【摘要】 针对建筑能耗数据存在的强波动性、高复杂性等特性，提出了一种基于完全集合经验模态分解(CEEMDAN)与变分模态分解(VMD)的二次分解策略和长短期记忆网络(LSTM)与多元线性回归(MLR)的组合预测模型(CEEMDAN-VMD-LSTM-MLR)。首先，利用CEEMDAN算法对建筑能耗数据进行初次分解，并通过样本熵对各IMF分量进行复杂度评估与重构，形成高频、中频和低频3类时序子集。然后，采用VMD算法对高频分量进行二次分解，进一步降低能耗序列的复杂程度。具体针对不同子序列的数据特征，分别采用MLR模型与LSTM模型进行预测。实验结果表明，相较于传统的一次分解预测模型(如CEEMDAN-LSTM-MLR)及单一预测模型(如CEEMDAN-VMD-LSTM),所提出的CEEMDAN-VMD-LSTM-MLR模型在预测精度上表现出显著优势。研究成果为短期建筑能耗预测提供了一种新的思路和方法，对于推动建筑节能减排、保障能源供应稳定性具有重要意义。更多还原

【Abstract】 In response to the characteristics of strong volatility and high complexity inherent in building energy consumption data, a secondary decomposition strategy is proposed based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN) and Variational Mode Decomposition(VMD), along with a hybrid prediction model combining Long Short-Term Memory networks(LSTM) and Multiple Linear Regression(MLR), termed as CEEMDAN-VMD-LSTM-MLR. Initially, the CEEMDAN algorithm is employed to perform primary decomposition of the building energy consumption data. Subsequently, sample entropy is utilized to evaluate the complexity of each Intrinsic Mode Function(IMF) component, facilitating their reconstruction into three temporal subsets: high-frequency, mid-frequency, and low-frequency. The VMD algorithm is then applied to conduct secondary decomposition of the high-frequency components, further reducing the complexity of the energy consumption series. In light of the distinct data characteristics of different subsequences, the MLR model and LSTM model are respectively adopted for prediction. Experimental results demonstrate that the proposed CEEMDAN-VMD-LSTM-MLR model exhibits a notable advantage in prediction accuracy compared to traditional single-stage decomposition prediction models(e.g., CEEMDAN-LSTM-MLR) and standalone prediction models(e.g., CEEMDAN-VMD-LSTM). The findings of this study offer a novel perspective and methodology for short-term building energy consumption prediction, which holds significant implications for promoting energy conservation and emission reduction in buildings, as well as ensuring the stability of energy supply.更多还原