| 摘要 |
The present invention is a system and method for improved processing of flow data signals to provide output flow data signals which better approximate the true flows being measured. The general approach on which the invention is based is commonly known as flow data reconciliation. A dynamic simulation is run in parallel with a flow data reconciliation, and model predicted flows are used to complete balances wherever sensor measurements are unavailable. Model predicted compositions are also used to allow the computation of stream enthalpies so that enthalpy balances can be used in the reconciliation. The use of model predicted values of changes in inventory allow dynamic material balances to be used, making the reconciliation much more effective for processes with slow dynamic behavior. Weighting factors on sensors, computed based on the current value of the error of the sensor, reduce the impact of sensors with high errors. A multiplicative correction factor distributes corrections more evenly among streams with large and small flows. Statistical behavior of historical performance of sensors is used to reduce the adverse impact of rapid shifts in sensor accuracy and to detect unconverged reconciliation solutions.
|
