As modern power systems continue to evolve, accurate power load forecasting remains a critical issue. The phase space reconstruction method can effectively retain the chaotic characteristics of power load from a system dynamics perspective and thus is a promising knowledge-based preprocessing method for power load forecasting. However, limited by its fundamental theory, there is still a gap in implementing a multi-step forecasting scheme in current studies. To bridge this gap, this study proposes a novel multi-step forecasting approach by integrating the PSR with neural networks. Firstly, the useful features in the phase trajectory obtained from the preprocessing of PSR are discussed in detail. Through mathematical derivation, the equivalent characterization of the PSR and another time series preprocessing method, patch segmentation, is demonstrated for the first time. Based on this prior knowledge, an image-based modeling perspective with the global and local feature extraction strategy is introduced. Subsequently, a novel deep learning model, namely PSR-GALIEN, is designed for end-to-end processing, in which the Transformer Encoder and 2D-convolutional neural networks are employed for the extraction of the global and local patterns in the image, and a multi-layer perception based predictor is used for the efficient correlation modeling. Then, extensive experiments are conducted on five real-world benchmark datasets to verify the effectiveness as well as to have an insight into the detailed properties. The results show that, comparing it with six state-of-the-art deep learning models, the forecasting performance of PSR-GALIEN consistently surpasses these baselines, which achieves superior accuracy in both intra-day and day-ahead forecasting scenarios. At the same time, a visualization-based method is proposed to explain the attributions of the forecasting results.

翻译：随着现代电力系统的持续发展，精确的电力负荷预测仍是关键问题。相空间重构方法能够从系统动力学角度有效保留电力负荷的混沌特性，因此是一种具有前景的基于知识的电力负荷预测预处理方法。然而受其基础理论限制，当前研究中实施多步预测方案仍存在空白。为弥补这一不足，本研究提出一种融合相空间重构与神经网络的新型多步预测方法。首先，详细讨论了经相空间重构预处理所得相轨迹中的有效特征。通过数学推导，首次证明了相空间重构与另一种时间序列预处理方法——片段分割的等效表征关系。基于此先验知识，引入了结合全局与局部特征提取策略的图像化建模视角。随后，设计了一种新型深度学习模型PSR-GALIEN进行端到端处理，其中采用Transformer编码器与二维卷积神经网络提取图像中的全局与局部模式，并利用基于多层感知机的预测器进行高效相关性建模。接着，在五个真实世界基准数据集上开展大量实验以验证有效性并深入探究其详细特性。结果表明，与六种先进深度学习模型相比，PSR-GALIEN的预测性能始终超越这些基线模型，在日内与日前预测场景中均实现了更优的精度。同时，提出了一种基于可视化的方法来解释预测结果的归因。