Invalid formulations when a pair of conneted `unit` and `node` use different stochastic/temporal structures

[nnhjy]

For a (unit, node) pair connected with a unit__node relationship, if one assigns a dedicated stochastic structure to either of its elements in units_on__stochastic_structure or node__stochastic_structure, the resulted variable units_on is stemmed to be 0 and so is the variable unit_flow.

The problem arises from the separation of the constraints on the two variables due to the different stochastic structures. unit_flow follows the stochastic structure of its associated node. And its capacity constraint needs units_on with the same stochastic structure, which would otherwise bind the unit_flow to 0. Likewise, a units_on would be 0 unless there exists a unit_flow with the same stochastic structure to represent the demand from a node in the capacity constraint that links both.

What we would like to have for such a discrepancy? To allow flexibility in this regard, there needs some communication between the related variables taking different stochastic structures.

The issue could be reproduced with this database: sample_sto_structure.zip

[nnhjy]

Similar issues appear in the rolling forward mode with multiple temporal blocks, which were mentioned in issue #633.

In the case where the rolling window uses 2 or more consecutive temporal blocks to define its temporal structure, unit_flow and 'units_onwould work only if the associatedunitandnode` share the same temporal block in each temporal section. The issue could be more complicated. Hope this database sample_temp_structure.zip could illustrate that.

The system contains a unit unitB and a node nodeB. The model rolls 1h per solve with a temporal span of 2h. This 2h solving window consists of a "realise" temporal block followed by a "lookahead" temporal block, with which the available temporal block option for the model is designed as:

name	span	model default
realise	0h-1h	yes
lookahead	1h-2h	yes
realise_dedicated	0h-1h	no
lookahead_dedicated	1h-2h	no

Performances of unit_flow and units_on for all cases are summarised in the table below. The * cases of the big trouble. The term "effective" means the variable present a value as expected.

case	model entity	0h-1h	1h-2h	unit_flow behaviour	units_on behaviour
1	unitB	realise	lookahead	effective over full horizon	effective over full horizon
	nodeB	realise	lookahead
2	unitB	realise_dedicated	lookahead_dedicated	effective over full horizon	effective over full horizon
	nodeB	realise_dedicated	lookahead_dedicated
3	unitB	realise	lookahead	effective over full horizon	effective over full horizon
	nodeB	realise_dedicated	lookahead_dedicated
4	unitB	realise_dedicated	lookahead_dedicated	effective over full horizon	effective over full horizon
	nodeB	realise	lookahead
5*	unitB	realise	lookahead_dedicated	deactivated (appears as 0 only at the last window)	deactivated (appears as 0 only at the last window)
	nodeB	realise	lookahead_dedicated
6	unitB	realise_dedicated	lookahead	effective but missing the last window	effective over full horizon
	nodeB	realise_dedicated	lookahead
7*	unitB	realise	lookahead_dedicated	appears as 0 over full horizon except for the last window being effective	deactivated (appears as 0 only at the last window)
	nodeB	realise	lookahead
8*	unitB	realise	lookahead	deactivated (appears as 0 only at the last window)	appears as 0 over full horizon
	nodeB	realise	lookahead_dedicated
9	unitB	realise_dedicated	lookahead_dedicated	effective but missing the last window	effective but appears as 0 at the last window
	nodeB	realise_dedicated	lookahead
10	unitB	realise_dedicated	lookahead	effective but appears as 0 at the last window	effective over full horizon
	nodeB	realise_dedicated	lookahead_dedicated
11*	unitB	realise	lookahead_dedicated	same as in 7	same as in 7
	nodeB	realise_dedicated	lookahead_dedicated
12*	unitB	realise_dedicated	lookahead_dedicated	same as in 8	same as in 8
	nodeB	realise	lookahead_dedicated
13	unitB	realise	lookahead	same as in 9	same as in 9
	nodeB	realise_dedicated	lookahead
14	unitB	realise_dedicated	lookahead same as in 10	same as in 10
	nodeB	realise	lookahead
15*	unitB	realise	lookahead_dedicated	appears as 0 over full horizon	deactivated (appears as 0 only at the last window)
	nodeB	realise_dedicated	lookahead
16*	unitB	realise_dedicated	lookahead	same as in 8	same as in 8
	nodeB	realise	lookahead_dedicated