Having developed a discrete event simulator myself for my master's thesis I can guarantee that the usefulness of tools like this one depends heavily on the veracity and quality of the time parameters and the random distributions.
This one is a very cool exercise, but applicability is dubious.
Anyway, does anybody have any suggestion about job positions where the job requires developing this kind of stuff (simulators, process optimization, heuristics)? Being a fresh graduate in this time period is pretty bad, but doesn't hurt to do some research on interesting job positions. I only see web related job positions and "machine learning" job positions lately.
If you're interested in MIP/LP optimization models and have either AIMMS or GAMS experience and either CPLEX or GUROBI knowledge, the vendors of most power systems simulators probably have some openings. They're large codebases, but actively saving billions annually in the US. The problems of unit commitment and economic dispatch are well understood, but the business rules framework is massive for all the US RTO/ISOs and is always changing.
No problem. Vendors are GE, Siemens, and ABB. I'm sure at least one is hiring. They generally prefer some industry experience, but a Masters or PhD would likely go far, especially with some research projects.
These models are amongst the most difficult out there as far as size and time requirements. One of the founders of GUROBI got involved with the industry recently to try to help some researchers speed things up.
If you want a decent example/starter model to analyze, there is a unit commitment model someone made in Xpress that is free online if you Google for it that shows the fundamental formulation although they are much larger in practice.
Unit commitment is the MIP problem that combines linear and integer constraints and tries to determine the least cost set of units to bring online for each hour of the day. Constraints include things like the minimum amount of time the unit has to remain offline before being started up again, the minimum amount of time it has to run once online, how fast it can move (ramp), how much capacity it has..etc. You try to minimize the costs of starting up the unit, the cost of the unit just being online, and energy costs. The economic dispatch problem is much simpler and asks, with the set of units that were given to me by unit commitment, where should I set each one. The commitment problem runs the day ahead at hourly granularity for the next day and is also run periodically throughout the day. The dispatch problem generally runs every 5 minutes 24/7 365. There are also other constraints like not burning down the transmission grid.
The US military (and other too) use simulation, modeling, and analysis tools to help design all sorts of activities and equipment. Examples at http://jasp-online.org/model-simulations/
* I used ESAMS (Enhanced Surface-to-air Missile Simulation) and CODER (COnceptual DEsign Representation) to help flesh out design details for various US Air Force avionics platforms. ESAMS is a tools to help evaluate missile / aircraft scenarios. CODER is a discrete event simulation tool for designing and evaluating aircraft cockpit automation.
The RAND Corporation is a group that uses games and simulation to better understand many subjects (policy, healthcare, education, etc). See https://www.rand.org/search.html?query=games
No idea if they are hiring but a friend of mine works for a company called Sandtable and they mostly do advanced agent-based modelling. Based in London.
Hi, I'm the creator of this project. Seems like someone shared it here. This is a school project and a current work-in-progress, and I'm open to any feedback available. The usefulness of a simulator is heavily dependent on how well it approximates to reality, which I have yet to do. It is currently a baseline experiment without any reference to current research on COVID-19, but stay tuned as I'm working on it! And thanks for sharing :)
I am also taking AI this semester and was considering expanding upon something like this for my masters' capstone project. The application of machine learning to this type of situation is going to be very interesting to say the least.
Basically optimizations and "modelling" are the reason resource allocations are total failure in the case of rare emergencies, like ongoing.
You assume the worst "one in a 1000 years" case, multiply by 2..10, depending on your cash flow, and keep allocations up to date. Period.
And the simple truth again: you can't immediately allocate during a crisis.
I am modelling with American Community Survey complete raw data, physically in Berkeley, California.. using PostGIS, python and an SEIR model; Urban Planning background.. suggestions welcome
example output, US Persons by Age-Sex 50+ by Census Tract, nationwide.. exec. time 1230 ms. local, no clouds
-[ RECORD 221408 ]----+
mtable_2_pkey | 18744349
geoid | 08000US361198400084000000900
geo_name | Census Tract 9, Yonkers city,
Yonkers city, Westchester County, New York
Total_Population | 2307
Male | 1085
male over 50 est. | 285
Female | 1222
female over 50 est. | 272
-[ RECORD 221409 ]----+
mtable_2_pkey | 18754394
geoid | 08000US421338704887048000900
geo_name | Census Tract 9, York city, York city, York County, Pennsylvania
Total_Population | 7100
Male | 3934
male over 50 est. | 1453
Female | 3166
female over 50 est. | 1347
-[ RECORD 221410 ]----+
mtable_2_pkey | 18776479
geoid | 14000US42133000900
geo_name | Census Tract 9, York County,
Pennsylvania
Total_Population | 1169
Male | 667
male over 50 est. | 226
Female | 502
female over 50 est. | 240
This one is a very cool exercise, but applicability is dubious.
Anyway, does anybody have any suggestion about job positions where the job requires developing this kind of stuff (simulators, process optimization, heuristics)? Being a fresh graduate in this time period is pretty bad, but doesn't hurt to do some research on interesting job positions. I only see web related job positions and "machine learning" job positions lately.