VMD-WSLSTM Load Prediction Model Based on Shapley Values

Bilin Shao,Yichuan Yan,Huibin Zeng

doi:10.3390/en15020487

Bilin Shao, Yichuan Yan + Show 1 more

Open Access

https://doi.org/10.3390/en15020487

Copy DOI

Journal: Energies	Publication Date: Jan 11, 2022
Citations: 1	License type: CC BY 4.0

Affiliation: Xi'an University of Architecture and Technology

Abstract

Accurate short-term load forecasting can ensure the safe operation of the grid. Decomposing load data into smooth components by decomposition algorithms is a common approach to address data volatility. However, each component of the decomposition must be modeled separately for prediction, which leads to overly complex models. To solve this problem, a VMD-WSLSTM load prediction model based on Shapley values is proposed in this paper. First, the Shapley value is used to select the optimal set of special features, and then the VMD decomposition method is used to decompose the original load into several smooth components. Finally, WSLSTM is used to predict each component. Unlike the traditional LSTM model, WSLSTM can simplify the prediction model and extract common features among the components by sharing the parameters among the components. In order to verify the effectiveness of the proposed model, several control groups were used for experiments. The results show that the proposed method has higher prediction accuracy and training speed compared with traditional prediction methods.

Highlights

With the continuous progress of social science and technology, the application of electric power is becoming increasingly extensive, and there are more and more factors affecting the electric load, which leads to the non-smoothness and complexity of the electric load
The components obtained from the decomposition were fed into the WSLSTM model for prediction, and the final results were obtained after superposition
In order to verify the effectiveness of the feature selection method proposed in this paper, three models were used for controlled experiments: the first model takes all features as input (FF), the second model uses the Pearson correlation coefficient method to select those with high correlation as the optimal feature set (PF), and the third model uses the Shapley value for model feature selection (SF)

Summary

Introduction

With the continuous progress of social science and technology, the application of electric power is becoming increasingly extensive, and there are more and more factors affecting the electric load, which leads to the non-smoothness and complexity of the electric load. Accurate prediction of power load data is beneficial to the relevant departments for policy making and power dispatching, and it is of great significance to the development of power systems. Determining how to accurately forecast the power load is a topic worthy of study. Prediction by artificial intelligence algorithms is a current research hotspot in the field of load prediction, and artificial intelligence algorithms are more suitable for nonlinear data, such as random forests [1,2], artificial neural networks [3,4], and support vector machines [5]. LSTM is studied and applied in many fields, such as load prediction [7], action recognition [8], and speech recognition [9]

Methods

Results

Conclusion