pyomo | oriented algebraic modeling language in Python

I think there are 2 things I can add here that might be your missing links...

First, as mentioned in comments, the stuff you feed into the model must be legal expressions that do not depend on the value of the variables at time of creation, so len() etc. are invalid. Solution: use binary variables for those types of counting things and then sum them over appropriate indices.

Second, you are indexing your first variable correctly, but you have a second variable you need to introduce, namely, the decision to send worker w some material matl. See my example below that introduces this variable and then uses a big-M constraint to link the two decisions together.... Specifically, ensure the model delivers a required material to a worker for a task in which it is required.

Welcome to the site.

You are off to an "OK" start. Your model has quite a few errors in it.... did you look at the examples in the pyomo documentation?

A few suggestions:

1. Start with ConcreteModel and manually initialize the data. I think it is easier to do, esp. with python's ability to handle .csv files either manually (as I show below) or with pandas or csv_reader.

2. Throw out pandas for now. Use it from a separate file if need be to create the .csv files if you are comfortable with that, or just manually write them, or use csv_reader, but don't commingle pandas and pyomo till you get your feet on the ground. Same advice for numpy.

3. Use "flat file" format for your data, not tabular. It is easier to ingest. So, for example, create your distance table in a csv that has 3 columns as mine does and it is easier to read into a dictionary or, if you go to an AbstractModel it is in an easy format.

4. Use a small slice of your data and pprint() your model to make sure it makes sense and complies with your math model.

Well, I went a little geek on this. Got hooked, kinda interesting problem.

So, I made a bunch of changes and left some of your code in this example. I also chopped down a handful of the cost variables and made them rather simple as I was getting a little lost in the sauce and so that I was (mostly) convinced things were working, so the units/conversions/costs are a bit nonsensical now, but should be easily recovered.

Hopefully there are a couple concepts in here that you can use as you work through this. A few notes...

• Needed a binary variable to indicate that the plant was started, and another to keep track of whether it was "running" or not in a particular period, these were linked with a constraint
• Added a little trickery with the time windows to make a rolling evaluation period for total gas use
• Added a minimum use for the plant to run or else once it was "started" it could arbitrarily run with 0 gas when not profitable, now a minimum-run or off decision is forced

Plot shows pretty convincing evidence that it is running as hoped, it starts up, runs at max blast when price is high, and adheres to rolling gas limit, then shuts down and does it again.

After trying other open source solvers like cbc and having the same issue, I have seen that the error with cbc also gives the information:

Given that the model works on some data, but not on an alternate data source, we can obviously focus a bit on the data set (which isn't shown).

We have a huge clue in the error report that the problem is unbounded. This means that there is nothing to prevent the objective function from running away to infinity, or negative infinity in the case of a minimization problem. So, let's look at your objective function. You are:

This is just a hunch, but I think you have something odd going on within your namespace. model is not defined outside of your function and you aren't "catching" the instance you create when calling the function in this segment:

This is a very long post where I think you could have whittled it down to a more minimal example. I think you are asking if/how you can set the value of a particular variable (in your case the first indexed one) from some known value (in your case, something from the preceding run).

You can do this by simply assigning a value to the variable, and then "fixing" the value for that run, then go on to solve...

I had to improvise a bit without access to model.T and pe.value(eval(var_name)) but I think this is close to what you want. I used an f-string to clean up your var_name syntax. Next we check to see if a key is already in the_dict with .get() and extend or append as needed. Otherwise, we create a new key. Finally, we can just pass the_dict to the DataFrame constructor.

If this doesn't work for you, post a comment and I'll assist until it does. Providing a sample list for model.T would be a big help.

So here is what is going on. When you are initializing your parameter, you are failing to pass in a key:value pairing (like a dictionary) so pyomo is just initializing each element in your parameter to the whole array which is clearly not what you want.

Most of the time, the indexing sets are not just counting numbers, so this doesn't really pop up too often, because if your index is {'A', 'B', 'C', ...} then it becomes clear that you must pass in a key:value mapping in order to straighten that out.

Here are a couple lines to demo what I'm talking about. Bottom line: get numpy out of your model and make a dictionary of the stuff you want to use in the parameter. (This can also be done w/ pandas, but it isn't necessary). List of commands at end if you wish to copy/replicate/experiment.