Title: Short-term load forecasting with bidirectional LSTM-attention based on the sparrow search optimisation algorithm

Authors: Jiahao Wen; Zhijian Wang

Addresses: School of Information Science, Guangdong University of Finance and Economics, Guangzhou, China ' School of Information Science, Guangdong University of Finance and Economics, Guangzhou, China

Abstract: Aiming at the complexity and diversity of short-term power load data, a bidirectional long short-term memory (BILSTM) prediction model based on attention was proposed for the pretreatment collected data, and the different kinds of data were divided to obtain a training set and test set. The BILSTM layer was used for modelling to enable the extraction of the internal dynamic change rules of features and reduce the loss of historical information. An attention mechanism was used to give different weights to the implied BILSTM states, which enhanced the influence of important information. The sparrow search (SS) algorithm was used to optimise the hyperparameter selection process of the model. The test results showed that the performance of the proposed method was better than that of the traditional prediction model, and the root mean square errors (RMSEs) decreased by (1.18, 1.09, 0.60, 0.54) and (2.11, 0.45, 0.21, 0.11) on different datasets.

Keywords: short-term load prediction; sparrow search algorithm; neural network; weight assignment; attention mechanism.

DOI: 10.1504/IJCSE.2023.129154

International Journal of Computational Science and Engineering, 2023 Vol.26 No.1, pp.20 - 27

Received: 28 Feb 2022
Received in revised form: 21 Apr 2022
Accepted: 27 Apr 2022
Published online: 23 Feb 2023 *