3D-CNN-Based Spatial Load Forecasting Method Considering the Spatial-Temporal Influences of Multi-Source Data on Loads among Adjacent Cells

Accurate spatial load forecasting is an important aspect of distribution network planning and power equipment configuration. Current research in spatial load forecasting primarily relies on time series forecasting with insufficient consideration of the spatial influences among adjacent cells, which may not provide a comprehensive explanation for the development of spatial loads. Therefore, in this work, a spatial load forecasting method based on a three-dimensional convolutional neural network is proposed, taking into account the spatial-temporal influences of multi-source data on loads among adjacent cells. First, a three-dimensional data structure is proposed to establish a spatial-temporal dataset that incorporates the multi-source data into specific cells. Second, a spatial load forecasting method based on a three-dimensional convolutional neural network is proposed using the spatial-temporal dataset, which quantifies the nonlinear spatial-temporal influences of multi-source data on loads among adjacent cells in the target year. Finally, case studies on cells within an actual distribution network are performed to verify the effectiveness of the proposed method. The simulation results demonstrate that the proposed method can quantify the spatial-temporal influences of multi-source data on loads among adjacent cells and provide relatively more accurate long-term spatial load forecasting results for these cells than trend extrapolation-based SLF methods.