基于时频熵和神经网络的光伏发电功率预测模型

来源期刊:中南大学学报(自然科学版)2020年第1期

论文作者:孙志强 李东阳

文章页码:221 - 231

关键词:集合经验模态分解;Hilbert变换;时频熵;神经网络;光伏发电功率预测

Key words:ensemble empirical mode decomposition; Hilbert transform; time-frequency entropy; neural network; photovoltaic power prediction

摘    要:针对光伏电站并网后发电功率波动较大影响电网合理调度及平稳运行的问题,提出一种基于EEMD-Hilbert变换及时频熵的神经网络光伏发电功率预测方法。将电站光伏发电功率历史值按晴天、雨天及多云天分类并分别用集合经验模态分解(ensemble empirical mode decomposition,EEMD)方法分解成若干频率由高到低的内禀模态分量,经游程检测法重构成高频和中频分量,再由Hilbert变换得到高频和中频分量各数据点的频率与幅值,构造信号的能量谱图,提取含有时频信息的特征参数时频熵,将其与EEMD分解后得到的中频或高频分量以及温度、光照强度、风速、相对湿度等气象参数作为BP(back propagation)神经网络的输入,构建预测模型对不同天气条件下短期和超短期光伏发电功率进行预测。研究结果表明:此方法预测精度较高,发电功率预测误差基本上小于仅用气象数据直接预测的误差;晴天、雨天和多云天的短期光伏发电功率预测准确率分别为0.995,0.944和0.931,超短期预测准确率分别为0.996,0.984和0.991。

Abstract: Aiming at the problem that the power generation fluctuation of PV power grid is connected to the grid and affects the rational dispatching and smooth operation of the power grid, a neural network photovoltaic power generation prediction method based on EEMD-Hilbert transform and time-frequency entropy was proposed. The historical data of power generation photovoltaic power generation was classified according to sunny, rainy and cloudy days and decomposed into high frequency to low internal modal components by the ensemble empirical mode decomposition(EEMD) method. The weights were composed of high frequency and intermediate frequency components, the frequency and amplitude of each data point of the high frequency and intermediate frequency components were obtained by Hilbert transform. The energy spectrum of the signal was constructed, and the time-frequency entropy of the characteristic parameters containing time-frequency information was extracted. The intermediate frequency or high frequency component obtained by decomposition of EEMD and the meteorological parameters such as temperature, light intensity, wind speed and relative humidity were used as input of BP(back propagation) neural network to construct a prediction model of short-term and ultra-short-term photovoltaic generation under different weather conditions for power forecasting. The results show that the prediction accuracy of this method is relatively high, and the prediction error is basically smaller than the direct prediction error using only meteorological data. The short-term photovoltaic power prediction accuracy on sunny, rainy and cloudy days is 0.995, 0.944 and 0.931, respectively, and super short-term photovoltaic power prediction accuracy on sunny, rainy and cloudy days is 0.996, 0.984 and 0.991, respectively.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号