基于神经模块网络的汽轮机发电功率预测方法研究  

Research on Prediction Method of Steam Turbine Power Generation Based on Neural Module Network

作　　者：董艳妮 陈煦 张桁恺 王慧召 Dong Yanni;Chen Xu;Zhang Hengkai;Wang Huizhao(CHN Energy(Quanzhou)Thermal Power Co.,Ltd.,Quanzhou,Fujian 362801)

机构地区：[1]福建省泉州市泉港区国能(泉州)热电有限公司,福建泉州362801

出　　处：《现代工程科技》2024年第16期45-48,共4页Modern Engineering Technology

摘　　要：汽轮机对现代工业生产的稳定性和效率起着至关重要的推动作用。目前一些汽轮机发电功率预测方法存在建模能力不足、泛化能力不强等问题。文章聚焦于应用神经模块网络,以提升工业汽轮机发电功率的预测精度。通过将神经网络分解为专注于特定任务的模块,从而能够更有效地捕捉汽轮机系统中复杂的非线性关系。在汽轮机发电系统中,该网络自动学习了运行状态的特征,实现了对功率变化的准确预测。该方法的关键优势在于神经模块网络的模块化结构,提高了对系统行为的可解释性,并降低了手动提取特征的负担,为实现可持续、高效的发电过程奠定了坚实基础。Steam turbines play a crucial role in promoting the stability and efficiency of modern industrial production.At present,some steam turbine prediction methods have problems such as insufficient modeling ability and weak generalization ability.Research focuses on applying neural module networks to improve the prediction accuracy of industrial steam turbine power generation.By decomposing neural networks into modules focused on specific tasks,modular neural networks can more effectively capture complex nonlinear relationships in steam turbine systems.In the steam turbine power generation system,the network successfully automatically learned the characteristics of the operating state and achieved accurate prediction of power changes.The key advantage of this method lies in the modular structure of the neural module network,which improves the interpretability of system behavior and reduces the burden of manually extracting features,laying a solid foundation for achieving sustainable and efficient power generation processes.

关 键 词：神经模块网络 发电功率预测 特征表示 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TK472[自动化与计算机技术—计算机科学与技术]