Archive for March, 2010

DIAS for the NRS-034 Geyser Simulator at DUE 2010

Wednesday, March 31st, 2010

Background

During the 2010 Domestic Use of Energy Conference (DUE) held at the Cape University of Technology on the 30 and 31 March 2010, Enerweb’s Demand Intelligence Group presented 2 papers and presented a NRS-034 Workshop.

During the NRS034 workshop, a hot water geyser simulation application utilizing the Demand Intelligence Analytics Server (DIAS) was presented.

DIAS is a platform which enables Demand Intelligence decision applications to be deployed on the Desktop in a portable fashion and this software was distributed to the conference attendees on a CD.

DIAS to Deliver Desktop Decision Applications

In order to successfully deliver Desktop Decision Applications (in the Demand Intelligence functional area as per this conference) or in any other sector for that matter (e.g. decision applications in logistics, manufacturing, government) three critical components are required as shown in the diagram below. These three components are the Engineering/Statistical models, the Data Sets & Data Management, as well as a Deployment Technology (Software).

As per the diagram thus, three independent components are required in order to deliver a successful desktop decision application. These three components should preferable be independent allowing users to potentially mix and match vendors and technology suppliers in each of the spaces.

Engineering and/or Statistical models are thus developed, where these models typically rely very heavily on a good set of historically data (typically supplied by some type of warehouse) as well as the data management processes which ensure good quality data to be delivered consistently. A typical linear regression model thus, is developed using data as obtained from the user’s chosen database system (e.g. MySql, Oracle, Postgres etc). Once the model is available and has been calibrated it typically goes through a software development life-cycle and can then be deployed on a Desktop or in client server mode, typically executed in any one of a variety of deployment technologies (e.g. Python, VB, Octave, SAS, R, etc).

In the DIAS case, a Java wrapper (Jetty) and some open-source plumbing has been added to the widely popular R statistical engine allowing R scripts (the models) to be immediately rendered as Desktop Decision Applications without the need for a software development project.

This is made possible by the fact that some additional meta data (describing the input and output variables of a R function) are added to the original R script, and the DIAS engine interprets this information and automatically generates a desktop application (that runs from within a web browser) from this information.

Typical Geyser Simulator Results

Here follows a set of selected outputs from the hot water cylinder simulator:

You are now able to simulate the hot water geyser consumption patterns for groups of different consumers

Different control of the geysers can be simulated and the effects on the profile (before and after) can be predicted

Additional Demand Intelligence decision applications e.g. additional end-use simulators, short, medium and long term electrical load forecasting applications, etc. are currently under development

Step by Step guide to run the Geyser simulator

  1. Copy the PotableApps subdirectory and the Start.exe from your CD, to your chosen subdirectory on your computer’s hard drive
  2. Setup a Desktop shortcut to the Start.exe file, so that you can start the applications by clicking on the shortcut icon on your desktop
  3. Run Start.exe
  4. Click on the Geyser Profile Modelling menu item (it appears under the TOU Geyser Model heading) in order to start the DIAS server
  5. Accept any firewall warnings, as the DIAS server will now run on your machine and your own browser will connect to this server via this port
  6. Wait for your web browser to be automatically loaded with the DIAS start page
  7. Click on the “Dynamic Modelling DTPET and TOU on geysers” item in the task bar, which appears on the left hand frame of the browser
  8. Now pass through the 3 steps of the simulation by pressing “1. Run DTPET”, to move to the first blue panel to appear
  9. Go to the bottom of the first blue panel, and then press the “1. Run DTPet , to execute the first part of the model
  10. The output of this model will be the expected average monthly consumption for the group of interest which is now to be used in the 2nd part of the simulation. Move to the 2nd part of the simulation by going to the bottom of the screen and pressing the “Goto Step 2″ button.
  11. Wait for the 2nd blue panel to appear and move to the bottom of the screen, then press the “2. Run Penetration” button, in order to generate a linear model which calculates the geyser penetration levels from the consumption levels
  12. After the graph with the straight line fit on the measured data points appears, go to the bottom of the screen and press the “Goto Step 3″ button.
  13. Wait for the 3rd panel to appear which utilises the Excel input sheet (hotwater_2.xls, as found in YourDrive\YourDirectory\PortableApps\DIAS-Portable\App\DIAS\ebr_home\R\resources\water\hotwater_2.xls) move to the bottom of the screen and press the “3. Run Simulation” button
  14. After the simulation completes a set of hot water consumption patterns, before and after total control panels, as well as a comparison to actually measure profiles will be shown

Which DIAS Files are Needed

Two sets of files are required in order to execute this simulation:

  • The first set are the R (read script) task files, as well as the resource files (being supporting or information files). The R scripts are to be found in YourDrive:\YourDirectory\PortableApps\DIAS-Portable\App\DIAS\ebr_home\R\tasks. These files can be modified (after making a backup), if the user wants to build additional tasksor modify the core of the current simulation.
  • The resources files are to be found in YourDrive:\YourDirectory\PortableApps\DIAS-Portable\App\DIAS\ebr_home\R\resources. If the user wants to modify the physical parameters of the geyser simulation, the hotwater_2.xls file can be found in YourDrive:\YourDirectory\PortableApps\DIAS-Portable\App\DIAS\ebr_home\R\resources\water\hotwater_2.xls A zip file, containing the DIAS server version 0.5.0.1, the script and the resource files are available for download from this location.

What are the current limitations of this NRS-034 geyser simulation model?

Geyser Count: When simulating geyser counts in the order of thousands, the simulation time can be substantive and should be monitored closely. When comparing simulation profiles to actual measured in the field with a total population of around 1000 geysers, already provides simulation results which are practically realistic and completely usable for economic evaluation, design, and impact analysis studies.

Simulation Duration: The current simulation duration has been limited to a single 24 hour period, primarily for demonstration purposes. The simulation does assume that complete recovery of the hot water cylinder is achieved during a 24 hour period and thus there is nothing more to be learned by making the simulation longer than 24 hours. The reader is thus encouraged to choose the values of water consumption, ambient temperatures, etc which typify the type of analysis which is being targeted. The reader is also encouraged to modify the R script which performs the simulation, if inclined to do so and additional scenarios or extensions can be achieved.

Simulation Time Step: The current simulation time-step has been set to 300 seconds or 5 minutes. The R script has been configured to be able to handle different types of time steps, but if this is to be changed the reader should monitor and confirm results. For more information, please contact Schalk Heunis

DIAS PROJECT meets with Nelson Mandela Metropolitan University (NMMU)

Thursday, March 18th, 2010

A one-day workshop was held between the NMMU , Eskom and Enerweb on the 17 March 2009, at the NMMU Summerstrand Campus (South). The intention was to specifically discover, understand and then plan a future for the NMMU forecasting work, in collaboration with the Demand Intelligence Analytics Server (DIAS) delivery platform.

The day started off with a demonstration, by  Prof. Litvine on  the Eskom Forecast Consolidation Project, focussing on the web interfaces that have been developed thus far, with the aid of JSP pages, Web Mathematica, and the Mathematica models which he is building

Everybody participated in an overall architecture discussion, specifically relating to the interaction and collaboration of the Mathematica Tools, the Load Studies Data Store, a future planned Forecast Operational Data Store, and the DIAS delivery platform / productivity tool. In order to ensure future robustness, a long term, “production version”, was to be designed for

Everybody joined in and a road ahead was plotted and future work was identified, with the concept of a “task” being described, where this task must be written in such a way that it can be re-used by multiple platforms, and multiple technologies (e.g. SAS model manager). In addition, the task should be “wrapped” correctly, so that multiple re-use could be obtained from as many consumers as possible

Gerard gave an introductory DIAS slide presentation, where Schalk showed that with the addition of some metadata, added to a R written modelling task, that the DIAS engine is able to render a Desktop Decision Application without the need for any additional application coding (e.g. JSP’s)

Marius and Thomas identified that a DIAS adaptor for Web Mathematic is to be targeted next, adding to the list of DIAS supported platforms , which now already includes R, Java, Python and Octave

Gerard distributed 2 copies of DIAS , one to Thomas Monk , and one copy to Mark Nasila of the NMMU, showing them where they could find more information (i.e. this blog), and the HelloWorld example, so that they could test drive the DIAS platform.