fix: realtime + KPI power dedup by station prefix (v1.4.3)

Realtime endpoint was summing ALL device power readings, causing
double-counting when multiple devices share the same Sungrow station.
E.g. 10 devices × station-level power = 5x inflated total.

Fix: GROUP BY station prefix (first 3 chars of device name) and
take MAX per station. Same fix applied to KPI daily_generation.

Result: 5,550 kW → 1,931 kW (matches iSolarCloud's 2,049 kW
within the 15-min collection timing window).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend/app/api/v1/dashboard.py

@@ -105,21 +105,58 @@ async def get_overview(db: AsyncSession = Depends(get_db), user: User = Depends(
 
 @router.get("/realtime")
 async def get_realtime_data(db: AsyncSession = Depends(get_db), user: User = Depends(get_current_user)):
-    """实时功率数据 - 获取最近的采集数据"""
+    """实时功率数据 - 获取最近的采集数据，按站去重防止重复计数"""
     now = datetime.now(timezone.utc)
-    five_min_ago = now - timedelta(minutes=20)
+    window_start = now - timedelta(minutes=20)
-
-    latest_q = await db.execute(
-        select(EnergyData).where(
-            and_(EnergyData.timestamp >= five_min_ago, EnergyData.data_type == "power")
-        ).order_by(EnergyData.timestamp.desc()).limit(50)
-    )
-    data_points = latest_q.scalars().all()
 
+    # Get latest power per station (dedup by device name prefix)
+    # Sungrow collectors report station-level power, so multiple devices
+    # sharing the same station (AP1xx = Phase 1, AP2xx = Phase 2) report
+    # identical values. GROUP BY station prefix and take MAX to avoid
+    # double-counting.
+    from sqlalchemy import text as sa_text
     pv_ids = await _get_pv_device_ids(db)
     hp_ids = await _get_hp_device_ids(db)
-    pv_power = sum(d.value for d in data_points if d.device_id in pv_ids)
+
-    heatpump_power = sum(d.value for d in data_points if d.device_id in hp_ids)
+    # PV power: dedup by station prefix
+    if pv_ids:
+        pv_q = await db.execute(
+            select(
+                func.substring(Device.name, 1, 3).label("station"),
+                func.max(EnergyData.value).label("power"),
+            ).select_from(EnergyData).join(
+                Device, EnergyData.device_id == Device.id
+            ).where(
+                and_(
+                    EnergyData.timestamp >= window_start,
+                    EnergyData.data_type == "power",
+                    EnergyData.device_id.in_(pv_ids),
+                )
+            ).group_by(sa_text("1"))
+        )
+        pv_power = sum(row[1] or 0 for row in pv_q.all())
+    else:
+        pv_power = 0
+
+    # Heat pump power: dedup by station prefix
+    if hp_ids:
+        hp_q = await db.execute(
+            select(
+                func.substring(Device.name, 1, 3).label("station"),
+                func.max(EnergyData.value).label("power"),
+            ).select_from(EnergyData).join(
+                Device, EnergyData.device_id == Device.id
+            ).where(
+                and_(
+                    EnergyData.timestamp >= window_start,
+                    EnergyData.data_type == "power",
+                    EnergyData.device_id.in_(hp_ids),
+                )
+            ).group_by(sa_text("1"))
+        )
+        heatpump_power = sum(row[1] or 0 for row in hp_q.all())
+    else:
+        heatpump_power = 0
 
     return {
         "timestamp": str(now),

backend/app/api/v1/kpi.py

@@ -35,31 +35,32 @@ async def get_solar_kpis(db: AsyncSession = Depends(get_db), user: User = Depend
             "total_rated_kw": 0, "daily_generation_kwh": 0,
         }
 
-    # Get latest daily_energy per PV device for today
+    # Get latest daily_energy per station (dedup by device name prefix)
+    # Sungrow collectors report station-level data per device, so multiple
+    # devices sharing the same station report identical values.
+    # Group by station prefix (first 3 chars of name, e.g. "AP1" vs "AP2")
+    # and take MAX per station to avoid double-counting.
+    from sqlalchemy import text as sa_text
     daily_gen_q = await db.execute(
         select(
-            EnergyData.device_id,
+            func.substring(Device.name, 1, 3).label("station"),
             func.max(EnergyData.value).label("max_energy"),
+        ).select_from(EnergyData).join(
+            Device, EnergyData.device_id == Device.id
         ).where(
             and_(
                 EnergyData.timestamp >= today_start,
                 EnergyData.data_type == "daily_energy",
                 EnergyData.device_id.in_(pv_ids),
             )
-        ).group_by(EnergyData.device_id)
+        ).group_by(sa_text("1"))
     )
 
-    # Check if values are station-level (all identical) or device-level
     daily_values = daily_gen_q.all()
     if not daily_values:
         daily_generation_kwh = 0
     else:
-        values = [row[1] or 0 for row in daily_values]
+        daily_generation_kwh = sum(row[1] or 0 for row in daily_values)
-        # If all values are identical, it's station-level data — use max (not sum)
-        if len(set(values)) == 1 and len(values) > 1:
-            daily_generation_kwh = values[0]
-        else:
-            daily_generation_kwh = sum(values)
 
     # Performance Ratio (PR) = actual output / (rated capacity * peak sun hours)
     # Approximate peak sun hours from time of day (simplified)