Page 110 - Intelligent Digital Oil And Gas Fields

Data Filtering and Conditioning 79

application. Depending on the storage data duration and storage size, the
data historian software is set up to capture data with specific acquisition rates
(seconds, minute, hours); internally, the application performs a series of cal-
culations performing the following tasks:
• Transform electronic signal to physical data types (e.g., pressure, temper-
ature, rates, power, friction, choke settings, etc.).
• Convert raw data to specific units (e.g., psia, °F and Mscf/s, or bbl/day).
• Clean and filter bad data, data spikes, out-of-range data, and frozen data.
• Identify and replace missing data with rules, imputation, or
reconciliation.
• Down sample raw data in single data points, for example, 60s in 1min,
60min in 1h, etc., through time series averaging.
• Summarize high-frequency data to statistical values over lower
frequencies.
• Aggregate missing data or replace information for misleading data
through statistical interpolation.

3.2 BASIC SYSTEM FOR CLEANSING, FILTERING,
ALERTING, AND CONDITIONING

In a real-time process, data quality is a common and persistent issue,
and it takes time to repair or to replace the information. Typical problems
include the following.
• Missing data: Most of the time this occurs when connection is lost
between SCADA and RTU systems. This generally happens in hostile
environments, with extremely low or high temperatures, or winds at
high velocity. The lost signal appears as a gap in the data (null value)
or as a “frozen” data point (flat line).
• Data out of range: Occasionally signal errors can occur in remote systems
or metering systems, such as non-calibrated meters equipment like ori-
fice gas meters or turbine meters, which result in values that are out of
range of the reasonable or acceptable range.
• Frozen data occur when a data value is repeated several times due to mal-
function of electrical equipment or no transmission. To avoid null or
zero values, the system automatically repeats the previous value. Most
of the time this is an issue related to the programming language of
PLC or RTU devices.
• Data spike is a natural error signal that is outside the manufacturer’s
tolerance; it is random, occurs infrequently, and is typically caused by