Merge pull request #840 from openego/features/#800-clustering-with-focus

Features/#800 clustering with focus

Author: KathiEsterl

doc/theoretical_background.rst

@@ -127,6 +127,8 @@ The procedures of the two methods are depicted in the following figure [Esterl20
 
 |
 
+Additionally, there is the option to perform a **clustering with a focus on a defined region**, in which the spatial distribution of clustered buses is intentionally weighted toward the corresponding region. This allows a higher spatial resolution within and around the region of interest, while areas farther away are represented with fewer buses.
+
 In general, the clustering of the **sector-coupled system** is divided into two steps:
 First, the electrical and gas grid are clustered independently using one of the methods described above. Afterwards, nodes of the other sectors (hydrogen, heat, e-mobility and DSM nodes) are mapped according to their connection to electricity or gas buses and aggregated to one node per carrier.
 

etrago/appl.py

@@ -23,7 +23,7 @@
         "q_allocation": "p_nom"
     },
     "start_snapshot": 1,
-    "end_snapshot": 2,
+    "end_snapshot": 168,
     "solver": "gurobi",
     "solver_options": {},
     "model_formulation": "kirchhoff",
@@ -63,31 +63,36 @@
     "extra_functionality": {},
     "network_clustering_ehv": {
         "active": false,
-        "busmap": false
+        "busmap": false,
+        "cpu_cores": 4
     },
-    "network_clustering": {
-        "active": true,
-        "method": "kmedoids-dijkstra",
-        "n_clusters_AC": 30,
-        "cluster_foreign_AC": false,
-        "method_gas": "kmedoids-dijkstra",
-        "n_clusters_gas": 15,
-        "n_clusters_h2": 15,
-        "cluster_foreign_gas": false,
-        "k_elec_busmap": false,
-        "k_gas_busmap": false,
-        "bus_weight_tocsv": null,
-        "bus_weight_fromcsv": null,
-        "gas_weight_tocsv": null,
-        "gas_weight_fromcsv": null,
-        "line_length_factor": 1,
-        "remove_stubs": false,
-        "use_reduced_coordinates": false,
-        "random_state": 42,
-        "n_init": 10,
-        "max_iter": 100,
-        "tol": 1e-6,
-        "CPU_cores": 4
+    "network_clustering":  {
+        "method": {
+            "focus_region": null,
+            "per_country": true, 
+            "algorithm": "kmedoids-dijkstra",
+            "remove_stubs": false,
+            "use_reduced_coordinates": false,
+            "line_length_factor": 1,
+            "random_state": 42,
+            "n_init": 10,
+            "max_iter": 100,
+            "tol": 1e-6,
+            "cpu_cores": 4
+        },
+        "electricity_grid": {
+            "active": true,
+            "cluster_within_focus": false,
+            "n_clusters": 30,
+            "k_elec_busmap": false
+        },
+        "gas_grids": {
+            "active": true,
+            "cluster_within_focus": false,
+            "n_clusters_ch4": 15,
+            "n_clusters_h2": 15,
+            "k_ch4_busmap": false,
+        }
     },
     "spatial_disaggregation": null,
     "snapshot_clustering": {

etrago/args.json

@@ -42,6 +42,7 @@
     from etrago.cluster.spatial import (
         busmap_ehv_clustering,
         drop_nan_values,
+        focus_weighting,
         group_links,
         kmean_clustering,
         kmedoids_dijkstra_clustering,
@@ -195,7 +196,7 @@ def find_de_closest(network, bus_ne):
     # Do not apply this part if the function is used for creating the market
     # model. It adds one bus per country, which is not useful in this case.
     if apply_on != "market_model":
-        if (not etrago.args["network_clustering"]["cluster_foreign_AC"]) & (
+        if (etrago.args["network_clustering"]["method"]["per_country"]) & (
             cluster_met in ["kmeans", "kmedoids-dijkstra"]
         ):
             buses_orig = network.buses.copy()
@@ -515,9 +516,9 @@ def select_elec_network(etrago, apply_on="grid_model"):
     apply_on: str
         gives information about the objective of the output network. If
         "grid_model" is provided, the value assigned in the args for
-        ["network_clustering"]["cluster_foreign_AC""] will define if the
-        foreign buses will be included in the network. if "market_model" is
-        provided, foreign buses will be always included.
+        ["network_clustering"]["method"]["per_country""] will
+        define if the foreign buses will be included in the network.
+        If "market_model" is provided, foreign buses will be always included.
 
     Returns
     -------
@@ -536,10 +537,12 @@ def select_elec_network(etrago, apply_on="grid_model"):
             """Parameter apply_on must be either 'grid_model' or 'market_model'
             """
         )
-    settings = etrago.args["network_clustering"]
+    settings = etrago.args["network_clustering"]["electricity_grid"]
 
     if apply_on == "grid_model":
-        include_foreign = settings["cluster_foreign_AC"]
+        include_foreign = not etrago.args["network_clustering"]["method"][
+            "per_country"
+        ]
     elif apply_on == "market_model":
         include_foreign = True
     else:
@@ -556,7 +559,7 @@ def select_elec_network(etrago, apply_on="grid_model"):
             (elec_network.links.carrier == "AC")
             | (elec_network.links.carrier == "DC")
         ]
-        n_clusters = settings["n_clusters_AC"]
+        n_clusters = settings["n_clusters"]
     else:
         AC_filter = elec_network.buses.carrier.values == "AC"
 
@@ -572,7 +575,7 @@ def select_elec_network(etrago, apply_on="grid_model"):
         elec_network.buses = elec_network.buses[
             AC_filter & (elec_network.buses.country.values == "DE")
         ]
-        n_clusters = settings["n_clusters_AC"] - num_neighboring_country
+        n_clusters = settings["n_clusters"] - num_neighboring_country
 
     # Dealing with generators
     elec_network.generators = elec_network.generators[
@@ -813,14 +816,17 @@ def preprocessing(etrago, apply_on="grid_model"):
                 """)
         network.buses.country.loc[network.buses.country.isna()] = "DE"
 
-    if settings["k_elec_busmap"] is False:
+    if settings["electricity_grid"]["k_elec_busmap"] is False:
         busmap_foreign = unify_foreign_buses(etrago)
     else:
         busmap_foreign = pd.Series(name="foreign", dtype=str)
 
     network_elec, n_clusters = select_elec_network(etrago, apply_on=apply_on)
 
-    if settings["method"] == "kmedoids-dijkstra":
+    if (
+        settings["method"]["algorithm"] == "kmedoids-dijkstra"
+        or settings["method"]["focus_region"] is not None
+    ):
         lines_col = network_elec.lines.columns
 
         # The Dijkstra clustering works using the shortest electrical path
@@ -835,19 +841,8 @@ def preprocessing(etrago, apply_on="grid_model"):
         network_elec.lines = lines_plus_dc.copy()
         network_elec.lines["carrier"] = "AC"
 
-    # State whether to create a bus weighting and save it, create or not save
-    # it, or use a bus weighting from a csv file
-    if settings["bus_weight_tocsv"] is not None:
-        weight = weighting_for_scenario(
-            network=network, save=settings["bus_weight_tocsv"]
-        )
-    elif settings["bus_weight_fromcsv"] is not None:
-        weight = pd.read_csv(
-            settings["bus_weight_fromcsv"], index_col="Bus", squeeze=True
-        )
-        weight.index = weight.index.astype(str)
-    else:
-        weight = weighting_for_scenario(network=network, save=False)
+    # weight buses for clustering
+    weight = weighting_for_scenario(network=network, save=False)
 
     return network_elec, weight, n_clusters, busmap_foreign
 
@@ -883,9 +878,9 @@ def postprocessing(
         busmap : pandas.Series
             Updated mapping between buses and clusters
     """
-    settings = etrago.args["network_clustering"]
-    method = settings["method"]
-    num_clusters = settings["n_clusters_AC"]
+    settings = etrago.args["network_clustering"]["electricity_grid"]
+    method = etrago.args["network_clustering"]["method"]["algorithm"]
+    num_clusters = settings["n_clusters"]
 
     if not settings["k_elec_busmap"]:
         busmap.name = "cluster"
@@ -933,7 +928,9 @@ def postprocessing(
     )
 
     # merge busmap for foreign buses with the German buses
-    if not settings["cluster_foreign_AC"] and (apply_on == "grid_model"):
+    if etrago.args["network_clustering"]["method"]["per_country"] and (
+        apply_on == "grid_model"
+    ):
         for bus in busmap_foreign.index:
             busmap[bus] = busmap_foreign[bus]
             if bus == busmap_foreign[bus]:
@@ -952,7 +949,9 @@ def postprocessing(
         one_port_strategies=strategies_one_ports(),
         generator_strategies=strategies_generators(),
         aggregate_one_ports=aggregate_one_ports,
-        line_length_factor=settings["line_length_factor"],
+        line_length_factor=etrago.args["network_clustering"]["method"][
+            "line_length_factor"
+        ],
         bus_strategies=strategies_buses(),
         line_strategies=strategies_lines(),
     )
@@ -1139,17 +1138,39 @@ def run_spatial_clustering(self):
     -------
     None
     """
-    if self.args["network_clustering"]["active"]:
+    if self.args["network_clustering"]["electricity_grid"]["active"]:
         if self.args["spatial_disaggregation"] is not None:
             self.disaggregated_network = self.network.copy()
         else:
             self.disaggregated_network = self.network.copy(with_time=False)
 
         elec_network, weight, n_clusters, busmap_foreign = preprocessing(self)
 
-        if self.args["network_clustering"]["method"] == "kmeans":
-            if not self.args["network_clustering"]["k_elec_busmap"]:
-                logger.info("Start k-means Clustering")
+        focus_region = self.args["network_clustering"]["method"][
+            "focus_region"
+        ]
+        if focus_region:
+            weight = focus_weighting(
+                self,
+                elec_network,
+                weight,
+                focus_region=focus_region,
+                cluster_within=self.args["network_clustering"][
+                    "electricity_grid"
+                ]["cluster_within_focus"],
+                per_country=self.args["network_clustering"]["method"][
+                    "per_country"
+                ],
+                cpu_cores=self.args["network_clustering"]["method"][
+                    "cpu_cores"
+                ],
+            )
+
+        if self.args["network_clustering"]["method"]["algorithm"] == "kmeans":
+            if not self.args["network_clustering"]["electricity_grid"][
+                "k_elec_busmap"
+            ]:
+                logger.info("Start k-means Clustering AC")
 
                 busmap = kmean_clustering(
                     self, elec_network, weight, n_clusters
@@ -1159,9 +1180,14 @@ def run_spatial_clustering(self):
                 busmap = pd.Series(dtype=str)
                 medoid_idx = pd.Series(dtype=str)
 
-        elif self.args["network_clustering"]["method"] == "kmedoids-dijkstra":
-            if not self.args["network_clustering"]["k_elec_busmap"]:
-                logger.info("Start k-medoids Dijkstra Clustering")
+        elif (
+            self.args["network_clustering"]["method"]["algorithm"]
+            == "kmedoids-dijkstra"
+        ):
+            if not self.args["network_clustering"]["electricity_grid"][
+                "k_elec_busmap"
+            ]:
+                logger.info("Start k-medoids Dijkstra Clustering AC")
 
                 busmap, medoid_idx = kmedoids_dijkstra_clustering(
                     self,
@@ -1193,7 +1219,9 @@ def run_spatial_clustering(self):
 
         logger.info(
             "Network clustered to {} buses with ".format(
-                self.args["network_clustering"]["n_clusters_AC"]
+                self.args["network_clustering"]["electricity_grid"][
+                    "n_clusters"
+                ]
             )
-            + self.args["network_clustering"]["method"]
+            + self.args["network_clustering"]["method"]["algorithm"]
         )