LSTM Short-term Electricity Load Forecasting Model Based on Improved Sparrow Search Algorithm and Ensemble Learning

Accurate short-term power load forecasting is crucial for power system management and planning, significantly enhancing operational efficiency, optimizing dispatching, and conserving energy resources. Although the Long Short-Term Memory (LSTM) network has proven effective in power load forecasting,...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Engineering letters Jg. 33; H. 8; S. 2996
Hauptverfasser: Gong, Ze, Zhang, Yong
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Hong Kong International Association of Engineers 01.08.2025
Schlagworte:
Algorithms
Datasets
Effectiveness
Electrical loads
Electricity
Energy sources
Ensemble learning
Error reduction
Forecasting
Machine learning
Optimization
Root-mean-square errors
Search algorithms
ISSN:1816-093X, 1816-0948
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Accurate short-term power load forecasting is crucial for power system management and planning, significantly enhancing operational efficiency, optimizing dispatching, and conserving energy resources. Although the Long Short-Term Memory (LSTM) network has proven effective in power load forecasting, challenges remain in determining optimal hyperparameters and ensuring stable short-term load forecasting. To address these issues, this study proposes a combined model integrating the Adaptive Spiral Flight Sparrow Search Algorithm (ASFSSA) with the Adaptive Boosting (AdaBoost) ensemble learning algorithm for LSTM, aiming to achieve more precise short-term power load forecasting. Firstly, this study employs the ASFSSA algorithm to optimize the number of hidden layer neurons, the learning rate, and the Epochs of the Long Short-Term Memory (LSTM) model, aiming to determine the optimal hyperparameter combination. Subsequently, the AdaBoost algorithm adjusts weights, integrating several LSTM models into a robust predictor. This approach not only enhances the stability of short-term power load forecasting but also effectively reduces prediction errors. Experimental validation on power load datasets from Austria, Belgium, Hungary, and Luxembourg demonstrates that the ASFSS-LSTM-AdaBoost model excels in the evaluation metrics of Root Mean Square Error (RMSE), Coefficient of Determination (R2), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). In the Austrian electricity load dataset RMSE, MAE, MAPE, and R2 are 69.58 MW, 51.74 MW, 0.6%, and 0.996 respectively. The model exhibits higher accuracy in comparison with other algorithms and demonstrates the effectiveness and superiority of the proposed method in the field of short-term electricity load forecasting.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1816-093X
1816-0948