Accurate power load forecasting is critical for ensuring the stable operation of port microgrids. To address the inherent randomness and volatility of port power loads, this paper proposes a novel forecasting method that integrates optimized mode decomposition with error correction. The Pearson correlation coefficient is employed to evaluate the correlation between the load and influencing factors, such as production operations. Optimal parameters for the mode decomposition model are determined using the Sparrow Search Algorithm (SSA), which effectively decomposes the original load signal into distinct components that capture both the overall trend and local characteristics. Subsequently, a prediction and reconstruction model incorporating error correction is established. This model leverages forecasted error values to refine preliminary predictions, thereby significantly improving forecasting accuracy. Validation using real-world port operational data demonstrates the effectiveness of the proposed method across multiple time scales. Notably, the performance is particularly robust for 8-h-ahead and 24-h-ahead forecasts, achieving a Mean Absolute Percentage Error (MAPE) below 5% and forecasting accuracy as high as 95%.