A Novel Interpretable Short-Term Load Forecasting Method Based on Kolmogorov-Arnold Networks
Bozhen Jiang, Yidi Wang, Qin Wang, Hua Geng
Abstract
Short-term load forecasting (STLF) plays a crucial role in the efficient and economical management of power systems. While artificial neural networks have achieved significant success in STLF, they suffer from the limitation of providing a black box representation, making it challenging to obtain an analytical expression between features and loads. This limitation hampers subsequent quantitative analysis, which is crucial for artificial intelligence based decision-making processes. To address this issue, this paper proposes a novel STLF approach through the utilization of Kolmogorov-Arnold Networks (KANs). By leveraging KANs, the interpretability of model parameters can be enhanced. As a result, detailed analytical expressions of the model can be derived. To validate the proposed approach, we conducted experiments by comparing the forecasting performances among KANs, multi-layer perceptrons and XGBoost on a publicly available dataset from Switzerland. Numerical results demonstrate the effectiveness of the proposed KAN-based STLF method in accurately forecasting short-term loads. Additionally, the KAN-based approach provides the advantage of yielding the analytical expression for STLF, enabling further insights and analysis.