Page 380 - Design of Simple and Robust Process Plants

9.5 Closed Loop Steady-state Optimization 367
The system's architecture is shown in detail in Figure 9.8, including all the ele-
ments. The description is written for a steady-state optimization. The modules,
which are always complemented with a decider, will be described in sequence.

9.5.2.1 Data analysis (DA) (Jordache and Narasimham, 1999)
In this module, all plant data necessary for the optimization are retrieved from the
basic control system, to be completed with status data from the model-based control
platform. These data are analyzed for corruptness and used for the determination of
process condition, and completed with a decider before transfer to the next sequen-
tial module. The data required are:

. Status data, to reflect (digital information):
± operational state of the plants (is it in an operational optimization step?).
± operational state of the model-based control platform (is it active?).
± operational state of the control loops (are loops closed and available for imple-
mentation of downloaded optimal set-points? ± these are the DOFs for the
optimization).
± data analysis requested, set by a timer in the executive.
. Process operational conditions for (analog information):
± Steady-state Representing Values (SRVs)for the selection of SRVs, see Sec-
tion 9.6.5. (see Krist et al., 1994 for steady-state detection);
± Data reconciliation.
± Parameter estimation and optimization.
± Feed and environmental conditions, such as outside temperature.
± Constraint conditions, such as maximum operational pressures and tempera-
tures, speed, environmental loads.
The first task of the data analysis module is to check the existence of any corrupt
data; no signals, or out of a defined range, the first pass on gross error detection by
signal analysis. Gross errors in measurements are detected at different levels:
. Instrument level by signal analysis.
. BC (Basic control)level by comparison of redundant measurements such as
tank levels.
. OO level by data reconciliation and parameter estimation criteria and the
determination of outliers compared to simulation results.
These gross errors are selected based on decision criteria which depend on the
specifics of the measurement and the process. Gross errors detection during data
reconciliation are described in Section 9.5.2.2.
The second task is to measure steady-state condition. The detection of steady state
is not limited to systems planned to operate as such. They may also apply to
dynamic optimizations to measure process parameters that have an impact on per-
formance, such as the aging or fouling conditions; transient operations often also
start from a steady-state condition. In fouling or aging systems the dynamic optimi-
zations implement gradually a new set-point over a long time period, and in fact
move in discrete steps from one steady-state condition to the next within an opti-