Why Data Modeling Using the CIM is Important to Big Data Analytics (14PESGM2447) Margaret Goodrich Director of Systems Engineering SISCO, Inc. margaret@sisconet.com
Topics Legacy use of data modeling for analytics The legacy process for accommodating multiple application models Impact of the legacy approach Understanding the use case The CIM based process for analytic data modeling Summary 2
Legacy Approach for Analytic Data Modeling Each group looks at its own application needs and develops a data model that is optimized for its own use: Only data needed for its application is considered. New data model elements are added as needed based on needs of individual applications. The Ad-Hoc Approach Data Store 3
Line Rating Application Control Area Corridor Ad Hoc Approach for Line Rating Analytic Line Segment 1 Line Segment 2 Ambient Temp Wind Speed Wind Direction Current A Line LineTemp Sag B Line LineTemp Sag Line Rating App 4 Data Store
Ad Hoc Approach for Remedial Action Schemes Remedial Action Application Corridor North-South Interconnect Line Trip RAS Generator Trip RAS Airport Substation Sydney Sub West Dam Sub East Wind Sub Line Status Current Margin Line Rating RAS Arming C-RAS App Data Store 5 Line Rating App 5
Ad Hoc Approach for Disturbance Analysis Disturbance Monitor App Control Area Airport Sub Sydney Sub East Wind Sub Bus Monitoring Battery Breakers Transformer Voltage Level 138KV 69KV DFR1 West Dam Sub Disturbance Monitor C-RAS App Line Rating App Data Store 6
7 Condition Based Maintenance Circuit Breakers Sydney Sub 69KV 138KV Ad Hoc Approach for CBM Analytics Breaker Q1A1 Breaker Q1A2 Breaker Q1A3 Breaker Q2B1 Breaker Q2B2 Last Operate NumOperations Transformers Sydney Sub West Wind Sub Disturbance Monitor C-RAS App Line Rating App 7 CBM Data Store
Impact of Ad Hoc Approach for Application Data Models Each Application has its own data model. Disturbance Monitor CBM Next App? How Many? Impact of crossorganizational integration and data sharing ignored. Questions: How many different models exist in a utility? How is data kept in synch? C-RAS App Data Store Line Rating App 8 8
How Does This Ad-Hoc Approach Happen? Misunderstanding the Enterprise Integration Needs Limiting Integration to the Use Case for a specific application or project Is this really the use case that should drive choices? Outage Analysis Outage Management Application A System A 9
What breaks the Ad-Hoc Approach? Change Addressing change becomes too difficult when each application uses its own incompatible data modeling: Business needs demand organizational changes and new levels of data sharing and integration. New technology must be addressed (e.g. renewables, DER, deregulation, etc. Result: Application rewrites, reintegration, project delays, barriers to data sharing. The Bigger the data, the more the negative impact will be of not using a consistent common data model. 10 10
CIM Helps Manage Change The model driven process captures change and creates incremental updates Existing Model Change: new, delete, modify Modeling Tools and Processes The individual applications can be updated and kept synchronized with each other. Incremental Update Model Store Incremental or Partial CIM-XML File or Updates from ESB 11 11
Model-Driven Data Using CIM User Requirements CIM is a Model that is flexible to accommodate: Extensions for non-standard business needs Eliminate the complexity of unused models Profiles are created based on use cases to address specific needs Instances created to relate existing data to the CIM Profile schema Use Cases Profile Schema for Data Templates Extensions CIM Model Model can be used to configure analytics. Analytics use models to access data eliminating custom tag name dependency. Instance File for Application Data MODEL and Data Store Application Data 12 12
Use Cases Rule! Use Cases defines the requirements needed to define data models, what data must be exchanged between which systems, how the data is used, who uses the data, why, what applications are needed, etc. Without a good understanding of the use cases an analytic design architecture cannot be developed. You don t need to define ALL use cases up front. Using a common model (CIM) as the starting point enables all analytics to use common model constructs instead of creating new models for each new use case (analytic). 13
Profile = Exchange Contract of CIM information Selection of which classes and attributes are of interest Selection of relationships (e.g. associations) are of interest (e.g. to create a containership ). Add extensions Make optional attributes/associations mandatory Why? Because unique use cases have a different needs! 14
Profiling is: needed to create a contract that represents what information is to be exchanged based on the requirements defined in the use case. typically a subset of the entire information model. used to generate messages as well as file definitions for analytic data modeling. 15
The process of profiling: Defining information to be exchanged based on a Use Case Step 1: Develop model Iteration Step 2: Decide on profile Proposed Standard Extensions XSD RDFS RDFS or OWL Step 3: Implementation: Create adapters /configuration Messages Files Databases 16
CIM Data Models Deliver Flexibility Multiple uses cases can be addressed with one profile. Multiple profiles can be supported for use cases that can t share a profile A disciplined modeling process with CIM provides models optimized for all applications Use Case Use Case Use Case CIM Profile Development IEC 61850 Other Models Profile 1 Profile 2 CIM Data Models 17 17
CIM Is The Only Choice for the Model-Driven Utility Internationally accepted IEC 61970 and IEC 61968 standards. Developing your own comprehensive utility data model to replace CIM will take many decades of effort. How many experts can your utility hire to design this from scratch? CIM is specifically designed to be adapted to fit the needs of individual utility use cases: Extensions Profiles/subsets Messages Integration Patterns New applications can extend independently yet share the existing models where needs overlap without breaking existing applications and integration. 18
SUMMARY Power system data models provide standardized context for data simplifying data management: Eliminating data source dependencies from analytics. Use of common semantic model between applications enables data sharing. Model management practical for large complex systems compared to tag management. CIM is an industry standard models that exists, has a defined process for adapting to individual needs and is ready to be used. 19
Thank You Questions/Discussion Margaret Goodrich Director of Systems Engineering SISCO, Inc. margarat@sisconet.com www.sisconet.com 20