.png){kind=icon}
.png){kind=icon}
Return
Short-term Load Forecasting Method for Optimizing ETSformer based on the COA Algorithm
.png){kind=icon}
PRE
AI
DOI：10.2174/0123520965369669250411071016.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Introduction Accurate prediction of short-term power load is a key factor in ensuring the safe and economical operation of the power system. In order to enhance the accuracy of short-term load forecasting of the power system, this article proposes an innovative forecasting method that optimizes the ETSformer model based on the COA algorithm.Methods Firstly, the XGBoost algorithm is used to calculate the correlation between features and quantitatively analyze the specific impact of meteorological data on the load sequence. This helps screen out features with higher correlation, thereby improving prediction accuracy. Secondly, the filtered feature data is provided as input into the ETSformer model for training and prediction. By introducing the Exponential Smoothing Attention (ESA) mechanism and the Frequency Attention (FA) mechanism, efficient and accurate short-term load prediction is achieved. Finally, the COA algorithm is employed to optimize the hyperparameters of the ETSformer model, further enhancing its prediction accuracy.Results Comparative analysis with baseline models, such as ARIMA and BiLSTM, demonstrates that the model proposed in this article exhibits superior performance in short-term load forecasting tasks. Especially compared with the ARIMA model, this model reduced key indicators such as MAE, RMSE and MAPE by 44.6, 58.85, and 10.91%, respectively.Discussion The proposed method demonstrates strong performance in the field of short-term load forecasting. It addresses the limitations of traditional point forecasting methods, which are often difficult to apply effectively in real-world scenarios. Moreover, it overcomes the challenges of manual parameter tuning and the shortcomings of conventional optimization algorithms when dealing with high-dimensional data, thereby avoiding local optima. As a result, this method offers an efficient and practical solution for short-term load forecasting.Conclusion Our study demonstrates that the ETSformer model exhibits excellent performance in short-term load forecasting tasks. Integrating XGBoost for feature selection effectively mitigates the interference caused by features with low relevance to the prediction process. Furthermore, employing the COA algorithm to optimize model hyperparameters significantly enhances prediction accuracy. Compared to other baseline models, the proposed model achieves superior performance in short-term load forecasting tasks.
Keywords:
Short-term load forecasting
ETSformer
COA optimization
machine learning
hyperparameter optimization
neural network
Journal
R
IF:
0.5
Papers:
40
Citations:
0