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

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.