Running a model in hours, keeping track of days....

[karankhosla]

Hi Folks! It's been a while! I have an interesting problem to deal with.

I'm running a model in hours for a year - this is a client specification. I need to create an electricity load profile for a campus, for which I have data for different buildings over a twenty-four hour period over 3 seasons (summer, winter, monsoons) - see attached spreadsheet (Sheets "LoadFactor" and "TimeTracking"). Instead of creating 8760 data points is there any way I can create look-up tables that pick up hour by hour load profiles depending on the day, depending on season, or if it's a working day or not? This would imply that the model, which is run in hours, also has to keep track of the day. So for example, during summer working days, how can I get it to choose appropriate hour by hour load profile for residential buildings, without hard coding this data into excel.

Any and all advice is welcome!

Best,

Karan

[LAUJJL]

Hi Karan

Why does the client (the decision maker) want to run the model with a time step of an hour?

Is he an SD specialist?

Nothing proves that this is the right way of working.

Is he ready to spend with you the right amount of time during the modeling process and if not at least during the policy analysis phase?

What outcome does the client expect?

Which policies are concerned?

Are these policies applied on an hour basis? Could it be applied on a day basis?

How much does the client expect to earn, or what is the minimum and maximum? Is it reasonable?

I think that it is the responsibility of a good modeler to verify that there is a reasonable chance that the model generates a profit higher that its costs for the client (money and time spent). It is very rarely done.

I would work first on a highly aggregated model, to make the model simple to build and understand even if it does not deliver detailed results and with different time steps and time units(hour or day with a time step of 1/24). I could at the same time, test different strategies of integrating the data (lookup etc.) with a simple model.

After having first studied that first model, I would with care chose to disaggregate little step by little step the model and find progressively more fine tuned policies up to the point where more fine tuning does not add a sufficient added value to be performed.

Of course this takes more time to work like this, but it delivers model of high quality. It is too the reason why it is necessary to weight first the utility of building such a model.

Best regards.

JJ

[Administrator]

Assuming time in your model is in hours, you can get the hour in the day very easily using

Code: Select all

hour in day = modulo ( Time , 24 ) {units:hour}

And getting the day as well is easy.

Code: Select all

hours per day = 24 {units:hour/day}
current day = integer ( Time / hours per day ) {units:day}

Will this give you what you need to use the smaller lookups?

[karankhosla]

Thanks Admin (Tony, Tom?! I'm going to try it out and might have to get back to you!

JJ, good to hear from you! I have built a high level model which works very well - sometimes, no matter how silly it may seem, it is also important for a good modeller to do what the client wants to make sure his/her curiosity is satiated and that the model is relevant to his problem; this may mean (at least to the modeller) getting into minute details - but if at the end of the day, the extra work means that we have a model (or two, in this case) that makes sense, demonstrate reliability, confirm or challenge mental models consistently and foster buy-in from the client - then the work is worth it. Specifics aside, I whole heatedly agree with you!

[tomfid]

One strategy to think about, somewhat common in electric power modeling, would be to simulate only representative times per day or days per month/season. You could also potentially move some of the hourly detail off the time axis and into an array dimension, if serial correlation isn't a big deal.