diff --git a/RELEASE_NOTES.md b/RELEASE_NOTES.md
index d471f0e..0b51df5 100644
--- a/RELEASE_NOTES.md
+++ b/RELEASE_NOTES.md
@@ -20,5 +20,7 @@
 - Published a locale-aware `ColumnMapper` utility that reads the YAML schema so notebooks can seamlessly move between raw API headers, canonical identifiers, and localized display labels.
 
 ## Bug Fixes
+- `frequenz.lib.notebooks.reporting.utils.helpers.add_energy_flows()` now infers consumption totals from existing data when explicit consumption columns are missing, preventing inconsistent outputs in notebook pipelines that only provide grid and production inputs.
+- `frequenz.lib.notebooks.reporting.metrics.consumption()` reindexes optional production/battery inputs and raises a warning when inferred consumption turns negative so sign-convention issues are surfaced immediately.
 
 <!-- Here goes notable bug fixes that are worth a special mention or explanation -->
diff --git a/src/frequenz/lib/notebooks/reporting/metrics/reporting_metrics.py b/src/frequenz/lib/notebooks/reporting/metrics/reporting_metrics.py
index 9055fa4..4aff87d 100644
--- a/src/frequenz/lib/notebooks/reporting/metrics/reporting_metrics.py
+++ b/src/frequenz/lib/notebooks/reporting/metrics/reporting_metrics.py
@@ -96,7 +96,11 @@ def production_excess_in_bat(
     production_excess_series = production_excess(
         production, consumption, production_is_positive=production_is_positive
     )
-    battery = battery.astype("float64").clip(lower=0)
+    battery = (
+        battery.astype("float64")
+        .reindex(production_excess_series.index, fill_value=0.0)
+        .clip(lower=0)
+    )
     return pd.concat([production_excess_series, battery], axis=1).min(axis=1)
 
 
@@ -199,44 +203,60 @@ def production_self_share(
 
 
 def consumption(
-    df: pd.DataFrame, production_cols: list[str] | None, grid_cols: list[str]
+    grid: pd.Series,
+    production: pd.Series | None = None,
+    battery: pd.Series | None = None,
 ) -> pd.Series:
-    """Infer the consumption column from grid and production data if missing.
+    """Infer total consumption from grid import, on-site production, and battery power.
 
-    If a 'consumption' column is not present, it is computed as the total grid import
-    (sum of all grid columns) minus total production. Safely handles missing or
-    empty production columns by treating them as zero.
+    Computes: consumption = grid_import - production (raw production-neg values)
+                            - battery (raw - with positive and negative values).
 
     Args:
-        df: Input DataFrame containing grid and optional production columns.
-        production_cols: List of production column names (e.g., "pv", "chp", "battery" or "ev").
-            Can be None or empty if no on-site generation is present.
-        grid_cols: List of one or more grid column names.
+        grid: Series of grid import values (e.g., kW or MW).
+        production: Optional Series of on-site production values.
+            If None, production is treated as zero.
+        battery: Optional Series representing battery discharge/charge power.
+            Positive values increase inferred consumption (battery discharge),
+            while negative values decrease it (battery charging). If None, the
+            battery contribution is treated as zero.
 
     Returns:
         A Series representing inferred total consumption, named `"consumption"`.
 
     Raises:
-        ValueError: If `grid_cols` is empty.
+        ValueError: If `grid` is None.
+
+    Warns:
+        UserWarning: If negative inferred consumption values are detected,
+            which may indicate times of net export or a sign-convention mismatch.
     """
-    if "consumption" in df.columns:
-        return df["consumption"]
+    if grid is None:
+        raise ValueError("`grid` must be provided as a pandas Series.")
 
-    if not grid_cols:
-        raise ValueError("At least one grid column must be specified in grid_cols.")
+    grid_s = grid.astype("float64").fillna(0)
 
-    # Compute total grid import and total production
-    grid_total = df[grid_cols].sum(axis=1)
+    # Ensure raw production values are used (usually negative for production)
+    if production is None:
+        prod_s = pd.Series(0.0, index=grid_s.index)
+    else:
+        prod_s = production.astype("float64").reindex(grid_s.index, fill_value=0.0)
 
-    # Handle empty production columns safely
-    if production_cols:
-        production_total = df[production_cols].sum(axis=1)
+    if battery is None:
+        battery_s = pd.Series(0.0, index=grid_s.index)
     else:
-        # No production → production_total = 0
-        production_total = pd.Series(0, index=df.index)
+        battery_s = battery.astype("float64")
+        battery_s = battery_s.reindex(grid_s.index, fill_value=0.0)
 
-    # Compute inferred consumption (Series)
-    consumption = grid_total - production_total
-    consumption.name = "consumption"
+    result = (grid_s - prod_s - battery_s).astype("float64")
+    result.name = "consumption"
 
-    return consumption
+    if (result < 0).any():
+        warnings.warn(
+            "Negative inferred consumption detected. This can occur during net export "
+            "or due to a sign-convention mismatch between grid and production.",
+            UserWarning,
+            stacklevel=2,
+        )
+
+    return result
diff --git a/src/frequenz/lib/notebooks/reporting/utils/helpers.py b/src/frequenz/lib/notebooks/reporting/utils/helpers.py
index dc9df15..9f11115 100644
--- a/src/frequenz/lib/notebooks/reporting/utils/helpers.py
+++ b/src/frequenz/lib/notebooks/reporting/utils/helpers.py
@@ -23,6 +23,7 @@
 import pandas as pd
 
 from frequenz.lib.notebooks.reporting.metrics.reporting_metrics import (
+    asset_production,
     grid_feed_in,
     production_excess,
     production_excess_in_bat,
@@ -43,11 +44,14 @@ def _get_numeric_series(df: pd.DataFrame, col: str | None) -> pd.Series:
         col: Column name to retrieve. If None or missing, zeros are returned.
 
     Returns:
-        A float64 Series with non-negative values, matching the input index.
+        A float64 Series aligned to the input index.
     """
     if col is None:
-        return pd.Series(0.0, index=df.index, dtype="float64")
-    return df.reindex(columns=[col], fill_value=0)[col].astype("float64").clip(lower=0)
+        series = pd.Series(0.0, index=df.index, dtype="float64")
+    else:
+        raw = df.reindex(columns=[col], fill_value=0)[col]
+        series = pd.to_numeric(raw, errors="coerce").fillna(0.0).astype("float64")
+    return series
 
 
 def _sum_cols(df: pd.DataFrame, cols: list[str] | None) -> pd.Series:
@@ -67,32 +71,45 @@ def _sum_cols(df: pd.DataFrame, cols: list[str] | None) -> pd.Series:
     if not cols:
         return pd.Series(0.0, index=df.index, dtype="float64")
 
-    # Safely extract each column as a numeric, non-negative Series, then sum row-wise
+    # Safely extract each column as a numeric Series then sum row-wise
     series_list = [_get_numeric_series(df, c) for c in cols]
     return pd.concat(series_list, axis=1).sum(axis=1).astype("float64")
 
 
-# pylint: disable=too-many-arguments, too-many-locals
+def _column_has_data(df: pd.DataFrame, col: str | None) -> bool:
+    """Return True when the column exists and has at least one non-zero value."""
+    if col is None or col not in df.columns:
+        return False
+
+    series = pd.to_numeric(df[col], errors="coerce").fillna(0.0).astype("float64")
+    if series.empty or not series.notna().any():
+        return False
+
+    return not series.fillna(0).eq(0).all()
+
+
+# pylint: disable=too-many-arguments, too-many-locals, too-many-positional-arguments
 def add_energy_flows(
     df: pd.DataFrame,
     production_cols: list[str] | None = None,
     consumption_cols: list[str] | None = None,
-    battery_charge_col: str | None = None,
+    battery_cols: list[str] | None = None,
     production_is_positive: bool = False,
 ) -> pd.DataFrame:
     """Compute and add derived energy flow metrics to the DataFrame.
 
     This function aggregates production and consumption data, derives energy flow
     relationships such as grid feed-in, battery charging, and self-consumption,
-    and appends these computed columns to the given DataFrame.
+    and appends these computed columns to the given DataFrame. Columns that are
+    specified but missing or contain only null/zero values are ignored.
 
     Args:
         df: Input DataFrame containing production, consumption, and optionally
-            battery charge data.
+            battery power data.
         production_cols: list of column names representing production sources.
         consumption_cols: list of column names representing consumption sources.
-        battery_charge_col: optional column name for battery charging power. If None,
-            battery-related flows are set to zero.
+        battery_cols: optional column names representing signed battery power.
+            Positive values indicate charging, negative values indicate discharging.
         production_is_positive: Whether production values are already positive.
             If False, `production` is inverted before clipping.
 
@@ -106,47 +123,83 @@ def add_energy_flows(
     """
     df_flows = df.copy()
 
-    # Total production and consumption (returns pandas series with 0.0 for missing cols)
-    df_flows["production_total"] = _sum_cols(df_flows, production_cols)
-    df_flows["consumption_total"] = _sum_cols(df_flows, consumption_cols)
+    # Normalize production, consumption and battery columns by removing None entries
+    resolved_production_cols = [
+        col for col in (production_cols or []) if _column_has_data(df_flows, col)
+    ]
+    resolved_consumption_cols = [
+        col for col in (consumption_cols or []) if _column_has_data(df_flows, col)
+    ]
+    resolved_battery_cols = [
+        col for col in (battery_cols or []) if _column_has_data(df_flows, col)
+    ]
+
+    battery_power_series = _sum_cols(df_flows, resolved_battery_cols)
+    battery_charge_series = (
+        battery_power_series.reindex(df_flows.index).fillna(0.0).clip(lower=0.0)
+    )
+
+    # Compute total asset production
+    asset_production_cols: list[str] = []
+    for col in resolved_production_cols:
+        series = _get_numeric_series(
+            df_flows,
+            col,
+        )
+        asset_series = asset_production(
+            series,
+            production_is_positive=production_is_positive,
+        )
+        asset_col_name = f"{col}_asset_production"
+        df_flows[asset_col_name] = asset_series
+        asset_production_cols.append(asset_col_name)
+
+    df_flows["production_total"] = _sum_cols(df_flows, asset_production_cols)
 
-    # Surplus vs. consumption
+    # Compute total consumption
+    consumption_series_cols: list[str] = []
+    for col in resolved_consumption_cols:
+        df_flows[col] = _get_numeric_series(df_flows, col)
+        consumption_series_cols.append(col)
+
+    df_flows["consumption_total"] = _sum_cols(df_flows, consumption_series_cols)
+
+    # Surplus vs. consumption (production is already positive because of the above cleaning)
     df_flows["production_excess"] = production_excess(
         df_flows["production_total"],
         df_flows["consumption_total"],
-        production_is_positive=production_is_positive,
+        production_is_positive=True,
     )
 
     # Battery charging power (optional)
-    bat_in = _get_numeric_series(df_flows, battery_charge_col)
     df_flows["production_excess_in_bat"] = production_excess_in_bat(
         df_flows["production_total"],
         df_flows["consumption_total"],
-        bat_in,
-        production_is_positive=production_is_positive,
+        battery=battery_charge_series,
+        production_is_positive=True,
     )
 
     # Split excess into battery vs. grid
     df_flows["grid_feed_in"] = grid_feed_in(
         df_flows["production_total"],
         df_flows["consumption_total"],
-        bat_in,
-        production_is_positive=production_is_positive,
+        battery=battery_charge_series,
+        production_is_positive=True,
     )
 
     # If no production columns exist, set self-consumption metrics to zero
-    if production_cols:
+    if asset_production_cols:
         # Use total production for self-consumption instead of asset_production
         # (which may not exist)
         df_flows["production_self_use"] = production_self_consumption(
             df_flows["production_total"],
             df_flows["consumption_total"],
-            production_is_positive=production_is_positive,
+            production_is_positive=True,
         )
         df_flows["production_self_share"] = production_self_share(
             df_flows["production_total"],
             df_flows["consumption_total"],
-            production_is_positive=production_is_positive,
+            production_is_positive=True,
         )
     else:
         df_flows["production_self_use"] = 0.0