Generic Optimization mix model

[Piper]

Hi,

I'm a Vensimuser and I'm having some trouble creating an optimization model. I wanted to know if maybe Ventity might be the solution. Right now I'm are trying to create a generic mix optimization model with the following features:

- Multiple raw materials from multiple sources
- Multiple plants and lines and products
- Products with different receipts (using the raw materials)
- Products with different marginal value depending on the markets/plants/lines
- Mix evolution over time (months) considering stocks, demand and production
- Multiple constraints on:
o Min/max raw materials per source
o Min/max production per line/product (considering tons/h and total hours per product/line)
o Min/max stocks
o Min/max sales per product/market

Could you provide any insight into whether this is possible using Ventity vs Vensim? I posted on the Vensim forum and it seems like putting constraints on a multi-dimensional model like this is impractical in Vensim.

[tomfid]

Ventity might improve the representation of the problem if, for example, not all sources can be shipped to all factories. But it won't particularly help with the representation of the constraints.

I actually don't think representing the constraints is impractical in Vensim at all. It's probably a little more verbose than doing it in an LP or something like that, but also probably clearer to read. If diagram spaghetti is the problem, judicious use of shadow variables and separate views should take care of it.

What could become impractical is the size of the parameter space. If you start making arrays with [materials,sources,plants,lines,products}, you're searching a 1000-dimension parameter space for just 4 options per dimension. A linear or mixed integer program can handle more detail than that easily. No hill climbing optimizer, including Vensim and Ventity, can. That's the tradeoff you make for having more realistic dynamics.