Add quickfix for multiple subgrids #402

Delete small subgrids

ding0/core/__init__.py

@@ -45,6 +45,7 @@
 from geoalchemy2.shape import from_shape
 import subprocess
 import json
+from pyvis.network import Network as pynet
 
 if not 'READTHEDOCS' in os.environ:
     from shapely.wkt import loads as wkt_loads
@@ -389,6 +390,52 @@ def run_ding0(
         logger.info("STEP 9: Open all switch disconnectors in MV grid")
         self.control_circuit_breakers(mode='open')
 
+        # ==============================================
+        # Quickfix for https://github.com/openego/ding0/issues/402
+        # Removes all nodes belonging to subgraphs
+        for grid_district in self.mv_grid_districts():
+            subgraphs = [s for s in nx.connected_components(grid_district.mv_grid.graph) if len(s) > 1]
+            subgraph_max_len = 0
+            for s in subgraphs:
+                if len(s) > subgraph_max_len:
+                    subgraph_max_len = len(s)
+
+            # Do we have subgraphs?
+            if len(subgraphs) > 1:
+                # Log
+                errstr = ", ".join([f"Grid {_ + 1} ({len(s)} nodes)" for _, s in enumerate(subgraphs)])
+                logger.error(
+                    f"Grid in {str(grid_district)} has {len(subgraphs)} subgrids (cf. issue #402): {errstr}. "
+                    f"Only the largest graph will be retained."
+                )
+
+                # Plot with pyvis for debugging
+                net = pynet(
+                    directed=False,
+                    select_menu=True,
+                    filter_menu=True,
+                )
+                net.show_buttons()
+                graph_relabeled = grid_district.mv_grid.graph.copy()
+                graph_relabeled = nx.relabel_nodes(
+                    graph_relabeled,
+                    dict(zip(
+                        graph_relabeled.nodes(),
+                        [str(n) for n in graph_relabeled.nodes()]
+                    ))
+                )
+                for n1, n2, d in graph_relabeled.edges(data=True):
+                    d.pop("branch", None)
+                net.from_nx(graph_relabeled)
+                net.write_html(f"{str(grid_district)}_graph_debug.html", notebook=False)
+
+                # Remove all nodes belonging to isolated subgraphs
+                for component in subgraphs:
+                    if len(component) < subgraph_max_len:
+                        for node in component:
+                            grid_district.mv_grid.graph.remove_node(node)
+        # ==============================================
+
         logger.info("STEP 10: Do power flow analysis of MV grid")
         self.run_powerflow(session, method='onthefly', export_pypsa=False, debug=debug)
         if export_mv_figures:

ding0/grid/mv_grid/urban_mv_connect.py

@@ -41,6 +41,20 @@ def mv_urban_connect(mv_grid, osm_graph_red, core_graph, stub_graph, stub_dict,
     routed_graph_node_set = routed_graph_node_set - forbidden_object_set
     root_nodes_to_remove = []
 
+    # ==============================================
+    # Detect problem with multiple subgraphs, see https://github.com/openego/ding0/issues/402
+    from collections import Counter
+    stub_root_nodes = [s["root"] for s in stub_dict.values()]
+    if len(stub_root_nodes) > len(set(stub_root_nodes)):
+        duplicated_root_nodes = {r: c for r, c in Counter(stub_root_nodes).items() if c > 1}
+        for r, c in duplicated_root_nodes.items():
+            stubs = [s["comp"] for s in stub_dict.values() if s["root"] == r]
+            logger.error(
+                f"Multiple stubs share the same root node, this is likely to result in multiple graphs "
+                f"(cf. issue #402): Root: {r}, Count: {c}, Stubs: {stubs}"
+            )
+    # ==============================================
+
     for key, stub_data in stub_dict.items():
 
         # get root_node, loads (mv_station or mv_load) and cable distributor nodes in osm stub graph