<p>Conventional electricity consumption forecasting methods typically integrate multiple factors, including historical power consumption data, natural environmental conditions, and socio-economic activities, to predict future electricity consumption within specific regions and time periods. However, the black-box nature of the prediction process and the lack of interpretability often hinder model optimization and limit improvements in prediction accuracy. In this paper, we propose a hierarchical electricity consumption forecasting visualization system. First, a bottom-up hierarchical forecasting architecture based on LSTM-KAN model is designed to achieve multi-scale predictions at fine-grained node levels in the underlying layers. Then, these predictions are progressively aggregated in an interpretable manner to generate comprehensive consumption forecasts. Furthermore, we introduce a spatiotemporal coupled visual mapping method that presents prediction results and error distributions across different hierarchical levels and temporal scales through multiple dimensions. This approach supports interactive error backtracking and dynamic parameter adjustment, enabling visual validation and iterative optimization of prediction outcomes. Case studies using real-world urban electricity consumption data, quantitative analyses, and expert interviews demonstrate the effectiveness and practical utility of our method in optimizing electricity forecasting models and enhancing prediction accuracy.</p> Graphical Abstract <p></p>