发电技术 ›› 2025, Vol. 46 ›› Issue (6): 1212-1222.DOI: 10.12096/j.2096-4528.pgt.24127

• 发电及环境保护 • 上一篇    

基于改进型鲸鱼算法优化长短期记忆神经网络的过热汽温预测研究

冯磊华1, 聂超鹏1, 郭奇峰2, 杨锋3, 何金奇4, 毛玉成1   

  1. 1.长沙理工大学能源与动力工程学院,湖南省 长沙市 410114
    2.新疆其亚铝电有限公司,新疆维吾尔自治区 阜康市 831100
    3.华自科技股份有限公司,湖南省 长沙市 410006
    4.陕西高业能源科技有限公司,陕西省 西安市 710061
  • 收稿日期:2024-10-08 修回日期:2024-12-17 出版日期:2025-12-31 发布日期:2025-12-25
  • 作者简介:冯磊华(1980),女,博士,副教授,研究方向为发电过程建模与优化控制、智慧电厂,fengleihua80@126.com
    聂超鹏(1999),男,硕士研究生,研究方向为发电过程建模与优化控制,本文通信作者,ncp08288@163.com
    郭奇峰(1971),男,研究方向为复杂工业过程建模与仿真,elgqf@163.com
    杨锋(1979),男,高级工程师,研究方向为发电过程建模与智能控制,35770385@qq.com;
    何金奇(1980),男,高级工程师,研究方向为热工过程建模与智能控制,2308296920@qq.com
    毛玉成(2001),男,硕士研究生,研究方向为发电过程建模与优化控制,15898318905@163.com
  • 基金资助:
    湖南省自然科学基金项目(2018JJ3552);陕西高业能源科技有限公司横向项目(20220030139)

Superheated Steam Temperature Prediction Based on Long Short-Term Memory Neural Network Optimized by Novel Whale Optimization Algorithm

Leihua FENG1, Chaopeng NIE1, Qifeng GUO2, Feng YANG3, Jinqi HE4, Yucheng MAO1   

  1. 1.College of Energy and Power Engineering, Changsha University of Science and Technology, Changsha 410114, Hunan Province, China
    2.Xinjiang Qiya Aluminum Electric Co. , Ltd. , Fukang 831100, Xinjiang Uygur Autonomous Region, China
    3.HNAC Technology Co. , Ltd. , Changsha 410006, Hunan Province, China
    4.Shaanxi Gaoye Energy Technology Co. , Ltd. , Xi’an 710061, Shaanxi Province, China
  • Received:2024-10-08 Revised:2024-12-17 Published:2025-12-31 Online:2025-12-25
  • Supported by:
    Natural Science Foundation of Hunan Province(2018JJ3552);Horizontal Project of Shaanxi Gaoye Energy Technology Co., Ltd(20220030139)

摘要:

目的 火电机组在参与灵活调峰的过程中,机组频繁升降负荷,导致过热汽温波动较大,影响机组运行的安全性和经济性,准确预测过热汽温的变化趋势至关重要。为此,提出一种基于改进型鲸鱼算法(novel whale optimization algorithm,NWOA)优化长短期记忆(long short-term memory,LSTM)神经网络的过热汽温预测模型。 方法 使用主元分析法(principal component analysis,PCA)进行变量选择,以消除初始数据中的冗余变量;采用非线性收敛因子调整策略、自适应权重策略和动态螺旋更新策略对鲸鱼算法(whale optimization algorithm,WOA)进行改进,以提高算法的寻优精度及全局寻优能力;针对新疆某电厂直流锅炉建立过热汽温预测的LSTM模型,并利用改进的鲸鱼算法对该模型进行参数寻优,以解决其最优超参数组合难以选取的问题;运用电厂数据进行预测仿真。 结果 改进后的算法预测精度更高,在机组升降负荷期间,均可以较好预测到过热汽温的变化趋势,模型的平均绝对误差及均方根误差相比于改进前模型分别降低了7.02%和6.22%。 结论 该改进算法有效提高了过热汽温预测模型的精度,为过热汽温控制系统的优化设计提供了基础。

关键词: 碳达峰, 碳中和, 调峰, 过热汽温预测, 长短期记忆神经网络, 改进型鲸鱼算法, 主元分析法

Abstract:

Objectives In the process of participating in flexible peak shaving, thermal power units frequently increase and decrease load, leading to significant fluctuations in superheated steam temperature, which affects the operational safety and economic performance of the units. Accurately predicting the change trend of superheated steam temperature is crucial. Therefore, a superheated steam temperature prediction model based on long short-term memory (LSTM) neural network optimized by the novel whale optimization algorithm (NWOA) is proposed. Methods Principal component analysis (PCA) is used for variable selection to eliminate redundant variables in the initial data. The whale optimization algorithm (WOA) is improved by nonlinear convergence factor adjustment strategy, adaptive weighting strategy, and dynamic spiral update strategy to enhance the accuracy and global search ability of the algorithm. An LSTM model for superheated steam temperature prediction is established for a direct current boiler in a power plant in Xinjiang. The improved whale algorithm is used to optimize the parameters of this model, addressing the challenge of selecting the optimal hyperparameter combination. The prediction simulation is carried out using the power plant data. Results The improved algorithm demonstrates higher prediction accuracy. It can better predict the change trends of superheated steam temperature during periods of load increase and decrease. The mean absolute error and root mean square error of the model are reduced by 7.02% and 6.22%, respectively, compared to the model before improvement. Conclusions The improved algorithm effectively enhances the accuracy of the superheated steam temperature prediction model, providing a basis for the optimal design of the superheated steam temperature control system.

Key words: carbon peak, carbon neutrality, peak shaving, superheated steam temperature prediction, long short-term memory neural network, novel whale optimization algorithm, principal component analysis

中图分类号: