A load forecasting approach for integrated energy systems based on aggregation hybrid modal decomposition and combined model

Abstract

To improve the granularity of load decomposition within integrated energy system (IES), a novel multiple load forecasting approach is proposed. This method leverages an aggregated hybrid modal decomposition (AHMD) strategy and a composite model approach to thoroughly investigate the inherent characteristics of load profiles. Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) is utilized for the primary decomposition of electric, cooling, and heating loads, yielding a series of subsequences. To further refine the analysis, Fuzzy Dispersion Entropy (FDE), and Over-zero Rate (OZR) are employed to aggregate the decomposed subsequences. Subsequently, Successive Variational Mode Decomposition (SVMD) is applied to break down the high-frequency components in a secondary manner. The adoption of the rolling decomposition strategy effectively circumvents the information leakage problem caused by traditional modal decomposition-prediction methods. In concert with environmental features, the Multiple Linear Regression (MLR) model forecasts the low-frequency components. Meanwhile, the Temporal Convolutional Network-Bidirectional Gated Recurrent Unit (TCN-BiGRU) fusion neural network is deployed for predicting mid-frequency and high-frequency components. Additionally, a Multi-Head Attention (MHA) mechanism is introduced to effectively assign weights to the features. The predictions of each component are superimposed to obtain the final forecast. An example of an integrated campus energy system is used to compare the predictive effectiveness of different models. The results show that the proposed method exhibits superior prediction accuracy across multiple load types. This approach significantly enhances energy efficiency and fosters the advancement of sustainable energy practices.

Tasks

Reproductions