fix: better implementation and moved to timely-beliefs

Signed-off-by: F.N. Claessen <felix@seita.nl>

flexmeasures/data/models/planning/storage.py

@@ -182,7 +182,8 @@ def _prepare(self, skip_validation: bool = False) -> tuple:  # noqa: C901
                 query_window=(start, end),
                 resolution=resolution,
                 beliefs_before=belief_time,
-                to_instantaneous=True,
+                as_instantaneous_events=True,
+                boundary_policy="first",
             )
         if isinstance(soc_minima, Sensor):
             soc_minima = get_continuous_series_sensor_or_quantity(
@@ -192,7 +193,8 @@ def _prepare(self, skip_validation: bool = False) -> tuple:  # noqa: C901
                 query_window=(start, end),
                 resolution=resolution,
                 beliefs_before=belief_time,
-                to_instantaneous=True,
+                as_instantaneous_events=True,
+                boundary_policy="max",
             )
         if isinstance(soc_maxima, Sensor):
             soc_maxima = get_continuous_series_sensor_or_quantity(
@@ -202,7 +204,8 @@ def _prepare(self, skip_validation: bool = False) -> tuple:  # noqa: C901
                 query_window=(start, end),
                 resolution=resolution,
                 beliefs_before=belief_time,
-                to_instantaneous=True,
+                as_instantaneous_events=True,
+                boundary_policy="min",
             )
 
         device_constraints[0] = add_storage_constraints(

flexmeasures/data/models/planning/tests/conftest.py

@@ -386,6 +386,7 @@ def add_soc_targets(db, add_battery_assets, setup_sources) -> dict[str, Sensor]:
         ("soc-targets (1h)", timedelta(minutes=60), timedelta(minutes=0)),
         ("soc-targets (15min)", timedelta(minutes=15), timedelta(minutes=0)),
         ("soc-targets (15min lagged)", timedelta(minutes=15), timedelta(minutes=5)),
+        ("soc-targets (5min)", timedelta(minutes=5), timedelta(minutes=0)),
         ("soc-targets (instantaneous)", timedelta(minutes=0), timedelta(minutes=0)),
     ]
 

flexmeasures/data/models/planning/tests/test_solver.py

@@ -1870,6 +1870,13 @@ def test_battery_storage_efficiency_sensor(
             "14:00:00",
             "15:00:00",
         ),
+        # A value defined in a finer resolution is downsampled to match the power sensor resolution.
+        # Only a single value coincides with the power sensor resolution.
+        (
+            "soc-targets (5min)",
+            "14:00:00",
+            "14:15:00",  # todo: change to "14:00:00" after timely-beliefs makes it possible to resample to a certain frequency and event resolution simultaneously
+        ),
         # A simple case, SOC constraint sensor in the sample resolution as the power sensor.
         (
             "soc-targets (15min)",

flexmeasures/data/models/planning/utils.py

@@ -329,7 +329,8 @@ def get_series_from_quantity_or_sensor(
     query_window: tuple[datetime, datetime],
     resolution: timedelta,
     beliefs_before: datetime | None = None,
-    to_instantaneous: bool = False,
+    as_instantaneous_events: bool = True,
+    boundary_policy: str | None = None,
 ) -> pd.Series:
     """
     Get a time series given a quantity or sensor defined on a time window.
@@ -357,16 +358,17 @@ def get_series_from_quantity_or_sensor(
             event_starts_after=query_window[0],
             event_ends_before=query_window[1],
             resolution=resolution,
-            # resolution=resolution if not to_instantaneous else timedelta(0),  # NotImplementedError in timely-beliefs
+            # resolution=resolution if not boundary_policy else timedelta(0),  # NotImplementedError in timely-beliefs
+            # frequency=resolution,
             beliefs_before=beliefs_before,
             most_recent_beliefs_only=True,
             one_deterministic_belief_per_event=True,
         )
+        if as_instantaneous_events:
+            bdf = bdf.resample_events(timedelta(0), boundary_policy=boundary_policy)
         time_series = simplify_index(bdf).reindex(index).squeeze()
         time_series = convert_units(time_series, quantity_or_sensor.unit, unit)
 
-        if to_instantaneous and not quantity_or_sensor.event_resolution == timedelta(0):
-            time_series = time_series.ffill(limit=1)
     else:
         raise TypeError(
             f"quantity_or_sensor {quantity_or_sensor} should be a pint Quantity or timely-beliefs Sensor"
@@ -384,7 +386,8 @@ def get_continuous_series_sensor_or_quantity(
     beliefs_before: datetime | None = None,
     fallback_attribute: str | None = None,
     max_value: float | int | pd.Series = np.nan,
-    to_instantaneous: bool = False,
+    as_instantaneous_events: bool = False,
+    boundary_policy: str | None = None,
 ) -> pd.Series:
     """Creates a time series from a quantity or sensor within a specified window,
     falling back to a given `fallback_attribute` and making sure no values exceed `max_value`.
@@ -412,7 +415,8 @@ def get_continuous_series_sensor_or_quantity(
         query_window=query_window,
         resolution=resolution,
         beliefs_before=beliefs_before,
-        to_instantaneous=to_instantaneous,
+        as_instantaneous_events=as_instantaneous_events,
+        boundary_policy=boundary_policy,
     )
 
     # Apply upper limit

requirements/3.10/app.txt

@@ -323,7 +323,7 @@ tabulate==0.9.0
     # via -r requirements/app.in
 threadpoolctl==3.2.0
     # via scikit-learn
-timely-beliefs[forecast]==2.1.0
+timely-beliefs[forecast]==2.2.0
     # via -r requirements/app.in
 timetomodel==0.7.3
     # via -r requirements/app.in

requirements/3.11/app.txt

@@ -321,7 +321,7 @@ tabulate==0.9.0
     # via -r requirements/app.in
 threadpoolctl==3.2.0
     # via scikit-learn
-timely-beliefs[forecast]==2.1.0
+timely-beliefs[forecast]==2.2.0
     # via -r requirements/app.in
 timetomodel==0.7.3
     # via -r requirements/app.in

requirements/3.8/app.txt

@@ -336,7 +336,7 @@ tabulate==0.9.0
     # via -r requirements/app.in
 threadpoolctl==3.2.0
     # via scikit-learn
-timely-beliefs[forecast]==2.1.0
+timely-beliefs[forecast]==2.2.0
     # via -r requirements/app.in
 timetomodel==0.7.3
     # via -r requirements/app.in

requirements/3.9/app.txt

@@ -326,7 +326,7 @@ tabulate==0.9.0
     # via -r requirements/app.in
 threadpoolctl==3.2.0
     # via scikit-learn
-timely-beliefs[forecast]==2.1.0
+timely-beliefs[forecast]==2.2.0
     # via -r requirements/app.in
 timetomodel==0.7.3
     # via -r requirements/app.in

requirements/app.in

@@ -34,7 +34,7 @@ pyomo>=5.6
 tabulate
 timetomodel>=0.7.3
 # significantly faster db queries
-timely-beliefs[forecast]>=2.1
+timely-beliefs[forecast]>=2.2
 python-dotenv
 # a backport, not needed in Python3.8
 importlib_metadata