作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2022, Vol. 48 ›› Issue (11): 69-76. doi: 10.19678/j.issn.1000-3428.0063153

• 人工智能与模式识别 • 上一篇    下一篇

基于TLBGA-GRU神经网络的短期负荷预测

吴铁洲1,2, 邹智1,2, 姜奔1, 张晓星1   

  1. 1. 湖北工业大学 电气与电子工程学院, 武汉 430063;
    2. 湖北工业大学 太阳能高效利用及储能运行控制湖北省重点实验室, 武汉 430063
  • 收稿日期:2021-11-05 修回日期:2021-12-22 发布日期:2022-01-12
  • 作者简介:吴铁洲(1966—),男,教授、博士,主研方向为电力系统大数据挖掘、储能关键技术、系统集成;邹智、姜奔,硕士研究生;张晓星,教授、博士。
  • 基金资助:
    国家自然科学基金(51677058)。

Short-Term Load Forecasting Based on TLBGA-GRU Neural Network

WU Tiezhou1,2, ZOU Zhi1,2, JIANG Ben1, ZHANG Xiaoxing1   

  1. 1. School of Electrical and Electronic Engineering, Hubei University of Technology, Wuhan 430063, China;
    2. Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System, Hubei University of Technology, Wuhan 430063, China
  • Received:2021-11-05 Revised:2021-12-22 Published:2022-01-12

摘要: 短期负荷预测在电网调度安排和电力市场交易中发挥着重要作用,预测精度高,有利于提高发电设备的利用率和经济调度的有效性。为充分挖掘负荷数据中时序性特征的联系,解决神经网络中由超参数的随机选取导致的预测精度下降问题,提出一种基于教与学的遗传算法(TLBGA)和门控循环单元(GRU)神经网络的短期负荷预测方法。利用灰色关联分析法对原始数据进行相关度分析,剔除冗余特征,使输入与输出保持较好的映射关系,在遗传算法中加入一种基于教与学优化的新型变异算子,用于防止其出现早熟收敛问题,从而提高解的质量。在此基础上,运用改进后的TLBGA算法对GRU神经网络模型进行超参数寻优,更新GRU的模型超参数并使其性能达到最佳状态,以提高负荷预测的精度。对欧洲某地区的电力负荷数据集和美国PJM电力市场公开负荷数据集进行预测,结果表明,该方法的预测精度分别达到了97.1%和97.2%,相比反向传播神经网络、循环神经网络及GRU神经网络模型,具有更高的预测精度。

关键词: 负荷预测, 变异算子, 遗传算法, 门控循环神经网络, 超参数寻优

Abstract: Short-term load forecasting plays an important role in the power grid dispatching arrangement and power market transaction.The higher the forecasting accuracy is, the more beneficial it is to improve the utilization rate of power generation equipment and effectiveness of economic dispatching.To fully excavate the connection of time series features in load data and solve the problem of a decrease in the prediction accuracy owing to the random selection of super-parameters in a neural network, this paper proposes a Teaching-Learning-Based Genetic Algorithm(TLBGA), and Gates Recurrent Units(GRU) neural network for short-term load forecasting.First, the gray correlation method is used to analyze the correlation degree of the original data and eliminate the redundant features such that the input and output can maintain a good mapping relationship.A new mutation operator based on teaching and learning optimization is added to the GA algorithm to prevent premature convergence and improve the quality of solution.Then, the improved TLBGA algorithm is used to optimize the GRU neural network model, update its hyperparameters, and enable its performance to reach the best state to improve the accuracy of its load prediction.The proposed method is used to predict the power load dataset of a certain region in Europe and the open load dataset of PJM power market in the United States.The results show that the prediction accuracy of this method can reach 97.1% and 97.2%, respectively.Compared with the BP neural network, Recurrent Neural Network (RNN), and GRU neural network model, the method proposed in this paper has higher prediction accuracy.

Key words: load forecasting, mutation operator, Genetic Algorithm(GA), gated Recurrent Neural Network(RNN), hyperparameter optimization

中图分类号: