基于MSWOA改进Attention-BiGRU模型的电力负荷预测  

Power Load Forecasting Based on MSWOA Improved Attention-BiGRU Model

作　　者：王童 WANG Tong(School of Computer Science and Technology,Zhejiang Sci-Tech University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学计算机科学与技术学院,浙江杭州310018

出　　处：《软件导刊》2023年第10期84-89,共6页Software Guide

基　　金：激光与物质相互作用国家重点实验室开发基础研究课题(SKLLIM2113)。

摘　　要：为了提高短期电力负荷的预测精度,提出一种基于混合策略改进鲸鱼优化算法(MSWOA)的Attention-BiG-RU短期电力预测模型。该模型首先利用双向长短期记忆网络(BiGRU)对电力数据的时序特征信息进行双向提取,并引入Attention机制,根据提取信息的特点,对隐藏状态的信息赋予不同比重,以增大重要信息的影响。针对模型的参数选取问题,通过MSWOA算法自动选取神经网络的模型参数,并优化网络模型参数使预测效果最优。通过对电力负荷数据进行训练与预测,与BiGRU、Attention-BiGRU、鲸鱼优化算法(WOA)改进的Attention-BiGRU等模型的预测结果进行对比。实验结果表明,该优化模型的预测精度达到98.829%,相较于传统的WOA,改进后的WOA对Atten-tion-BiGRU网络模型的优化有更好的效果,且相较于人工选择参数的神经网络模型具有更高的准确性和稳定性。In order to improve the accuracy of short-term power load forecasting,this paper proposes a mixed strategy based improved whale optimization algorithm(MSWOA)improved Attention-BiGRU short-term electric power forecasting model.The model first uses bidirectional gating recurrent unit(BiGRU)to extract the information of temporal characteristics of power data in both directions,and introduces Attention mechanism to give different weights to the information of hidden states according to the characteristics of extracted information to increase the influence of important information.To deal with the parameter selection problem of the model,the parameters of the neural network model are automatically selected by the MSWOA algorithm,and the parameters of the network model are optimized to make the optimal prediction effect.And by training and prediction of electric load data,the prediction results are compared with those of BiGRU,Attention-BiGRU,and the whale optimization algorithm(WOA)improved Attention-BiGRU models.The test results show that the prediction accuracy of the optimiza⁃tion model proposed in this paper reaches 98.829%,which has better results compared with the traditional WOA model for the improved Atten⁃tion-BiGRU network model,and has higher accuracy and stability compared with the neural network model with manually selected parameters.

关 键 词：电力预测 BiGRU 注意力机制 改进鲸鱼优化算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]