Short-Term Bus Load Forecasting for Substations Based on Secondary Decomposition and ISSA-TCN

Bus load forecasting is crucial for the stable operation of substations. With the integration of renewable energy, electricity consumption patterns have diversified, increasing volatility in bus loads. To address this, we propose a short-term forecasting method using secondary decomposition and an improved Sparrow Search Algorithm optimized Temporal Convolutional Network (ISSA-TCN). The original bus load is first decomposed using CEEMDAN. The high-frequency components, which exhibit strong fluctuations, undergo a secondary decomposition via VMD to extract their characteristics effectively. We assess the complexity of these subcomponents by calculating their sample entropy and reconstruct them into three groups. These groups are then fed into the ISSA-TCN model, which integrates Tent chaotic mapping, adaptive weighting, Levy flight, and variable spiral search strategies to enhance forecasting accuracy. The method is validated through case studies using real bus load data, and comparative analyses with other models demonstrate its superior performance, contributing to better operational management of substations.