基于非线性块式递推偏最小二乘法的电站热力过程动态数据检验

Dynamic Data Validation of Thermodynamic Systems in Power Plants Based on Nonlinear Block-wise Recursive Partial Least Squares

数据检验与校正是保证热力过程测量数据准确性的重要手段,是火电机组监控系统安全可靠运行的重要前提。针对电站热力过程的特点,提出基于非线性块式递推偏最小二乘的动态过程数据检验方法。该方法采用基于样条变换偏鲁棒M-回归方法,将队列中原始测量数据的非线性关系转化为拟线性关系后,再进行加权块式递推偏最小二乘回归,实现了非线性动态系统的在线建模。将该方法应用于某电站热力系统非线性动态过程的数据检验中,结果表明所提出的数据检验方法不仅具有良好的鲁棒性,而且具有较强的追踪过程变化的能力,可满足非线性动态过程在线监测的需要,具有较高的实用价值。

Data validation is not only the important way to ensure the accuracy of the measurement data, but also the basis of reliable operation of the monitoring system in power plant. A dynamic data validation method based on nonlinear block-wise recursive partial least squares was proposed which is suitable for the characteristic of the data in thermodynamic system. The nonlinear relationship of raw data was transformed into quasi-linear by partial robust M-regression of splines （PRMS）. Then the weighted block-wise recursive partial least squares method was implemented to establish a dynamic model. The data sets from a power plant unit were validated by the proposed method. The results show that the presented data validation method have either a good robustness or a strong tracing ability when process change. It can validate the bad data effectively. It can meet the needs of online monitoring nonlinear dynamic process and has high practical value.