Using Python to Filter and Calculate Average Data in OSIsoft PI with AFSDK

Using Python to Filter and Calculate Average Data in OSIsoft PI with AFSDK

Working with OSIsoft PI for data analytics is a common task in industries requiring time-series data management. One essential activity in this context is computing averages across data points while applying specific filters, such as removing negative values.

In this guide, I’ll walk you through the process of calculating average daily data for multiple PI tags in Python, excluding any negative data points, using the power of AFSDK (Asset Framework Software Development Kit). The challenge is that the direct usage of FilteredSummaries for a PIPointList in the code may not function as expected. Here’s how you can address it by leveraging AF attributes.

Understanding the Functionality

When working with the PI System, a common requirement is to filter out certain values (e.g., negative values) before calculating summaries like averages. Although there is a syntax for filtering, it may not apply directly to a list of PI Points using PIPointList. Hence, an alternative approach using AF attributes is necessary.

Approach Using AFAttributes

To calculate an average while excluding negative values, we need to utilize AF Attributes that allow us to access the FilteredSummaries method properly. Here's how to do it:

1. Generate AFAttributes: Create AFAttributes from the existing PIPoints. These attributes should belong to an AFAttributeList instead of a PIPointList.

2. Dynamically Create and Use Attributes: Leverage the AFSDK library to dynamically build AFAttributes that aren’t tied to a base element, enabling the use of a self-referential filtering syntax.

3. Filter and Compute: Use the FilteredSummaries method with the necessary filtering logic for the AFAttributes, which allows proper filtering and calculation of the average.

Step-by-Step Code Solution

Below is a modified version of the approach to tackle this problem using Python and AFSDK:

from OSIsoft.AF import AFAttributeList
from OSIsoft.AF.Data import *
from OSIsoft.AF.Asset import AFAttribute, AFAttributeTemplate
from OSIsoft.AF.Time import AFTimeSpan, AFTime
from OSIsoft.AF.PI import PIPoint

# Assume piPoints_list is a list of PIPoint objects
# time and timespan are defined as per the requirement

attributes = []
for point in piPoints_list:
    attr = AFAttribute(PIPoint\_(point))
    attributes.append(attr)

attribute_list = AFAttributeList(attributes)

time_range_start = AFTime('2023-01-01T00:00:00Z')  # example start time

timeSpan = AFTimeSpan(Duration=86400)  # 24 hours span for daily data

results = attribute_list.FilteredSummaries(
    time_range_start, timeSpan,
    ". > 0",
    AFSummaryTypes.Average,
    AFCalculationBasis.TimeWeighted,
    AFSampleType.ExpressionRecordedValues,
    AFTimeSpan.Parse("10m"),
    AFTimestampCalculation.EarliestTime
)

Conclusion

By embracing AF Attributes for filtering in combination with Python, this approach enables robust data manipulation and filtering on PIPoints, akin to advanced calculations possible in tools like PI DataLink. Dynamically creating non-element AFAttributes allows for seamless calculation of filtered summaries. This solution can be extended to include more complex filtering criteria as required by different industrial scenarios.

Remember, mastering such techniques improves your efficiency in managing large-scale PI System data operations, aiding in comprehensive data analysis and insightful decision-making.