基于CEEMDAN-WOA-BiLSTM的短期光伏功率预测

Short-Term Photovoltaic Power Forecasting Based on CEEMDAN-WOA-BiLSTM

中文关键词:光伏功率预测  自适应噪声完备集合经验模态分解  鲸鱼优化算法  双向长短期记忆神经网络

keywords:photovoltaic power forecasting  CEEMDAN  WOA  BiLSTM

中文摘要:针对光伏功率数据随机性强、不确定性高以及单一模型预测精度有限等问题，提出一种基于自适应噪声完备集合经验模态分解（complete ensemble empirical mode decompo-sition with adaptive noise，CEEMDAN）-鲸鱼优化算法（whale optimization algorithm，WOA）-双向长短期记忆神经网络（bidirectional long short-term memory，BiLSTM）的短期光伏功率组合预测方法。首先，利用CEEMDAN方法对相似日聚类后的光伏发电功率时间序列信号进行分解，将具有高复杂度的光伏数据转化为具有低复杂度的分量；其次，构建基于BiLSTM的光伏功率预测模型，并利用WOA优化BiLSTM超参数提高模型泛化能力；最后，以澳大利亚某太阳能中心的光伏数据进行算例分析。分析结果表明，所构建的组合预测方法与LSTM、BiLSTM、VMD-WOA-BiLSTM和CEEMDAN-BiLSTM预测方法相比，具有较高的预测精度。

Abstract:To address the challenges posed by the strong randomness and uncertainty nature in photovoltaic（PV）power data，as well as the limited prediction accuracy of a single model，this study proposes a hybrid short-term PV power forecasting method integrating complete ensemble empirical mode decomposition with adaptive noise （CEEMDAN），whale optimization algorithm （WOA），and bidirectional long short-term memory （BiLSTM）. First，the CEEMDAN method is employed to decompose the PV power time series signals after similar-day clustering ，transforming highly complex PV data into low-complexity components. Secondly，a BiLSTM-based PV power prediction model is constructed，and the WOA is utilized to optimize the hyperparameters of BiLSTM，thereby enhancing the model’s generalization capability. Finally，a case analysis is conducted based on the photovoltaic data from a solar energy center in Australia.The analysis results show that the constructed combined forecasting method has higher prediction accuracy compared with the LSTM，BiLSTM，VMD-WOA-BiLSTM and CEEMDAN-BiLSTM forecasting methods.