Changing Model Result Years

[vsaini]

Dear Dr. Antti,

Greetings of the day.

I have made a model that have the following data Periods (2012, 2017, 2022, 2027, 2032, 2037, 2042, 2047, 2052) and Result Periods (2012, 2017, 2022, 2027, 2032, 2037, 2042, 2047, 2052). My base year for the model is 2012.

The Beginning and End year of the data periods are provided as shown in image below (Refer Image : Model without Change)

 

Now I wanted to generate the result for 2015, 2020, 2025, 2030, 2035, 2040, 2045, 2050 without changing the data year, so I undertook the following:

·        Added result time periods 2015, 2020, 2025, 2030, 2035, 2040, 2045, 2050

·        Now I ran the model (without changing any input data) by specifying the model milestone years as shown below in image (Refer Image : Model with Change)

And so on for 2050. Please note here, I have made a change in the base year 2012 by making its beginning and end years as same.

Objective value without change 46,339,808.0000

Objective value with change 10,700,604.0000

The issue is my objective value changes drastically after the changes moreover significant changes observed in installed capacity mix after the change. 

Can you please suggest whether I am doing the right way and what could be the possible reasons for such drastic change?
Regards
Vinay Saini

[Antti-L]

Well, I have never changed the Milestone years (result years) in ANSWER myself (I am mostly using VEDA).
But yes, I think you are doing the it the right way.

When using VEDA-FE, changing the Period definition is equally easy, and requires a similar procedure. To my knowledge, it has been done by numerous users without problems. I am also myself doing that frequently. However, in general, it requires that you are using the inter-extrapolation options for any attributes that are not inter-extrapolated by default, but which should be interpolated/extrapolated in the model. See TIMES-Interpolate.pdf for details.

Concerning your case, it is indeed strange that you have such big differences!
It is hard for me to imagine what could be causing them without seeing the model. But you should be able to see it yourself, by analysing the differences in the model results (between the old and new period definition). It might be due to some very simple modelling error, or due to a combination of many constraints that are incompletely specified under the new period definition.

[vsaini]

Dear Dr. Antti,
Thank you for your reply. I guess the problem could be because of inter-extrapolation. I will check it out and thanks for the document.
I read the document and I guess I will be able to solve this issue, but can you please explain what is EPS value means in the document. I could not found its meaning in the TIMES documents.

Regards
Vinay Saini

[Antti-L]

In GAMS, EPS is short for Epsilon, meaning an infinitesimally small positive number.
See:  Extended Range Arithmetic And Error Handling

Think about it as an "explicit" zero value, as opposed to a normal zero value, which is not stored by GAMS, and can thus be considered an "implicit" zero value. All missing data points for time-series parameters have "implicit" zero values.