基于TVF-EMD-ELM的超短期光伏功率预测

Ultra-short-term Photovoltaic Power P

针对光伏功率预测方法精度不高和时效性低的问题,提出了一种基于时变滤波经验模态分解(TVF-EMD)和极限学习机(ELM)相结合的超短期光伏功率预测方法。首先应用TVF-EMD方法对光伏功率数据进行分解,以便得到一组相对稳定的分量,降低不同功率影响因素之间的交互影响。然后采用ELM神经网络模型,根据各分量的特点构建不同的预测模型,来预测各个分量的值。将ELM预测的各分量值相加,从而获得最终的预测结果。算例结果表明该方法的有效性,相比传统模型其归一化均方根误差值降低了25.8%,标准平均绝对误差下降了17.97%,相关系数提高了8.3%。

Aiming at the problems of low accuracy and low timeliness of photovoltaic power prediction method,an ultra short-term photovoltaic power prediction method based on the combination of Time Varying Filtering-Empirical Mode Decomposition(TVF-EMD)and Extreme Learning Machine(ELM)was proposed.First,the TVF-EMD method was used to decompose the photovoltaic power data,so as to obtain a group of relatively stable components and reduce the interaction between different power influencing factors.Then the ELM neural network model was used to construct different prediction models according to the characteristics of each component to predict the value of each component.The values of each component predicted by ELM were added to obtain the final prediction result.Example results show the effectiveness of the proposed method.Compared with the traditional model,the normalized root mean square error decreases by 25.8%,the standard mean absolute error decreases by 17.97%,and the correlation coefficient increases by 8.3%.