Two-Stage Combined Model for Short-Term Electricity Forecasting in Ports

With an increasing emphasis on energy conservation, emission reduction, and power consumption management, port enterprises are focusing on enhancing their electricity load forecasting capabilities. Accurate electricity load forecasting is crucial for understanding power usage and optimizing energy allocation. This study introduces a novel approach that transcends the limitations of single prediction models by employing a Binary Fusion Weight Determination Method (BFWDM) to optimize and integrate three distinct prediction models: Temporal Pattern Attention Long Short-Term Memory (TPA-LSTM), Multi-Quantile Recurrent Neural Network (MQ-RNN), and Deep Factors. We propose a two-phase process for constructing an optimal combined forecasting model for port power load prediction. In the initial phase, individual prediction models generate preliminary outcomes. In the subsequent phase, these preliminary predictions are used to construct a combination forecasting model based on the BFWDM. The efficacy of the proposed model is validated using two actual port data, demonstrating high prediction accuracy with a Mean Absolute Percentage Error (MAPE) of only 6.23% and 7.94%. This approach not only enhances the prediction accuracy but also improves the adaptability and stability of the model compared to other existing models.