独立成分分析方法在电站热力过程数据检验中的应用

Application of Independent Component Analysis on the Data Validation of Thermodynamic System in Power Plant

提出了基于独立成分分析的电站热力过程数据检验方法,通过对测量数据的特征提取,对原测量数据中的信号进行分离,再通过独立成分分析进行信号复原,实现测量数据的重构。对独立成分分析算法、成分数目选择和初始化、检验阈值进行了讨论,给出了不良数据检验的流程,该方法计算量不大,可以用于在线数据检测。选用300MW机组热力系统21个实际测量参数作为算例,表明提出的数据检验方法能有效地检测到给水温度测量数据中的不良值,并可以给出可信的重构数据。对独立成分的信号分离能力进一步分析验证,表明了独立成分对测量数据中信息的表达能力。

A data validation method based on the independent component analysis （ICA） was proposed, which was applied to the thermodynamic system in power plant. The data detection and reconstruction can be implemented on the base of the separating of independent source signals from the measured data. The ICA algorithm as well as the IC selection, initialization and thresholds were analyzed. Then the framework of the method was presented, which can be used for online data validation since it has less computing complex. Then, 21 data sets from a 300MW unit were validated by the proposed method. The results revealed its validity to the detection of bad data. The ability of signal separating of these ICs was also analyzed.