This paper proposes a novel approach, addressing the underexplored potential of convolutional networks in long-term time series forecasting research. Specifically, we propose a novel multi-scale time series reshape module that effectively captures cross-period patch interactions and variable dependencies. Based on this, we develop MS-DFTVNet, a multi-scale 3D deformable convolutional framework specialized for long-term forecasting. Furthermore, to address the uneven distribution of temporal features, we introduce a context-aware dynamic deformable convolution mechanism, enabling better capture of complex temporal patterns.
