZENG Liang;CHANG Yufang(School of Electrical and Electronic Engineering,Hubei University of Technology,Wuhan 430068,Hubei Province,China;Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System(Hubei University of Technology),Wuhan 430068,Hubei Province,China)
出处
《电网技术》
EI
CSCD
北大核心
2020年第3期917-926,共10页
Power System Technology
基金
国家自然科学基金项目(41601394、61903129)
湖北工业大学博士科研启动基金项目(BSQD2017008).
关键词
光伏功率预测
时间序列
变分模态分解
深度回声状态网络
稀疏高斯混合过程专家模型
photovoltaic power prediction
time series
variational modal decomposition(VMD)
deep echo state network(DESN)
mixture of sparse gaussian process experts model(MSGP)
