The use of conventional process simulators is commonplace for system design and is growing in use for online monitoring and optimization applications. While these simulators are extremely useful, additional value can be extracted by combining simulator predictions with field inputs from measurement devices such as flowmeters, pressure and temperature sensors. The statistical nature of inputs (e.g., measurement uncertainty) are typically not considered in the forward calculations performed by the simulators and so may lead to erroneous results if the actual raw measurement is in error or biased.
A complementary modeling methodology is proposed to identify and correct measurement and process errors as an integral part of a robust simulation practice. The studied approach ensures best quality data for direct use in the process models and simulators for operations and process surveillance. From a design perspective, this approach also makes it possible to evaluate the impact of uncertainty of measured and unmeasured variables on CAPEX spend and optimize instrument / meter design.
In this work, an extended statistical approach to process simulation is examined using Data Validation and Reconciliation, (DVR). The DVR methodology is compared to conventional non-statistical, deterministic process simulators. A key difference is that DVR uses any measured variable (inlet, outlet, or in between measurements), including its uncertainty, in the modelled process as an input, where only inlet measurement values are used by traditional simulators to estimate the values of all other measured and unmeasured variables.
A walk through the DVR calculations and applications is done using several comparative case studies of a typical surface process facility. Examples are the simulation of commingled multistage oil and gas separation process, the validation of separators flowmeters and fluids samples, and the quantification of unmeasured variables along with their uncertainties. The studies demonstrate the added value from using redundancy from all available measurements in a process model based on the DVR method.
Single points and data streaming field cases highlight the dependency and complementing roles of traditional simulators, and data validation provided by the DVR methodology; it is shown how robust measurement management strategies can be developed based on DVR's effective surveillance capabilities. Moreover, the cases demonstrate how DVR-based capex and opex improvements are derived from effective hardware selection using cost versus measurement precision trade-offs, soft measurements substitutes, and from condition-based maintenance strategies.