add smooth blocks options

india_forecast_app/app.py

@@ -149,6 +149,7 @@ def get_model(
     hf_repo: str,
     hf_version: str,
     name: str,
+    smooth_blocks: Optional[int] = 0,
 ) -> PVNetModel:
     """
     Instantiates and returns the forecast model ready for running inference
@@ -160,6 +161,7 @@ def get_model(
             hf_repo: ID of the ML model used for the forecast
             hf_version: Version of the ML model used for the forecast
             name: Name of the ML model used for the forecast
+            smooth_blocks: Number of blocks to smooth the forecast by
 
     Returns:
             A forecasting model
@@ -169,7 +171,13 @@ def get_model(
     model_cls = PVNetModel
 
     model = model_cls(
-        asset_type, timestamp, generation_data, hf_repo=hf_repo, hf_version=hf_version, name=name
+        asset_type,
+        timestamp,
+        generation_data,
+        hf_repo=hf_repo,
+        hf_version=hf_version,
+        name=name,
+        smooth_blocks=smooth_blocks,
     )
     return model
 
@@ -363,6 +371,7 @@ def app_run(timestamp: dt.datetime | None, write_to_db: bool = False, log_level:
                     hf_repo=model_config.id,
                     hf_version=model_config.version,
                     name=model_config.name,
+                    smooth_blocks=model_config.smooth_blocks,
                 )
                 ml_model.site_uuid = site.site_uuid
 

india_forecast_app/models/all_models.yaml

@@ -7,24 +7,28 @@ models:
     version: ae07c15de064e1d03cf4bc02618b65c6d5b17e8e
     client: ruvnl
     asset_type: wind
+    smooth_blocks: 4
   - name: windnet_india_mo
     type: pvnet
     id: openclimatefix/windnet_india
     version: 546baded3d4216736d8ee8d6798d47235bd72b08
     client: ruvnl
     asset_type: wind
+    smooth_blocks: 7
   - name: windnet_india_mo_v2
     type: pvnet
     id: openclimatefix/windnet_india
     version: 165267d34500cf0c881ed70d9318421f4e0d10f1
     client: ruvnl
     asset_type: wind
+    smooth_blocks: 4
   - name: windnet_india_mo_v3
     type: pvnet
     id: openclimatefix/windnet_india
     version: 990bed9ad1dbb10515f830c181609b10e72c975e
     client: ruvnl
     asset_type: wind
+    smooth_blocks: 7
    # RU client solar
   - name: pvnet_india
     type: pvnet

india_forecast_app/models/dummy.py

@@ -5,6 +5,7 @@
 import datetime as dt
 import math
 import random
+from typing import Optional
 
 import pandas as pd
 import pytz
@@ -21,13 +22,14 @@ def version(self):
         return "0.0.0"
 
     def __init__(
-            self,
-            asset_type: str,
-            timestamp: dt.datetime,
-            generation_data: dict[str, pd.DataFrame] = None,
-            hf_version: str = None,
-            hf_repo: str = None,
-            name: str = None,
+        self,
+        asset_type: str,
+        timestamp: dt.datetime,
+        generation_data: dict[str, pd.DataFrame] = None,
+        hf_version: str = None,
+        hf_repo: str = None,
+        name: str = None,
+        smooth_blocks: Optional[int] = 0,
     ):
         """Initializer for the model"""
         self.asset_type = asset_type

india_forecast_app/models/pvnet/model.py

@@ -7,6 +7,7 @@
 import os
 import shutil
 import tempfile
+from typing import Optional
 
 import numpy as np
 import pandas as pd
@@ -63,6 +64,7 @@ def __init__(
         hf_repo: str,
         hf_version: str,
         name: str,
+        smooth_blocks: Optional[int] = 0,
     ):
         """Initializer for the model"""
 
@@ -72,6 +74,7 @@ def __init__(
         self.name = name
         self.site_uuid = None
         self.t0 = timestamp
+        self.smooth_blocks = smooth_blocks
         log.info(f"Model initialised at t0={self.t0}")
 
         self.client = os.getenv("CLIENT_NAME", "ruvnl")
@@ -181,14 +184,17 @@ def predict(self, site_id: str, timestamp: dt.datetime):
                     ) * smooth_values[final_gen_index + idx]
                 log.debug(f"New values are {values_df['forecast_power_kw']}")
 
-        if self.asset_type == "wind":
-            # Smooth with a 1 hour rolling window
-            # Only smooth the wind else we introduce too much of a lag in the solar
-            # going up and down throughout the day
+        if self.smooth_blocks > 0:
+            log.info(f"Smoothing the forecast with {self.smooth_blocks} blocks")
             values_df["forecast_power_kw"] = (
-                values_df["forecast_power_kw"].rolling(4, min_periods=1).mean().astype(int)
+                values_df["forecast_power_kw"]
+                .rolling(window=self.smooth_blocks, min_periods=1, center=True)
+                .mean()
             )
 
+        # convert to int
+        values_df["forecast_power_kw"] = values_df["forecast_power_kw"].astype(int)
+
         # remove any negative values
         values_df["forecast_power_kw"] = values_df["forecast_power_kw"].clip(lower=0.0)
 

india_forecast_app/models/pydantic_models.py

@@ -32,14 +32,22 @@ class Model(BaseModel):
         60,
         title="Average Minutes",
         description="The number of minutes that results are average over when "
-                    "calculating adjuster values. "
-                    "For solar site with regular data, 15 should be used. "
-                    "For wind sites, 60 minutes should be used.",
+        "calculating adjuster values. "
+        "For solar site with regular data, 15 should be used. "
+        "For wind sites, 60 minutes should be used.",
+    )
+    smooth_blocks: int = Field(
+        0,
+        title="Smooth Blocks",
+        description="The number of blocks to smooth the forecast by. "
+        "We use a central smoothing and each block is 15mins, "
+        "so 5 would be smoothing from 30s before and afterwards",
     )
 
 
 class Models(BaseModel):
-    """ A group of ml models """
+    """A group of ml models"""
+
     models: List[Model] = Field(
         ..., title="Models", description="A list of models to use for the forecast"
     )