钢铁企业蒸汽管网数据驱动建模与数据校正技术研究

Measurement data of high quality is the basis for optimization and reliable operation of the steam networks in Iron & Steel Enterprises. The researched methods for steam pipe network modeling, fault data monitoring, gross error detection and data reconciliation based on the model, are of great significance for reducing the comprehensive energy consumption of Iron & Steel Enterprises. Based on the large number of stored historical data and the experts' experience and a large number, applying Least Squares Support Vector Machine the steam networks model can be set up. By combining with the method of ICA-PCA and the model, the fault data monitoring problem of multi-variable system with abnormal distribution can be solved. With the evidence theory, spatial distance detection and clustering method, the gross errors and dynamic outliers should be detected. By combining with the robust estimation and the intelligent algorithm, the static and dynamic data reconciliation problem can be solved. So the data's quality of integrity, accuracy and consistency is overall improved. Through the research of this project, a set of theory and methods including data-driven modeling, fault data monitoring, gross error detection and data reconciliation for the steam networks in Iron& Steel Enterprises can be expected. The software of modeling and data rectification for the universal usage in complex industrial process will be derived as well. The research will promote the large amount of information stored in the enterprises playing an important role in the process production and management.

高质量的测量数据是钢铁企业蒸汽管网优化可靠运行的基础。本课题研究蒸汽管网建模及基于该模型的故障数据监控、显著误差检测和数据协调的方法，对降低钢铁企业的综合能耗有重要意义。 本项目基于专家经验和大量存贮的历史数据，采用最小二乘支持向量机蒸汽管网进行建模；结合ICA-PCA的方法和管网模型解决非正态分布的多变量系统故障数据监控问题；采用证据理论与空间距离聚类的方法检测显著误差和动态离群值；结合鲁棒估计与智能算法解决静态与动态数据协调问题。提高测量数据的完整性、准确性和一致性。 通过本项目的研究，可望得到关于钢铁企业蒸汽管网数据驱动建模、故障数据监控、显著误差检测和数据协调有关的理论与方法，编写在复杂工业过程中具有一家通用性的建模与数据校正软件。发挥企业信息化对生产和管理中的重要作用。

本项目研究了钢铁企业蒸汽管网多层次建模及基于该模型的故障数据监控、显著误差检测和数据协调的方法。.项目的主要成果包括：1)基于专家经验和历史数据，提出了多角度多层次建模的方法，尤其是多汽源蒸汽管网建模的方法；2) 提出基于数据驱动的不同状态PCA模型和分段线性PCA的故障数据监控方法；3)提出数据冗余性判定方法，利用空间冗余性结合TBM(Transferable Belief Model, TBM)证据理论进行显著误差判定的方法；4)提出通过损耗和间接法确定加权矩阵以改善数据协调效果的方法；5)提出蒸汽管网流量数据校正的整体方案。.通过本项目的研究，形成了钢铁企业蒸汽管网数据驱动建模、故障数据监控、显著误差检测和数据协调有关的理论与方法，编写了复杂工业过程中具有一家通用性的建模与数据校正软件。对发挥企业信息化在生产管理中的作用有重要意义。