IIC Journal of Innovation 16th Edition | Page 55

Design and Implementation of a Digital Twin for Live Petroleum Production Optimization
Simulation Input :
The simulation schematic shown in Figure 10 is Graphic User Interface ( GUI ) representation of a simulation file of the commercial physics-based simulator , such as Ledaflow or Olga . Each GUI based case has an associated input file that can be broadly divided into :
1 . Static parameters : Inputs that are fixed for the entire run of the simulations , such as the well completion and design data , reservoir fluid properties , pipeline and separator properties .
2 . Dynamic parameters : Inputs that can vary as a function of time such as the gas injection rate , reservoir pressure , produced gas to liquid ratio and water cut , sales gas back pressure , other wells gas .
3 . A system has been set up to write simulation input files based on the parameters obtained from a queue of simulations stored on a No-Sql Database such as MongoDB or PostgreSQL . The architecture of this system is further elaborated in a subsequent section .
Simulation inputs are queued on the No-Sql Database based on the type of parameters as described in Table 1 below . The parameters described in the “ knowns ” section in Table 1 are directly recorded from field data . These parameters are updated in the simulation queue based on timely trends in field data . The value of these parameters is based on the exact operating range observed in the processed field data . The “ unknowns ” correspond to parameters whose values are difficult to measure yet have a significant sensitivity .
These may include static parameters such as the tubing friction factor , or , dynamic parameters which vary at a high rate such as reservoir pressure . Since the input values for these parameters are unknown , a wide range of possibilities within the bounds of physical guardrails are input for these parameters . The “ approximations ” column in Table 1 refers to parameters that have some sample data from the field , but not precise , live , or well-specific data . These parameters can be approximated within a smaller range because of their static nature and relative insensitivity .
The simulation queue consists of combinations of the known , unknown and approximate parameters . New simulations are added to the queue based on the rate of change of the field data . After a point of time , further simulation may not be necessary on a well , as historical simulation may have covered the operating range . The ranges of the unknown parameters also narrow down with time as the inverse model provides estimates based on history matching . The details of the inverse model are beyond the scope of this paper , and shall be elaborated in subsequent publications .
- 50 - March 2021