Model Creation
While planning a plant, comprehensive stationary calculations are carried out with the simulation programs. The prerequisite here is that for all components the specification data and, if needed, the characteristic lines are determined or adjusted to the results of the component design calculations. It stands to reason to use the created computing model also for the check calculation or for plant monitoring.
An essential prerequisite for the data validation is a valid model. For this reason, the process engineering of the power plant should completely be modelled as auxiliary conditions based on the equations. Approximate relationships (e.g. steam cone law) can be included in the error calculation in this way, in which the formal result of the calculation is transferred as estimated value to a pseudo measurement point and the specification of a suitable confidence interval. The measurement value can be filled from a preceding simulation calculation by using the faulty relationship. Further, the highest possible measurement value redundancy must be given. The degree of freedom, which results from the difference of the number of the necessary measurement points for a simulation run and the actual number of measurement points included, should, therefore, be as high as possible.
The data validation adjusts the given measurement values to the model i.e. the deviations between the reality and the calculation model (respective measurement value and validated value) are indicated as measurement value errors. The total of the error squares is thereby kept to a minimum. Contrary to the measurement value equations, the model equations are fulfilled strictly.
While evaluating the measurement errors, therefore, it must always be noted that basically even model errors (errors in the topology, in process-engineering equations, in the specification data of the components) can be contained in them. To exclude these errors made by mistake, the underlying model must be subject to an exact checking in the simulation itself, before the raw data are analyzed (see error analysis). The quality of the model can be tracked depending upon the time regarding the individual measurement values at the changes of the evaluation criteria as per the inequation 141 VDI 2048 sheet 1 and for the complete model at the Chi^2-test ratio (the relative total error square divided by the value of the Fisher-distribution as per the inequation 141 VDI 2048 sheet 1).
If a deterioration is determined, then the possible reasons could be changes in the component properties (specification data) or else errors appearing in recording the measurement errors. Accordingly, an adaptation of the model or else a processing of the measurement points must be done.
Model Checking
A successful simulation design calculation must be present. The validation is normally done in the off-design calculation mode. No non-trivial double naming (trivial double naming on the logical lines can be ignored) may be present after a simulation calculation in its error analysis, as otherwise the data validation can be hampered. A double naming represents an over-determination of the simulation equation system and is permitted only for the flow rate in closed cycles; however, it must get removed through a splitting of the cycle (insertion of a flow splitter).
Double naming in pressure or specific enthalpy indicate an over-determination in the specifications for the simulation and must be thoroughly eliminated. Closed loops without pressure gradients are also to be avoided. While mixing, it must be noted tat the auxiliary line connected via a throttle lies at a higher pressure level, which can be achieved by integrating a pipeline with pressure loss in the main line.
For the validation, it is recommended to deactivate the controllers in the model, because a fixed value regulation of a validation would possibly run in the opposite direction.
Setting the validation parameters
The method, according to which the validation is to be done, must be decided and also the calculation parameters, which are to be made the basis for this.
Specification of the measurement value parameters
Along with the measured value, the necessary additional settings (e.g. weights or the confidence intervals) are to be entered for each measurement point.
If correlations are present between the measurement values, these must also be entered.