Advances in PI System Streaming Analytics and Other External Calculation Engines

Similar documents
Transcription:

Advances in PI System Streaming Analytics and Other External Calculation Engines Stephen Kwan, OSIsoft Product Manager Tim Choo, MATLAB Production Server Product Manager 1

Goals and Objectives Use PI System as the data infrastructure Enable users to execute advanced streaming calculations Execution based on new events into the PI System or by clock Encourage reuse of existing domain expertise Support different personas Engineers, designers, users, etc. Retain ease of use, feature set and existing workflow 2

Advanced Streaming Calculations Asset Analytics released in 2014 PE replacement Leverages AF and PI System Configuration experience Robust engine with scheduler Widely used, but users want more Advanced calculations More flexibility Retain ease of use 3

Integration with 3 rd Party Calc Engines Calc engines for computational mathematics exist Take advantage of existing applications MATLAB is the first 3 rd party integration with Asset Analytics MATLAB is used by millions of engineers and scientists Many PI users have existing MATLAB code Collaboration between MathWorks and OSIsoft Mutual feedback for improvements Faster time to market 4

Personas: Streaming vs. Adhoc Adhoc investigation and analytics development Online, scheduled, streaming analytics 5

Integration with MATLAB Production Server Analytics Development HTTP/HTTPS MATLAB Production Server MATLAB Compiler SDK MATLAB Package Code / test Requirements PI Asset Framework 2018 MATLAB Production Server R2018a MATLAB, MATLAB Compiler and MATLAB Compiler SDK R2018a 6

OSIsoft Headquarters Leverage PI System to support the facility Collect data from Building Management System (BMS) Operational excellence Energy management Optimize energy usage Anomaly detection Single pane of glass How can we predict Energy Usage? 7

MATLAB and MATLAB Production Server Tim Choo MATLAB Production Server Product Manager

MathWorks is the leading provider of technical computing software Founded in 1984 Revenues ~$1B in 2017 ~4000 employees worldwide More than 2 million users in 175+ countries 95% of technical support calls reach an engineer with an advanced technical degree in under 30 seconds MATLAB SIMULINK Technical computing Simulation and model based design 9

Why choose MATLAB? MATLAB lets you focus on solving your problems Productive environment tuned for engineering and scientific work Ready to use with toolboxes that work out of the box Ready to run on production systems without rewriting code Reliable entrusted to send a spacecraft to Pluto, create certified code for medical devices Execution speed with optimized code that leverages GPUs, clusters, and clouds 10

Railway Systems Automotive Aeronautics Retail Off-highway vehicles Industrial Automation Data Analytics Finance Internet Logistics Oil & Gas Clean Energy Medical Devices Healthcare Management 11

Predictive Maintenance Toolbox New Design and test condition monitoring and predictive maintenance algorithms Feature extraction for designing condition indicators Machine learning and time-series models for remaining useful life (RUL) estimation 12

Analyzing HVAC data with MATLAB follows four basic steps 13

1 Access and Explore Data 2 Preprocess Data Its easy to explore, clean, and preprocess data with MATLAB 14

3 Develop Predictive Models Its easy to build a predictive model in MATLAB Use apps to help understand methods and workflows 15

3 Develop Predictive Models Its easy to try different models on the same data Neural Networks Time series models (ARIMA, GARCH,..) 16

3 Develop Predictive Models MATLAB s Regression Learner App lets you train with multiple algorithms in parallel 17

4 Integrate with Production Systems Deploy your predictive model as a reliable and scalable service with MATLAB Production Server MATLAB Compiler SDK Analytics Development MATLAB Deploy Package Code / test Access Integrate Data sources MATLAB Production Server Worker processes Request Broker Enterprise Application < > Mobile / Web Application Scale and secure 3 rd party dashboard 18

4 Integrate with Production Systems The Deployment Tool with Compiler SDK makes it easy to package and deploy your predictive model 19

4 Integrate with Production Systems MATLAB Production Server operationalizes your predictive model as a scalable and reliable service Predictive Model MATLAB Production Server AF with Asset Analytics Worker processes Request Broker Call analytic functions using REST API 20

Enhanced Data handling Complex Data Types Support MATLAB data needs Asset Analytics mostly handled single values only Introduce support for multiple values as arrays in analyses Configuration UI improvements Enhance existing functions for single values and arrays PI data infrastructure = first class citizen 21

User Experience Retain existing user experience External codes are treated as Expression functions Authoring, testing, preview, scheduling, backfill, recalculation, etc. remain the same 22

Data Retrieval and Array Operation Functions 23

Example Use Cases I Of need these the data last values, last 50 for archive 50 the I want archived last values, day to apply values in I 30 only a second transform want increments ones to > them 185

DEMO 25

26

EBSILON Heat Balance Analysis Single Asset Model Full Plant Model 27

Integration with EBSILON via enegrid Coming soon Asset Framework Asset Analytics HTTPS https://partners.osisoft.com/directory/partner/463286/vtu-energy-gmbh 28

Do more with your PI data Real time asset performance calculations Closed balances of mass and energy Calculate parameters that are not measured Data reconciliation via mass and energy balances What-if calculations Max. capacity Cost of next MWh Power/steam reserve Etc. 100+ component library Coming soon https://partners.osisoft.com/directory/partne r/260707/steag-energy-services-gmbh 29

Conclusions Asset analytics integration with 3 rd party calculation engines can help solve new problems Full use of your PI data infrastructure Execute your custom functions online in a streaming fashion Functions are treated just like any built in functions Almost zero learning curve for existing users Support different use cases Adhoc investigation vs. streaming calculations 30

More Details and Q/A with Developers 14:30 16:15 Room 117, P1 Level HowTo: Streaming Calculations with the PI System and MATLAB and other Computation Engines Abstract: In this live How-To, we will demonstrate the native integration between asset analytics and MATLAB Production Server introduced with AF 2018 enabling you to execute your custom functions in a streaming fashion leveraging the PI System data infrastructure. In addition, integration with other computation engines will also be demonstrated. 31

Stephen Kwan skwan@osisoft.com Product Manager OSIsoft, LLC Tim Choo Tim.choo@mathworks.com Product Manager MathWorks, Inc. 32

Questions? Please wait for the microphone State your name & company Please rate this session in the mobile app! Search OSIsoft in your app store 33

34