Apascaritei, Bogdan:
ISBN 9783899598063

Error Analysis for Energy Process Simulations # Pb., 132 S., 87 Abb., 9 Tab.


power plant simulation
solvability
validation
L1-L2-norm (least sum of squared or absolute residuals)
occurrence matrix
Householder reflections
IRLS (iteratively re weighted least squares)

The error analysis for energy process simulations is of increasing interest due to the possibility of cycle optimization resulting in resource consumption minimization. The error analysis methods presented in this paper deal with system solvability and data validation. Nowadays system solvability verification is mainly done by checking to see if the number of equations is equal to the number of variables involved and perhaps by searching for under-determined subsystems by occurrence matrices. Based on the Householder decomposition method presented in this work, linear dependent equations causing singularities can be identified. Data validation is a challenging task mainly because of the nonlinear nature of the systems. In the domain of power plant simulation data validation is done by minimizing the sum of the residuals squared (L2-Norm) as stated in the VDI 2048 standard. The L2-Norm validation technique can give very wrong estimations if pinch-points are present in a system. This problem can be overcome by minimizing the sum of the absolute residuals (L1-Norm). The L1-Norm methods were considered to be too slow and not stable enough as opposite to the L2-Norm methods. Therefore IRLS - iteratively re weighted least squares-method is used. It is to be noticed that the L1-Norm fault diagnosis method is already used in electrical engineering. Based on the method especially developed in this work to overcome the pinch-point problem, the L1-Norm validation technique has become a viable alternative to the L2-Norm method. The performance of the method was validated against the L2- Norm on illustrative examples, underlining its advantages over the L2-Norm.

	^