2 way 1 var in non-sync constraints

[jkiviluo]

There is a check that 2 way 1 var should not be in non-sync constraints. But the non-sync constraint has also been modified in the same commit (see ebfbefc). I am not sure what that modification is trying to do and it may be causing a problem with 2-way connections using more than 1 variable. I think the correct formulation should be simpler (the line deducting the v_flow from process_source_sink_eff should not be there? Because of the check, there are no 2-way 1-var flows in non-sync constraints (and they would not work anyway). That line is commented out below.):

s.t. non_sync_constraint{g in groupNonSync, (d, t) in dt} :
# Sum all incoming non-synchronous flows to the group nodes (and possibly decrease them with penalty)
  # Include incoming non-sync flows if they come from outside of the node group (and ignore sync and non-sync flows within the node group)
  + sum {(p, source, sink) in process_source_sink : (g, sink) in group_node && (p, sink) in process__sink_nonSync && (p, g) not in process__group_inside_group_nonSync}
    ( + v_flow[p, source, sink, d, t] 
	    * p_entity_unitsize[p]  
		* step_duration[d, t] )
  # Assumes that exogenous inflows are always non-synchronous (there is no separate parameter for this)
  + sum {(g, n) in group_node} p_positive_inflow[n,d,t]
  - vq_non_synchronous[g, d, t] * group_capacity_for_scaling[g, d]
  <=
# Sum all outgoing flows from the group nodes and multiply that with the non-sync limit
  ( + sum {(p, source, sink) in process_source_sink_noEff : (g, source) in group_node && (p,g) not in process__group_inside_group_nonSync}
      ( + v_flow[p, source, sink, d, t] 
		  * p_entity_unitsize[p] 
	  ) * step_duration[d, t]
    + sum {(p, source, sink) in process_source_sink_eff: (g, source) in group_node && (p,g) not in process__group_inside_group_nonSync}
      ( + v_flow[p, source, sink, d, t]
	      * ( if (p, 'min_load_efficiency') in process__ct_method then pdtProcess_slope[p, d, t]
	          else 1 / pdtProcess[p, 'efficiency', d, t] )
	      * ( if p in process_unit then 1 / ( p_process_sink_coefficient[p, sink] * p_process_source_coefficient[p, source])
	          else 1 )
  	    + ( if (p, 'min_load_efficiency') in process__ct_method then 
	        + ( + (if p in process_online_linear then v_online_linear[p, d, t]) 
	            + (if p in process_online_integer then v_online_integer[p, d, t])
	          )
		      * pdtProcess_section[p, d, t]
	      )
        # -(if (p,g) in process__group_inside_group_nonSync then v_flow[p, source, sink, d, t] else 0)
	  )	* p_entity_unitsize[p]
		* step_duration[d, t]
    # Add exogenous outflow (demand)
    + sum {(g, n) in group_node} -p_negative_inflow[n,d,t]
  ) * pdGroup[g, 'non_synchronous_limit', d]
;

[jkiviluo]

@ArttuTupala Can you check this logic? I think the line is from you and I am not sure what was the aim.

[jkiviluo]

The potential fix is now also in the https://github.com/irena-flextool/flextool/tree/fix-non-sync branch.

[ArttuTupala]

The idea of the non-sync constraint currently is that the non-sync flow to the group is smaller than the flows out of the group, internal demand and the losses of the flows between the members of the group times the limit. This is the reason why the flow out with eff is not restricted to flows only to outside of the group and the flows inside the group are subtracted. Although, the mistake here might be that the subtraction does not only apply to the flows with efficiency.