Design and Implementation of a Digital Twin for Live Petroleum Production Optimization
Fig . 4 : Template for Digitized well metadata .
DATA PROCESSING
The sensor data and metadata were recorded and digitized from various sources . Further the data processing steps subsequent to this includes :
● Mapping data
● Profiling
● Cleanup
● Transformation
● Labeling events
● Flagging changes
Mapping : As described in the previous section , Sensor data and metadata come in different formats and are stored in different formats . Similar devices from different providers or SCADA systems often have different naming schemes , these sensor tags need to be mapped to their appropriate devices and assets . After the name mapping the time series sensor data is mapped to its asset metadata by merging the two data sources by asset ID .
Data profiling : After mapping sensor data and metadata , sensor data will be going through an EDA ( Exploratory Data Analysis ) process which we call as data profiling step . In this step , we look into some specific properties of well ’ s sensor data to determine if it meets our model ’ s requirements and in result we can create a cohort of wells which have sufficient data properties to be used in our models . Those properties are including but not limited to Monitored Time , Lapse Time / Monitor Time ratio , Zeros ratio , and Frequency . Monitor Time quantifies how much
IIC Journal of Innovation - 43 -