Using Future Data to Predict your process - Future Data to Predict... · Using Future Data to Predict…

Download Using Future Data to Predict your process - Future Data to Predict... · Using Future Data to Predict…

Post on 07-Sep-2018

212 views

Category:

Documents

0 download

TRANSCRIPT

<ul><li><p>2015 OSIsoft TechCon </p><p>Using Future Data </p><p>to Predict your Process </p></li><li><p>2015 TechCon Session </p><p>2 | P a g e </p><p>OSIsoft, LLC </p><p>777 Davis St., Suite 250 </p><p>San Leandro, CA 94577 USA </p><p>Tel: (01) 510-297-5800 </p><p>Web: http://www.osisoft.com </p><p> 2015 by OSIsoft, LLC. All rights reserved. </p><p>No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or </p><p>by any means, mechanical, photocopying, recording, or otherwise, without the prior written permission </p><p>of OSIsoft, LLC. </p><p>OSIsoft, the OSIsoft logo and logotype, PI Analytics, PI ProcessBook, PI DataLink, ProcessPoint, PI Asset </p><p>Framework (PI AF), IT Monitor, MCN Health Monitor, PI System, PI ActiveView, PI ACE, PI AlarmView, PI </p><p>BatchView, PI Coresight, PI Data Services, PI Event Frames, PI Manual Logger, PI ProfileView, PI </p><p>WebParts, ProTRAQ, RLINK, RtAnalytics, RtBaseline, RtPortal, RtPM, RtReports and RtWebParts are all </p><p>trademarks of OSIsoft, LLC. All other trademarks or trade names used herein are the property of their </p><p>respective owners. </p><p>U.S. GOVERNMENT RIGHTS </p><p>Use, duplication or disclosure by the U.S. Government is subject to restrictions set forth in the OSIsoft, </p><p>LLC license agreement and as provided in DFARS 227.7202, DFARS 252.227-7013, FAR 12.212, FAR </p><p>52.227, as applicable. OSIsoft, LLC. </p><p>Published: May 8, 2015 </p></li><li><p>Using Future Data to Predict your Process </p><p>3 | P a g e </p><p>Contents Objectives ................................................................................................................................................. 4 </p><p>Prerequisites ............................................................................................................................................. 4 </p><p>Materials ................................................................................................................................................... 4 </p><p>Instructions ............................................................................................................................................... 4 </p><p>1. Introduction ...................................................................................................................................... 5 </p><p>2. Challenge ........................................................................................................................................... 7 </p><p>3. Use PI Coresight to visualize raw data .............................................................................................. 9 </p><p>4. Integrate future data in PI AF.......................................................................................................... 11 </p><p>4.1. Associate future tags with PI AF Attributes ............................................................................ 12 </p><p>4.2. Create asset-based analytics to leverage future data in PI AF................................................ 15 </p><p>5. Create an element relative PI Coresight display ............................................................................. 25 </p><p>6. Create a PI DataLink Report to Schedule Maintenance Teams (Time allowing) ............................ 26 </p><p>7. Create a PI ProcessBook Display to show anticipated production from solar panels (Time </p><p>allowing) .................................................................................................................................................. 33 </p><p>Annex A: List of PI AF substation parameters ......................................................................................... 42 </p><p>Annex B: PI AF Categories ....................................................................................................................... 44 </p><p>OSIsoft Virtual Learning Environment ........................................................................................................ 46 </p></li><li><p>2015 TechCon Session </p><p>4 | P a g e </p><p>Objectives </p><p>o Define future data in the PI System o List the ways future data can be injected into the PI System o Describe why PI AF is the best container for metadata o Configure AF templates that reference these future PI points o Create AF elements based on those templates o Create AF analyses that that shows delta between forecast and actual o Create PI Event Frames to capture the metadata around the results o Create a PI Coresight Display to show your data o Create a PI DataLink Report and a PI ProcessBook Display to show your data (optional time </p><p>permitting) </p><p>Prerequisites </p><p>o Some familiarity with PI Data Archive and PI Asset Framework o Some familiarity with the PI Visualization clients (PI Coresight, PI ProcessBook, PI DataLink) </p><p>Materials </p><p>Provided for you: </p><p>o PI Data Archive server with PI Points and data o PI Asset Framework with a base hierarchy o PI Clients installed: PI Coresight 2015, PI ProcessBook 2015, PI DataLink 2014 SP1 </p><p>Instructions </p><p> This lab consist of 5 exercises (2 optional). The exercises can be performed independently by using the CaliforniaWeather_Completed </p><p>database for sections 5,6, and 7 </p><p> For each exercise, general instructions are provided followed by step-by-step instructions. If you feel comfortable with the general instructions, go ahead and try to do the exercises on </p><p>your own. Dont hesitate to ask questions to the lab instructor if needed. </p><p> If you choose to follow the step-by-step instructions, complete each action marked by a red diamond symbol. </p></li><li><p>Using Future Data to Predict your Process </p><p>5 | P a g e </p><p>1. Introduction In this session, we will describe one way to use future data in the PI System. Lets start with a </p><p>common definition, in this case from Wikipedia: </p><p>Forecasting is the process of making statements about events whose actual outcomes (typically) have </p><p>not yet been observed. A commonplace example might be estimation of some variable of interest at </p><p>some specified future date. Prediction is a similar, but more general term. Both might refer to formal </p><p>statistical methods employing time series, cross-sectional or longitudinal data, or alternatively to less </p><p>formal judgmental methods. Usage can differ between areas of application: for example, in hydrology, </p><p>the terms "forecast" and "forecasting" are sometimes reserved for estimates of values at certain specific </p><p>future times, while the term "prediction" is used for more general estimates, such as the number of times </p><p>floods will occur over a long period.1 </p><p>And in the PI System, we define future data as time-series data that isn't measured and collected in </p><p>real-time from machine sensors, and will typically have a timestamp in the future. From its primary </p><p>function as a data historian, the PI Data Archive is designed to preserve all time-series data, which </p><p>includes future data. As a result, future or "non-measured" data will exist in the future but also in </p><p>the past (e.g., important forecasts kept for subsequent analysis, optimization, or auditing). </p><p>Future data is now handled natively by PI System 2015 including storage, analytics, data access, </p><p>reporting, and visualization and is most useful when compared or combined with real-time or </p><p>historical data. </p><p>Although the internals of future data in the PI System and the configuration and </p><p>administration tasks associated to it are not in the scope of this lab, lets take a quick look </p><p>at how future data can be stored into the PI System. </p><p>In PI System 2015, a new concept has been introduced: future PI Points. Those can be </p><p>created by setting the new future attribute to 1 at point creation. This attribute can be </p><p>set using the PI SMT Point Builder or the Excel PI Builder add-in and cannot be modified </p><p>after the point has been created. </p><p>Tags with the attribute set to 0 referred to as historical tags are identical to the tags in all previous </p><p>versions and will reject data with timestamps more than 10 minutes into the future. </p><p> 1 http://en.wikipedia.org/wiki/Forecasting </p></li><li><p>2015 TechCon Session </p><p>6 | P a g e </p><p>The picture below shows a couple ways to create future tags </p><p>Generally speaking, future tags should be used when storing data that is not collected sequentially in </p><p>chronological order. For example, process or operational data should be kept in historical tags because it </p><p>is measured and collected in real time. On the other hand, forecasts or any form of predictive data over </p><p>an arbitrary time range are perfectly suited for future tags. Data for all future tags is stored in </p><p>completely separate archive files. Future data never moves to the historical archive files even after the </p><p>future data ages into the past. The archive files for future tags will be created automatically as data is </p><p>written to the tags. All future archive files will be created in the directory specified during installation, </p><p>and this location can be changed any time afterwards using the new Archives plug-in included with PI </p><p>SMT 2015. </p><p>Writing data to future tags is just like writing data to historical tags just with different timestamps. </p><p>Thus, in general, you should be able to use all your favorite programs and tools for writing data. The </p><p>allowed timestamp range is January 1970 to January 2038. Some common options are the following: </p><p>o Standard OSIsoft interfaces like the PI UFL, PI RDBMS, or PI HTML Interfaces o Configured analyses and calculations with Asset Analytics, starting with PI Server 2015 Beta 3 o A custom application using PI AF SDK, PI Web API, or PI OLEDB Enterprise o PI SMT Archive Editor </p><p>The ways in which you use future data are only as limited as your need and creativity. In this example </p><p>we will use the analytics capability of the PI Server 2015 to determine forecast error from actual, create </p><p>more forecast values, etc. </p><p>For more information, please watch our webinar titled, Capturing Future Data with PI Server 2015 BETA, </p><p>available on our website (www.osisoft.com) under Resources Webinars Webinars on demand </p><p>(http://www.osisoft.com/Templates/item-abstract.aspx?id=11985) </p></li><li><p>Using Future Data to Predict your Process </p><p>7 | P a g e </p><p>2. Challenge You have been hired as a consultant by Sunshine Corp, LLC to help them take advantage of the new </p><p>future data features of the PI System 2015. </p><p>On your first day, youre provided with weather data from weather stations in the following airports, all </p><p>in the beautiful state of California: </p><p>o Bakersfield Meadows Field Airport (KBFL) o Los Angeles International Airport (KLAX) o Palo Alto Airport (KPAO) o Porterville Municipal Airport (KPTV) o San Francisco International Airport (KSFIO) o San Jose International Airport (KSJC) o Oakland International Airport (KOAK) </p><p>For each one of the above airports, the following PI Points are available </p><p>Real-time data tags </p><p>Future data tags </p><p> NWS__PressureIn NWS__RelativeHumidity NWS__Temp NWS__Visibility NWS__Weather NWS__WindDirection NWS__WindSpeed </p><p> .ChanceOfFog.Forecast .ChanceOfFrost.Forecast .ChanceOfHighTemp.Forecast .ChanceOfOvercast.Forecast .ChanceOfRain.Forecast .ChanceOfRemDry.Forecast .ChanceOfSnow.Forecast .ChanceOfSunshine.Forecast .ChanceOfThunder.Forecast .ChanceOfWindy.Forecast .CloudCover.Forecast .DewPoint.Forecast .FeelsLike.Forecast .HeatIndex.Forecast .Humidity.Forecast .Precipitation.Forecast .Pressure.Forecast .Temperature.Forecast .Visibility.Forecast .WeatherDescription.Forecast .WindChill.Forecast .WindDir16Point.Forecast .WindDirDegree.Forecast .WindGust.Forecast .WindSpeed.Forecast </p></li><li><p>2015 TechCon Session </p><p>8 | P a g e </p><p>After discussing with your customer, you decide to take the following approach: </p><p>o Look at some of the raw data points using PI Coresight o Put some context around the data by creating a structure in PI Asset Framework (PI AF) o Build a few PI AF analyses to compare forecasts with simulated process data, and create </p><p>aggregated calculations using future data. </p><p>o Use PI Coresight to create a display showing significant weather data for different sites o Create an event report using PI DataLink to keep track of significant weather events </p></li><li><p>Using Future Data to Predict your Process </p><p>9 | P a g e </p><p>3. Use PI Coresight to visualize raw data First we want to build a PI Coresight display holding all forecasted temperature. </p><p>The display should look as follows: </p><p>Play around with this display and explore PI Coresight functionalities. Try to answer the following </p><p>questions: </p><p> How is the current time represented? What does the Now button do? Do you notice any particular behavior? (Hint: Navigate in the </p><p>past until the current time is not displayed on the trend, then do the same in the future) </p></li><li><p>2015 TechCon Session </p><p>10 | P a g e </p><p>Step-by-step instructions: </p><p> Open Internet Explorer, navigate to the PI Coresight landing page (http://localhost/Coresight), and create a new display. </p><p> In the Search Pane, search for all temperature forecast tags on the PI Data Archive Server. </p><p> Drag all the forecast PI tags returned by the search onto the display, and display them as value symbols. </p><p> Build a trend underneath the values showing all forecasted temperatures. Using the Timebar control, change the time range of the trend to show exactly 1 day in the past </p><p>and 1 day in the future. Notice the dotted line representing the current time. </p><p> Modify the trend to show one scale for all traces and compare the temperature forecasts between them. </p><p> Navigate in the past/future. Do the values displayed by the Value symbols change? When? Change the trend to a Table symbol, and note the Maximum/Minimum values of the forecasted </p><p>temperature over 48 hours periods. </p><p> PI Coresight automatically saves your work. Change the table back to a trend and modify the display with the name of your choice. </p></li><li><p>Using Future Data to Predict your Process </p><p>11 | P a g e </p><p>4. Integrate future data in PI AF Our customer has a PI AF Structure that already holds real-time data. </p><p>Take some time to familiarize yourself with the database (how is it organized, what are the templates, </p><p>which data references are used, etc.). Pay close attention to the template NWS_WeatherStation, as it is </p><p>the one we are going to use the most throughout the rest of the lab. </p></li><li><p>2015 TechCon Session </p><p>12 | P a g e </p><p>4.1. Associate future tags with PI AF Attributes We now need to integrate the future data tags into our PI AF Structure. There are several ways to do </p><p>this. We could, for example, create an element template that only holds forecast data and have our </p><p>forecasted and observed weather live in separate elements. Instead, we will simply extend the </p><p>NWS_WeatherStation element template, and use AF Categories to discriminate forecasted weather </p><p>from actual observations. This approach is simpler, and will also be very convenient when we get to use </p><p>the PI Coresight Element-Relative features to compare forecasted and observed weather conditions </p><p>across different cities. </p><p>An Excel file called NWS_WeatherStation+FutureData.xlsx is available under C:\TechCon_Lab\ and has </p><p>references to Future Data tags. That file can be used in conjunction with the PI Builder Excel add-in to </p><p>integrate fu...</p></li></ul>

Recommended

View more >