Abstract/Summary
Calpine is one of North America’s largest power generation companies, with almost 80 generation facilities, including solar battery and geothermal plants. With AVEVA eDNA enterprise data management system deployed across Calpine's generation facilities, the sunsetting of the outdated platform provided both a daunting hurdle and a golden opportunity to modernize their enterprise data infrastructure to support predictive analytics and community data sharing solutions.

Having now successfully migrated almost half of their sites and over 300 years of data, the Calpine migration approach optimizes more than just the cumbersome data migration. Learn how calculations, reports, visualizations, and infrastructure improvements are addressed as part of this comprehensive migration strategy. This presentation discusses the struggles and successes of a complex migration and the roadmap for the future of Calpine's cloud-hosted centralized data infrastructure leveraging the AVEVA PI Data Infrastructure and AVEVA Predictive Analytics.

Challenge
With almost 80 plants running AVEVA eDNA as their local historian, Calpine knew the announcement to retire the antiquated system in mid-2025 would mean a foundational shift in their operational data infrastructure. This presentation tells the story of Calpine’s daunting migration challenge, some initial struggles, and how they aligned on a roadmap to achieve success.

• Almost 80 plants running AVEVA eDNA as an on-premise historian
  • Includes Battery Plants (some solar) & Geothermal (Geyser sites)
  • Averaging almost 20 years history per site
  • Over 500,000 tags
  • Over 100,000 calculated values
• The slow, manual process to export data files, edit formatting, and import data for migration was unmanageable at this scale and left uncertainty around import data accuracy.
• There was no out-of-the-box solution for converting calculations or displays to PI System equivalents.
• Each plant was unique, having their own asset models and many custom-built solutions developed on the eDNA platform. This lack of standards was difficult to support and maintain.
• The on-premise server architecture and eDNA applications for display clients and spreadsheet plug-ins were difficult to install and support.
• Many plants had to maintain a direct OPC connection between the plant systems and their customers for power data sharing.

Solution/Journey
Calpine decided to migrate on-premise plant historian instances and their archives to a cloud-hosted, centralized PI System. The first attempt at migration was performed on four (4) plants. The Calpine migration team ran the AVEVA eDNA data export utility, manually edited the output files for proper formatting, and then imported to PI Data Archive with a PI Interface for UFL. This migration process was slow, tedious, and prone to mistakes. More critically, there was no straightforward way to move eDNA calculations into the PI System and all calculated values had to be reverse engineered and manually recreated in PI AF Analytics. If they were to get through their entire fleet, drastic improvements had to me made.

At this point, Calpine engaged RoviSys, an AVEVA Endorsed System Integrator Partner, for input on how to more effectively address the migration needs. The team selected six (6) more plants as a pilot for the RoviSys proposed approach.

• RoviSys developed a custom application to parse eDNA data files to automatically format the contents and insert them into a PI Data Archive.
• A programmatic data validation process was included to ensure accuracy of migrated data.
• Automated scripts were developed to programmatically translate compatible calculations from eDNA to PI AF Analytics.
• A template-based PI AF Hierarchy was developed for each plant, including a calculation warehouse for the converted calculations.
  • This enabled standard client solutions on a common PI System Data Infrastructure (low-code/no-code).
• Template-based PI Vision displays and DataLink reports replace disparate eDNA client solutions
  • These web-based visualizations on a central PI Vision server reduced the complexity of distributed thick client installations.
• Moved customer data sharing from direct OPC connections with plant systems to leveraging PI Cloud Connect as an interim solution. Calpine in the planning stages of moving to AVEVA Data Hub for comprehensive data sharing community between customers and plants.
• The data infrastructure is feeding into AVEVA Predictive Analytics (PRiSM). Currently configured to use existing eDNA source and moving to leverage PI System upon completion of the migration.
  • PRiSM provides over 2,600 predictive models for the plants. (NOTE: Specific use cases can be added for presentation content)

Benefits/Roadmap
With significant improvements in the migration performance throughout the pilot sites, the team turned focus to Calpine’s additional ERCOT facilities to even further optimize the process. The confidence of the team enabled them to start doing multiple sites in parallel to accelerate the migration timeline even more. As a result of the parallel efforts, the current migration strategy and the roadmap ahead provides the following benefits and highlights:

• A small task team has begun prototyping an API-based migration utility in parallel. This solution programmatically connects to eDNA to further streamline the data migration process and provide a fully-automated, end-to-end data migration.
• The migration team has begun investigating RoviSys’ Vision Accelerator solution to programmatically convert eDNA displays to native PI Vision displays and further reduce the burden on Calpine’s internal resources from manually recreating valuable displays.
• The resulting IT infrastructure is highly simplified and much more secure.
  • Calpine now has one enterprise PI System feeding many internal applications from a common standardized data infrastructure.
• The ongoing data migration is dependable and efficient, with no manual data manipulation and significantly reduced opportunity for human error.
• Over 300 years of data and thousands of calculations have been successfully migrated to the PI System.
• A custom-built PItoPI Interface regulates data flow to prevent Production Server performance issues and delayed access to real-time data.
• The built-in data validation process provides certainty in the accuracy of the migrated data.
• Over 70% of calculations are now automatically converted to PI AF Analytics
  • Previously: ~2hrs/calc for manual recreation; Now: ~70% automatically recreated
    • ROI Time Savings: ~70% calcs count * 2hrs
• Template-based PI AF Infrastructure for sustainable, standardized growth
  • Moving towards a new automated ESG Reporting solution based on PI
  • PI AF hierarchy supports a standardized Compliance reporting solution
• Machine Learning Models for Gas Turbine Lifecycle Tracking (equivalent operating hours analysis) are running on standardized PI System data. The models provide monthly health summary indicators for critical assets.
• Over 2,600 AVEVA Predictive Analytics models are running today and moving towards leveraging data from the PI System once eDNA is decommissioned.
• Current customers on direct OPC connections are migrating to PI Cloud Connect and eventually AVEVA Data Hub for data sharing.

Measurable ROI
The calculation migration approach developed by RoviSys eliminated 70% of the manual effort to convert existing eDNA calculated values to an equivalent calculated value in the AVEVA PI System. This will save an estimated 140,000 development hours of effort to develop 70,000 calculations which will be automatically converted by the RoviSys migration solution.