How to constrain the overall renewable share for all nodes?

[MiaoCheng0517]

Dear support team and everyone, I met a question which is slimilar to #845

I need to add a constraint on the overall renewable share of all nodes. I use the set map_spatial_hierarchy mentioned by @behnam-zakeri in the link above. However, it does not work for my model.

Specifically, I assigned two spatial levels in my model, as you can see in the following code:

scenario.add_set("node", "China")
scenario.add_set("node", province)
scenario.add_set("lvl_spatial", ["country", "province"])
scenario.add_set("map_spatial_hierarchy", ["country", "China", "World"])
for pro in province:
    scenario.add_set("map_spatial_hierarchy", ["province", pro, "China"])
scenario.set("map_spatial_hierarchy")

where province contains several names of China's province. For example, province = ["Beijing","Tianjin","Shanghai","Chongqing"]

Refering to https://github.com/iiasa/message_ix/blob/main/tutorial/westeros/westeros_share_constraint.ipynb, I could add constraints on a single node, take Chongqing for example, the code is as follows:

df = pd.DataFrame(
    {
        "shares": ["share_renewable_electricity"],
        "node_share": "Chongqing",
        "node": "Chongqing",
        "type_tec": "electricity_total",
        "mode": "standard",
        "commodity": "electricity",
        "level": "secondary",
    }
)
scenario.add_set("map_shares_commodity_total", df)
df = pd.DataFrame(
    {
        "shares": ["share_renewable_electricity"],
        "node_share": "Chongqing",
        "node": "Chongqing",
        "type_tec": "electricity_renewable",
        "mode": "standard",
        "commodity": "electricity",
        "level": "secondary",
    }
)
scenario.add_set("map_shares_commodity_share", df)
df = pd.DataFrame(
    {
        "shares": "share_renewable_electricity",
        "node_share": "Chongqing",
        "year_act": [2068, 2073],
        "time": "year",
        "value": [0.9, 0.95],
        "unit": "-",
    }
)
scenario.add_par("share_commodity_lo", df)

With these setting, I can observe that the renewable share in "Chongqing" is limited to 0.9 and 0.95 in 2068 and 2073, respectively.

To constrain the overall renewable share in China, I substituted every "Chongqing" (totally five) in the above block with "China". However, the output is not as wish. The overall renewable share in all province is obviously exceed 0.9 in 2068 and 0.95 in 2073.

Thank you very much for your reading and answering!

[glatterf42]

Hi @MiaoCheng0517, thanks for this question :)
Thanks for searching the tutorials, I think the one you found is the right one for this case, if our code supports this. Please clarify for me: what kind of bound do you want to add for the renewable share?
The code you posted here is adding adding a share_commodity_lo, which defines a lower bound. If you want to define an upper bound instead, please try using share_commodity_up. If you want to limit the model to these exact numbers, you can add both -- the upper and the lower bound.

[MiaoCheng0517]

Thank you for your response. I want to define the lower bound of renewable share. But the renewable share constraint on the parent node doesn't seem to have been applied.

I believe the confusion was caused by an error in my description, which led to a misunderstanding. The overall renewable share in all procinve is obviously lower than 0.9 in 2068 and 0.95 in 2073. I apologize for my carelessness.

[khaeru]

@MiaoCheng0517, it is hard for us to know what precisely is happening and provide advice without knowing the exact contents/structure of the model you have built.

However, it is possible to make some guesses. My guess here is you are (reasonably) expecting a feature that the GAMS implementation of MESSAGE does not have.

In particular, you put constraints on technologies that output to (that is: have output parameter values for):

• commodity: electricity
• level: secondary
• node: Chongqing

These technologies are in two groups called electricity_renewable and electricity_total. (You don't say what technologies you have put in these groups.)

This activates the equation called SHARE_CONSTRAINT_COMMODITY_LO, which is documented here.

Then, you change to n="China". Separately, you have used map_spatial_hierarchy to put n="Chongqing" and some other nodes (your message says for pro in province: but you don't show what province is) under n="China".

Two key points here:

1. Again, I am guessing, but you do not have any technologies (in the electricity_renewable or electricity_total group) that output to:

   • commodity: electricity
   • level: secondary
   • node: China

   In other words, there is no data in output with these keys. Rather, I guess, your technologies all output to their respective node_dest, which is the same as node_loc, e.g. "Chongqing".

   If this is not the case, please provide further information.

2. The equation SHARE_CONSTRAINT_COMMODITY_LO does not automatically compute aggregates over the commodity outputs to "spatial hierarchy levels" that are lower than any respective level. You can confirm this by:

   • Looking at the documentation page I linked above. Notice that n^L appears as a subscript—i.e. output and other parameter/variable data are handled for particular nodes, but no mention is made of "child" nodes at lower.
   • Looking at the GAMS code directly:

     message_ix/message_ix/model/MESSAGE/model_core.gms

     Lines 1485 to 1517 in 00241f6

     	SHARE_CONSTRAINT_COMMODITY_LO(shares,node_share,year,time)$( share_commodity_lo(shares,node_share,year,time) )..
     	* total input and output by `type_tec_share` technologies mapped to respective commodity, level and node
     	SUM( (node,location,type_tec_share,tec,vintage,mode,commodity,level,time2)$(
     	( map_shares_commodity_share(shares,node_share,node,type_tec_share,mode,commodity,level) OR
     	map_shares_commodity_share(shares,node_share,node,type_tec_share,'all',commodity,level) ) AND
     	cat_tec(type_tec_share,tec) AND
     	map_tec_act(location,tec,year,mode,time2) AND
     	map_tec_lifetime(location,tec,vintage,year)
     	),
     	(
     	output(location,tec,vintage,year,mode,node,commodity,level,time2,time) +
     	input(location,tec,vintage,year,mode,node,commodity,level,time2,time)
     	) *
     	duration_time_rel(time,time2) *
     	ACT(location,tec,vintage,year,mode,time2)
     	)
     	=G=
     	share_commodity_lo(shares,node_share,year,time) * (
     	* total input and output by `type_tec_total` technologies mapped to respective commodity, level and node
     	SUM( (node,location,type_tec_total,tec,vintage,mode,commodity,level,time2)$(
     	( map_shares_commodity_total(shares,node_share,node,type_tec_total,mode,commodity,level) OR
     	map_shares_commodity_total(shares,node_share,node,type_tec_total,'all',commodity,level) ) AND
     	cat_tec(type_tec_total,tec) AND
     	map_tec_act(location,tec,year,mode,time2) AND
     	map_tec_lifetime(location,tec,vintage,year)
     	),
     	(
     	output(location,tec,vintage,year,mode,node,commodity,level,time2,time) +
     	input(location,tec,vintage,year,mode,node,commodity,level,time2,time)
     	) *
     	duration_time_rel(time,time2) *
     	ACT(location,tec,vintage,year,mode,time2)
     	) ) ;

     See that it does not identify the sub-nodes of any given node and compute their activity/output/input.

In other words, because of (1), your model does not output anything to (n=China), and so the constraints have no effect.

Let's please confirm this is a correct understanding of your situation. After that, we can talk about how to achieve what you intend.

[glatterf42]

Thanks for jumping in here, @khaeru. I have to say, though: I was also just studying the docs and this exact GAMS code and would not have arrived at your conclusions. In my view, this equation is not well documented, in particular, I don't see any documentation of the required map_shares_commodity_total and map_shares_commodity_share. So I, for one, don't understand why you can specify node_share and node and what these mean in this context. From our docs, "The implementation is generic and flexible, so that any combination of commodities, levels, technologies and nodes can be put in relation to any other combination" makes it sound like our code can handle the desired use case here. My interpretation would have been that we just need to figure out the right combination of node_share and node.
Maybe @behnam-zakeri again has more insight into the spatial implementation?

[MiaoCheng0517]

Thanks @khaeru @glatterf42 for seriously considering the questions I raised.

1. @khaeru 's conjecture about my model was correct. My technologies all output to their respective node_dest, which is the same as node_loc, e.g. "Chongqing". No technology output to the parent node "China".
2. According to @glatterf42 's insight, I changed all node_share in sets "map_shares_commodity_total" and "map_shares_commodity_renewable" to the parent node "China", and kept all node in these two sets as individual child node, e.g. "Chognqing". The node_share in the parameter "share_commodity_lo" was assigned as "China". Then it worked. The output shows that the renewable share of each child node might not necessarily exceed 0.9, but the aggregate exceeds 0.9.

Thank you again for your response and discussion.