[Stoves] Using a genetic algorithm and IBM Watson for power grid expert system

[Brown, Henry, DoIT]

Building intelligent power grids ( I will send attachments upon request)



I worked developing training scenarios on an EMS/SCADA for Public Service Company of NM (PNM).

I used PI Historian to predict wind farm load fluctuations.

http://www.pnm.com/systems/nmwec.htm

Wind Power Grids in NM:

http://scratch.mit.edu/projects/GeneMachine/733654



I built genetic algorithms to predict Human Genome contig maps at Los Alamos National Laboratory Human Genome Project. I created a Mycin like AI that used expertise from many users to predict the human genome.



Recently my son and I used a similar approach to build an Exoplanet modeling tool using Kepler data. It predicts simple patterns in planet systems using a java graphics tool called processing (http://processing.org).

http://www.challenge.nm.org/archive/10-11/finalreports/35.pdf



Recent work by IBM Watson using Hadoop and Map/Reduce will be applied to medical diagnostics using Mycin as the model.

Oncotype-DX has revolutionized cancer treatment by collecting tumor samples and characterizing cancer DNA. It allows some women to avoid chemotherapy.
http://www.genomichealth.com/

http://gcat.davidson.edu/Pirelli/index.htm


IBM has announced it will now use Watson at Columbia University and University of Maryland to resurrect MYCIN.
MYCIN was an automated blood diagnostic expert system which combined blood diagnostics from many doctors.
MYCIN failed because diagnostics took weeks in the 1970’s. Today a similar program can be refined daily to increase accuracy.
http://www.time.com/time/business/article/0,8599,2049826,00.html
http://en.wikipedia.org/wiki/Mycin
How to build Watson from free software:
https://www.ibm.com/developerworks/mydeveloperworks/blogs/InsideSystemStorage/entry/ibm_watson_how_to_build_your_own_watson_jr_in_your_basement7?lang=en
New breast cancer tumor diagnostics create an opportunity for IBM to provide the genome community with Watson as a solution to tumor diagnostics data. The Oncotype DX test of Genomic Health (Redwood City, CA) was approved by Medicare 6/2010.
http://www.genomichealth.com/
This test prescribes specific treatments for women with breast cancer based on their tumor markers.
It is a revolution in breast cancer. Colon and lung cancer may also use a similar test soon.
These tests become more accurate as data is entered into treatment databases.
A new DNA chip developed by Mass. Genl. Hospital and Johnson and Johnson may predict cancer decades before tumors grow.
Could IBM Watson be used to cross analyze blood samples for this new test?
http://www.technologytransfertactics.com/content/2011/01/05/mass-general-partners-with-jj-to-commercialize-cancer-test/
http://nextbigfuture.com/2011/01/microfluidics-chip-that-captures-rare.html
http://www.technologyreview.com/blog/editors/26218/
http://www.technologyreview.com/biomedicine/19969/
http://www.ncbi.nlm.nih.gov/pubmed/20930119
http://www.aacr.org/home/public--media/stand-up-to-cancer/su2c-dream-teams/bioengineering-and-clinical-applications-of-circulating-tumor-cell-chip.aspx
UIMA is the brains of Watson. It is an automated annotation system.
http://uima.apache.org/
UIMA could be invaluable in tracking research in cancer molecular biology.
Cancer research is a rapidly evolving field with many research projects.
Genetically cancer is a context switching machine that is constantly changing.
One treatment is usually followed by another as the cancer changes.
Treatments are growing exponentially and can be matched to genetic markers.
There needs to be a group of cancer UIMA systems updating each other similar to the Human Genome databases and Genbank.
http://www.ncbi.nlm.nih.gov/genbank/
Genbank was annotated by researchers by hand. UIMA may allow the automated annotation of genomic data.
A data classification system needs to involve today’s microarrays with tomorrow’s advances in bio chips.
http://uima.apache.org/
http://abhishek-tiwari.com/2009/04/3r-systems-reading-reasoning-and.html
http://www.youtube.com/watch?v=jAegU3aZbWI
http://www.computer.org/portal/web/csdl/doi/10.1109/ISM.Workshops.2007.24
http://www.orbitproject.org/resource/colorectal-cancer-uima-annotators-prostate-cancer-uima-annotators
Could MutationFinder cross correlate tumor markers to treatments similar to Oncotype-DX?
http://mutationfinder.sourceforge.net/Caporaso2007a.pdf

I have created a powerpoint on Comparative Genomics in Cancer.
It is 3 MB in size and I will send if you are interested. It shows how to use processing to build genetic network maps of cancer.

Power grids act deterministically most of the time.
Breast cancer is also beginning to become predictable with Oncotype DX.
By building multiple network responses the grid and cancer may become more predictable over time.

Could power grid modeling and medical diagnostics have parallels?
Gene networks and electric networks can now be modeled using processing.
See attached PNM Power Grid Model.doc

Could algorithms for both seek irregularities and approach solutions similarly?

Smart grids will be challenged by new technology and will need to adapt.
Example:
If Ecat could be attached to a Thermo Electric Generator a constant flow of electricity produced locally could decentralize the grid.
Ecat Scratch Cartoon

http://scratch.mit.edu/projects/GeneMachine/1791470


Could local EMS/SCADA take over for public utility and coop EMS/SCADA?
Could cloud computing reconfigure the national grid to adapt using Hadoop?
Would fluctuation of load be balanced by local production?
Could a processing model for local zones of control work?
Would simulations show a model that stabilizes without external transmission?
Would external transmission become a cost customers would tolerate?


Henry Brown
hbrown at sisna.com<mailto:hbrown at sisna.com>
505 795-3680




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20111014/6fa99d91/attachment.html>