基于CEEMDAN-VMD-BiGRU模型的印刷园区电力负荷预测

Electricity Load Forecasting for Printing Industrial Parks Based on CEEMDAN-VMD-BiGRU Model

为确保印刷园区绿色低碳运行,充分了解和预测电力负荷需求至关重要。然而,电力负荷预测面临数据复杂性、噪声干扰等问题,导致难以捕捉信号的非线性特征。为解决这一问题,本研究提出了基于CEEMDAN-VMD-BiGRU模型的印刷园区电力负荷预测方法。首先,使用自适应噪声完备集合经验模态分解(CEEMDAN)对印刷园区电力负荷数据进行分解,并计算各模态分量的排列熵(PE)。其次,利用变分模态分解(VMD)对熵值高频模态分量进行二次分解,从而减弱序列的非平稳性。最后,使用各分量作为双向门控循环单元(BiGRU)预测模型的输入进行训练,将模型结果线性叠加,得到最终预测结果。以某印刷园区电力负荷数据为研究对象,对提出的模型进行与BiGRU模型、CEEMDAN-BiGRU模型预测结果的比较。实验结果表明:融合VMD方法的二次分解模型R2达到97.13%,RMSE为25.354,MAE为62.776,相较于其他模型,本研究提出的预测模型性能更佳。

To ensure the green and low-carbon operation of the printing park,a thorough understanding and prediction of electricity demand is crucial.However,electricity load forecasting faces obstacles like data complexity and noise interference,which complicate the capture of nonlinear signal characteristics.In response,a method for forecasting electricity load in printing parks based on the CEEMDAN-VMD-BiGRU model was introduced.Initially,the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)was used to break down the electricity load data,after which the Permutation Entropy(PE)was computed for each mode component.Subsequently,the Variational Mode Decomposition(VMD)was employed to perform a secondary decomposition specifically on the high-entropy frequency mode component,with the goal of addressing non-stationarity within the sequence.Finally,each component was used as input for the Bidirectional Gated Recurrent Unit(BiGRU)neural network model,and the model results were linearly combined to obtain the final prediction.The study focused on the electricity load data of a printing park and compared the proposed model with the predictions of the BiGRU model and the CEEMDAN-BiGRU model.The experimental results indicated that the second-order decomposition model 2 incorporating VMD achieves an R of 97.13%,with RMSE of 25.354 and MAE of 62.776.Compared to other models,the proposed prediction model demonstrates superior performance.