PREDICTION METHOD FOR MONITORING PERFORMANCE OF POWER PLANT MEASURING INSTRUMENT

摘要

PPROBLEM TO BE SOLVED: To provide a prediction method for monitoring performance of power plant measuring instruments, wherein compared to an existing Kernel regression method, calculation accuracy of a prediction value is improved. PSOLUTION: Measuring instrument signals are displayed in a matrix, normalized, and separated into those for training, optimization, and test. A principal component is extracted. An optimal constant of an SVR model is obtained by a reaction analysis surface method using data for optimization. A training model is generated using the optimal constant. De-normalization is performed by de-normalizing the given normalization output into the original range to obtain a predicted value of a variable. Therefore, compared to an existing Kernel regression method, accuracy for calculating a prediction value is improved. PCOPYRIGHT: (C)2011,JPO&INPIT