zasadni uprava LP planneru

backend/services/planning_engine.py

@@ -1057,6 +1057,44 @@ def solve_dispatch(
                 prob += ge_bat[t] == 0
                 prob += z_export[t] == 0
 
+        # Ekonomické guardy: ceny v objective nestačí proti maskám / terminal SoC.
+        ref_buy_horizon = min(float(s.buy_price) for s in slots)
+        min_spread = float(degradation_cost_effective)
+        hp_rated_w = float(heat_pump.rated_heating_power_w)
+        soc_headroom_wh = max(
+            2000.0, 0.05 * float(battery.soc_max_wh)
+        )
+        for t in range(T):
+            s = slots[t]
+            buy_t = float(s.buy_price)
+            sell_t = float(s.sell_price)
+            load_t = float(s.load_baseline_w)
+            ev_cap_t = sum(
+                float(vehicles[e].max_charge_power_w)
+                for e in range(EV)
+                if (e == 0 and s.ev1_connected) or (e == 1 and s.ev2_connected)
+            )
+            pv_surplus_w = max(
+                0.0,
+                float(s.pv_a_forecast_w) + float(s.pv_b_forecast_w) - load_t,
+            )
+            # Ztrátový export FVE (sell ≪ buy): zakázat jen pokud jde energii do baterie.
+            # Výjimky: plná baterie (ventil), neriťitelné pv_b s přebytkem.
+            if sell_t < buy_t - min_spread:
+                block_loss_pv_export = not (
+                    float(s.pv_b_forecast_w) > 0 and pv_surplus_w > 0
+                )
+                if t == 0 and current_soc_wh >= float(battery.soc_max_wh) - soc_headroom_wh:
+                    block_loss_pv_export = False
+                if block_loss_pv_export:
+                    prob += ge_pv[t] == 0
+            # Drahý nákup oproti horizontu: import jen na load + EV + TČ, ne na grid-nabíjení.
+            if buy_t >= 0 and buy_t > ref_buy_horizon + min_spread:
+                prob += gi[t] <= load_t + ev_cap_t + hp_rated_w
+            # Anti souběžný vývoz FVE + významný import (mikrocyklus).
+            if buy_t > sell_t + min_spread and pv_surplus_w > 0:
+                prob += ge_pv[t] <= pv_surplus_w
+
     # Deadline constraints pro EV
     for e, session in enumerate(ev_sessions):
         if session and session.target_deadline and session.energy_needed_wh > 0:

backend/tests/test_planning_charge_slot_selection.py

@@ -3,11 +3,11 @@
 Logika je v DB: ems.fn_load_planning_slots_full. Kopie algoritmu pro unit testy bez PG.
 
 Charge mask:
-  A) PV-surplus: sell_price ASC, dokud PV nepokryje charge target.
-  B) Non-PV: AM/PM 50/50, OTE-first, buy_price ASC.
+  A) PV-surplus: store_score DESC, dokud PV nepokryje charge target.
+  B) Non-PV: AM/PM, OTE-first, buy≤ref+degrad, lookahead, cap 6 slotů.
 
 Discharge-export mask:
-  ref_buy = min(buy) mezi allow_charge sloty (arbitráž mezi sloty, ne sell vs buy ve stejném).
+  ref_buy = min(buy) celého horizontu.
   Top sloty dle sell_price desc kde sell > ref_buy + degradation.
 """
 
@@ -21,6 +21,31 @@ from zoneinfo import ZoneInfo
 from services.planning_engine import INTERVAL_H, PlanningSlot
 
 _PRAGUE = ZoneInfo("Europe/Prague")
+_LOOKAHEAD_SLOTS = 4
+_GRID_CHARGE_CAP_AM = 6
+_GRID_CHARGE_CAP_PM = 6
+_BUY_LOOKAHEAD_EPS = 0.05
+
+
+def _future_sell(slots: list[PlanningSlot], t: int) -> float:
+    tail = [float(slots[i].sell_price) for i in range(t + 1, len(slots))]
+    return max(tail) if tail else float(slots[t].sell_price)
+
+
+def _buy_min_next_n(slots: list[PlanningSlot], t: int, n: int = _LOOKAHEAD_SLOTS) -> float | None:
+    tail = [
+        float(slots[i].buy_price)
+        for i in range(t + 1, min(t + 1 + n, len(slots)))
+    ]
+    return min(tail) if tail else None
+
+
+def _store_score(slots: list[PlanningSlot], t: int) -> float:
+    s = slots[t]
+    buy = float(s.buy_price)
+    sell = float(s.sell_price)
+    fso = _future_sell(slots, t)
+    return fso - sell - max(0.0, buy - sell)
 
 
 def _select_charge_slots(
@@ -39,12 +64,21 @@ def _select_charge_slots(
     if energy_to_fill <= 0:
         return set(range(len(slots)))
 
+    reserve_wh = float(getattr(battery, "reserve_soc_wh", 0) or 0)
+    degrad = float(getattr(battery, "degradation_cost_czk_kwh", 0.15) or 0.15)
+    ref_buy = min(float(s.buy_price) for s in slots)
+
     eta = float(getattr(battery, "charge_efficiency", 1.0) or 1.0)
     max_p_w = float(getattr(battery, "max_charge_power_w", 0.0) or 0.0)
     per_slot_full_wh = max_p_w * eta * INTERVAL_H
-    charge_target_wh = max(energy_to_fill, 0) * charge_buf
+    if current_soc_wh >= reserve_wh:
+        charge_target_wh = max(energy_to_fill, 0.0)
+    else:
+        charge_target_wh = min(
+            max(energy_to_fill, 0.0) * charge_buf,
+            max(energy_to_fill, 0.0),
+        )
 
-    # AM/PM budget
     n_am = sum(1 for s in slots if _prague_hour(s) < 12)
     n_pm = len(slots) - n_am
     if n_am <= 0:
@@ -59,56 +93,67 @@ def _select_charge_slots(
 
     selected: set[int] = set()
 
-    # A) PV-surplus: cheapest sell_price first
+    # A) PV-surplus: highest store_score first
     pv_candidates: list[tuple[int, float, float]] = []
     for t, s in enumerate(slots):
         pv_surplus_w = max(0, s.pv_a_forecast_w + s.pv_b_forecast_w - s.load_baseline_w)
-        if pv_surplus_w > 0:
-            pv_candidates.append((t, float(s.sell_price), float(pv_surplus_w)))
+        if pv_surplus_w > 0 and float(s.sell_price) >= float(s.buy_price) - degrad:
+            pv_candidates.append((t, _store_score(slots, t), float(pv_surplus_w)))
 
-    pv_candidates.sort(key=lambda x: (x[1], x[0]))
+    pv_candidates.sort(key=lambda x: (-x[1], x[0]))
     cum = 0.0
-    for t, _sell, pv_surplus_w in pv_candidates:
+    for t, _score, pv_surplus_w in pv_candidates:
         if cum >= charge_target_wh:
             break
         selected.add(t)
         cum += min(pv_surplus_w, max_p_w) * eta * INTERVAL_H
 
-    # B) Non-PV grid charge — jen spot nákup (u fixed je buy všude stejný → jen FVE)
     if purchase_pricing_mode == "fixed":
         return selected
 
-    # B) Non-PV: AM budget (OTE-first)
+    def _grid_b_ok(t: int) -> bool:
+        s = slots[t]
+        if max(0, s.pv_a_forecast_w + s.pv_b_forecast_w - s.load_baseline_w) > 0:
+            return False
+        buy = float(s.buy_price)
+        sell = float(s.sell_price)
+        if buy > ref_buy + degrad:
+            return False
+        nxt = _buy_min_next_n(slots, t)
+        if nxt is not None and buy > nxt + _BUY_LOOKAHEAD_EPS:
+            return False
+        return True
+
+    # B) AM
     am_candidates = [
         (t, getattr(slots[t], "is_predicted_price", False), float(slots[t].buy_price))
         for t in range(len(slots))
-        if t not in selected
-        and max(0, slots[t].pv_a_forecast_w + slots[t].pv_b_forecast_w - slots[t].load_baseline_w) <= 0
-        and _prague_hour(slots[t]) < 12
+        if t not in selected and _grid_b_ok(t) and _prague_hour(slots[t]) < 12
     ]
     am_candidates.sort(key=lambda x: (int(x[1]), x[2], x[0]))
     cum = 0.0
+    grid_am = 0
     for t, _pred, _price in am_candidates:
-        if cum >= chg_am or per_slot_full_wh <= 0:
+        if cum >= chg_am or per_slot_full_wh <= 0 or grid_am >= _GRID_CHARGE_CAP_AM:
             break
         selected.add(t)
         cum += per_slot_full_wh
+        grid_am += 1
 
-    # B) Non-PV: PM budget (OTE-first)
     pm_candidates = [
         (t, getattr(slots[t], "is_predicted_price", False), float(slots[t].buy_price))
         for t in range(len(slots))
-        if t not in selected
-        and max(0, slots[t].pv_a_forecast_w + slots[t].pv_b_forecast_w - slots[t].load_baseline_w) <= 0
-        and _prague_hour(slots[t]) >= 12
+        if t not in selected and _grid_b_ok(t) and _prague_hour(slots[t]) >= 12
     ]
     pm_candidates.sort(key=lambda x: (int(x[1]), x[2], x[0]))
     cum = 0.0
+    grid_pm = 0
     for t, _pred, _price in pm_candidates:
-        if cum >= chg_pm or per_slot_full_wh <= 0:
+        if cum >= chg_pm or per_slot_full_wh <= 0 or grid_pm >= _GRID_CHARGE_CAP_PM:
             break
         selected.add(t)
         cum += per_slot_full_wh
+        grid_pm += 1
 
     return selected
 
@@ -138,13 +183,7 @@ def _select_discharge_export_slots(
     per_slot_wh = max_p_w * eta * INTERVAL_H
     discharge_target_wh = exportable_wh * discharge_buf
 
-    if charge_slots is None:
-        charge_slots = _select_charge_slots(slots, battery, current_soc_wh)
-
-    ref_buy = min(
-        (float(slots[t].buy_price) for t in charge_slots),
-        default=min(float(s.buy_price) for s in slots),
-    )
+    ref_buy = min(float(s.buy_price) for s in slots)
 
     if purchase_pricing_mode == "fixed":
         sell_min = degrad
@@ -182,9 +221,12 @@ def _slot(
     load: int = 2_000,
     hour_utc: int = 12,
     predicted: bool = False,
+    interval_start: datetime | None = None,
 ) -> PlanningSlot:
+    if interval_start is None:
+        interval_start = datetime(2026, 5, 19, hour_utc, 0, tzinfo=timezone.utc)
     return PlanningSlot(
-        interval_start=datetime(2026, 5, 19, hour_utc, 0, tzinfo=timezone.utc),
+        interval_start=interval_start,
         buy_price=buy,
         sell_price=sell,
         pv_a_forecast_w=0,
@@ -203,6 +245,7 @@ def _battery(
     uc_wh: float = 64_000.0,
     soc_max_pct: float = 95.0,
     min_soc_pct: float = 10.0,
+    reserve_soc_pct: float = 20.0,
     max_charge_w: float = 18_000.0,
     max_discharge_w: float = 18_000.0,
     charge_eff: float = 0.95,
@@ -213,6 +256,7 @@ def _battery(
     return SimpleNamespace(
         usable_capacity_wh=uc,
         min_soc_wh=min_soc_pct / 100.0 * uc,
+        reserve_soc_wh=reserve_soc_pct / 100.0 * uc,
         soc_max_wh=soc_max_pct / 100.0 * uc,
         max_charge_power_w=max_charge_w,
         max_discharge_power_w=max_discharge_w,
@@ -239,81 +283,82 @@ class SelectChargeSlotsTests(unittest.TestCase):
         )
         self.assertEqual(out, set(range(3)))
 
-    def test_pv_surplus_cheapest_sell_price_selected(self) -> None:
-        """PV-surplus sloty s nejnižší sell_price se vybírají přednostně."""
+    def test_pv_surplus_high_store_score_selected(self) -> None:
+        """Slot s vyšším store_score (lepší uložení vs export) má přednost."""
         slots = [
-            _slot(buy=1.0, sell=2.0, pv=8_000, load=2_000),
-            _slot(buy=1.0, sell=5.0, pv=8_000, load=2_000),
-            _slot(buy=1.0, sell=3.0, pv=8_000, load=2_000),
+            _slot(buy=1.5, sell=0.01, pv=8_000, load=2_000, hour_utc=8),
+            _slot(buy=1.5, sell=0.50, pv=8_000, load=2_000, hour_utc=9),
+            _slot(buy=0.5, sell=0.40, pv=8_000, load=2_000, hour_utc=10),
         ]
         battery = _battery(
             charge_buf=1.3, uc_wh=1_000.0, soc_max_pct=100.0, max_charge_w=6_000.0
         )
         out = _select_charge_slots(slots, battery, current_soc_wh=0.0)
-        self.assertIn(0, out, "Cheapest sell_price PV slot must be selected")
-        self.assertNotIn(1, out, "Expensive sell_price PV slot should be excluded")
+        self.assertIn(2, out, "Slot s lepší marží (nižší buy) má být vybrán")
+        self.assertNotIn(0, out, "Ztrátový sell≪buy slot nemá grid charge z masky A")
 
     def test_non_pv_slots_selected_with_am_pm_budget(self) -> None:
-        """Non-PV sloty se vybírají dle buy_price v rámci AM/PM rozpočtu."""
+        """Levný PM slot; AM s dražším buy než min v lookahead může být vynechán."""
         slots = [
-            _slot(buy=0.5, hour_utc=4),   # AM slot, cheap
-            _slot(buy=3.0, hour_utc=5),   # AM slot, expensive
-            _slot(buy=0.4, hour_utc=14),  # PM slot, cheap
-            _slot(buy=9.9, hour_utc=15),  # PM slot, expensive
+            _slot(buy=0.5, hour_utc=4),
+            _slot(buy=3.0, hour_utc=5),
+            _slot(buy=0.4, hour_utc=14),
+            _slot(buy=9.9, hour_utc=15),
         ]
         battery = _battery(
             charge_buf=1.3, uc_wh=5_000.0, soc_max_pct=100.0, max_charge_w=18_000.0
         )
         out = _select_charge_slots(slots, battery, current_soc_wh=0.0)
-        self.assertIn(0, out, "Cheapest AM slot must be selected")
-        self.assertIn(2, out, "Cheapest PM slot must be selected")
+        self.assertIn(2, out, "Nejlevnější buy v horizontu (PM) musí být vybrán")
+
+    def test_vt_before_nt_skips_expensive_pm_slot(self) -> None:
+        """Regrese home-01: 12:45 VT drahý, za 15 min NT levný → PM grid charge ne v 12:45."""
+        base = datetime(2026, 5, 21, 10, 45, tzinfo=timezone.utc)
+        slots = [
+            _slot(
+                buy=1.49,
+                sell=-0.04,
+                pv=0,
+                load=3_500,
+                interval_start=base,
+            ),
+            _slot(
+                buy=0.86,
+                sell=0.01,
+                pv=0,
+                load=3_500,
+                interval_start=base + timedelta(minutes=15),
+            ),
+            _slot(
+                buy=0.86,
+                sell=0.01,
+                pv=0,
+                load=3_500,
+                interval_start=base + timedelta(minutes=30),
+            ),
+        ]
+        battery = _battery(uc_wh=64_000.0)
+        soc = 0.31 * battery.usable_capacity_wh
+        out = _select_charge_slots(slots, battery, current_soc_wh=soc)
+        self.assertNotIn(0, out, "VT slot před levným NT nesmí dostat grid charge z masky B")
+        self.assertIn(1, out, "NT slot může být vybrán")
 
     def test_ote_slots_prioritized_over_predicted(self) -> None:
-        """OTE sloty (is_predicted_price=false) mají přednost před predikovanými."""
+        """Při stejné ceně má OTE (is_predicted=false) přednost před predikovaným."""
         slots = [
-            _slot(buy=3.56, hour_utc=13, predicted=False),  # OTE, dražší
-            _slot(buy=2.00, hour_utc=13, predicted=True),   # predicted, levnější
+            _slot(buy=2.00, sell=2.0, hour_utc=13, predicted=False),
+            _slot(buy=2.00, sell=2.0, hour_utc=13, predicted=True),
         ]
         battery = _battery(
             charge_buf=1.3, uc_wh=3_000.0, soc_max_pct=100.0, max_charge_w=18_000.0
         )
         out = _select_charge_slots(slots, battery, current_soc_wh=0.0)
-        self.assertIn(0, out, "OTE slot must be selected even if pricier than predicted")
-
-    def test_does_not_exclude_slot_just_because_pv_below_load(self) -> None:
-        """Regrese: sloty bez PV-surplus se vybírají přes AM/PM grid budget."""
-        slots = [
-            _slot(buy=0.4, pv=3_320, load=3_747, hour_utc=13),
-            _slot(buy=0.42, pv=2_116, load=3_747, hour_utc=13),
-            _slot(buy=0.44, pv=1_649, load=3_747, hour_utc=13),
-            _slot(buy=0.47, pv=1_276, load=3_747, hour_utc=13),
-        ]
-        battery = _battery()
-        out = _select_charge_slots(slots, battery, current_soc_wh=0.2 * battery.usable_capacity_wh)
-        for idx in (0, 1, 2, 3):
-            self.assertIn(idx, out)
-
-    def test_long_horizon_pv_surplus_does_not_exhaust_grid_budget(self) -> None:
-        """Regrese: v 96h horizontu nesmí PV-surplus sloty „vyžrat" grid rozpočet."""
-        cheap_grid = [_slot(buy=0.4 + 0.01 * i, pv=0, load=2_000) for i in range(40)]
-        pv_days = [_slot(buy=1.5, sell=1.5, pv=10_000, load=2_000) for _ in range(100)]
-        slots = cheap_grid + pv_days
-        battery = _battery(
-            charge_buf=1.3, uc_wh=64_000.0, soc_max_pct=95.0, max_charge_w=18_000.0
-        )
-        out = _select_charge_slots(slots, battery, current_soc_wh=0.2 * battery.usable_capacity_wh)
-        grid_selected = sum(1 for i in range(len(cheap_grid)) if i in out)
-        self.assertGreaterEqual(
-            grid_selected,
-            5,
-            "V dlouhém horizontu s mnoha PV-surplus sloty musí zůstat dostatek "
-            "grid slotů povolených pro nabíjení z levného importu.",
-        )
+        self.assertIn(0, out)
+        self.assertNotIn(1, out)
 
 
 class SelectDischargeExportSlotsTests(unittest.TestCase):
     def test_evening_sell_allowed_when_cheaper_than_ref_charge_buy(self) -> None:
-        """Regrese home-01: večer sell 3.3 > ref_buy 0.5 + degrad i když buy ve slotu je 5.6."""
         slots = [
             _slot(buy=0.50, sell=-0.30, hour_utc=6),
             _slot(buy=0.51, sell=-0.29, hour_utc=7),
@@ -325,8 +370,7 @@ class SelectDischargeExportSlotsTests(unittest.TestCase):
         discharge = _select_discharge_export_slots(
             slots, battery, current_soc_wh=0.2 * battery.usable_capacity_wh, charge_slots=charge
         )
-        self.assertIn(0, charge)
-        self.assertIn(2, discharge, "Evening sell must qualify vs ref buy, not same-slot buy")
+        self.assertIn(2, discharge)
         self.assertIn(3, discharge)
 
     def test_export_excluded_when_sell_below_ref_buy_plus_degradation(self) -> None:
@@ -335,16 +379,13 @@ class SelectDischargeExportSlotsTests(unittest.TestCase):
             _slot(buy=4.00, sell=0.50, hour_utc=18),
         ]
         battery = _battery(uc_wh=10_000.0, discharge_buf=2.0)
-        charge = _select_charge_slots(slots, battery, current_soc_wh=0.0)
         discharge = _select_discharge_export_slots(
-            slots, battery, current_soc_wh=0.0, charge_slots=charge
+            slots, battery, current_soc_wh=0.0
         )
-        self.assertNotIn(1, discharge, "sell 0.5 < ref 0.4 + 0.15")
+        self.assertNotIn(1, discharge)
 
 
 class FixedPurchasePricingTests(unittest.TestCase):
-    """purchase_pricing_mode=fixed: žádné grid CHARGE, export dle sell."""
-
     def test_fixed_skips_non_pv_grid_charge_slots(self) -> None:
         slots = [
             _slot(buy=6.35, sell=2.0, hour_utc=14, load=500),
@@ -357,7 +398,7 @@ class FixedPurchasePricingTests(unittest.TestCase):
             current_soc_wh=0.4 * battery.usable_capacity_wh,
             purchase_pricing_mode="fixed",
         )
-        self.assertEqual(out, set(), "fixed buy must not enable non-PV grid charge")
+        self.assertEqual(out, set())
 
     def test_fixed_allows_discharge_on_high_sell(self) -> None:
         slots = [

backend/tests/test_planning_dispatch_milp.py

@@ -1290,5 +1290,74 @@ class TerminalSocShadowTests(unittest.TestCase):
         )
 
 
+class SpreadGuardHome01EconomicsTests(unittest.TestCase):
+    """Regrese: sell≪buy (VT) nesmí vést k PV exportu + masivnímu grid importu ve stejném slotu."""
+
+    def test_loss_making_morning_and_vt_slot_avoid_export_and_grid_charge(self) -> None:
+        from test_planning_charge_slot_selection import (
+            _battery as mask_battery,
+            _select_charge_slots,
+            _select_discharge_export_slots,
+        )
+
+        base = datetime(2026, 5, 21, 8, 0, tzinfo=timezone.utc)
+        raw: list[tuple[float, float, int, int]] = [
+            (1.55, 0.01, 6_000, 2_000),
+            (1.55, 0.01, 6_500, 2_000),
+            (1.49, -0.04, 0, 3_500),
+            (0.86, 0.01, 0, 3_500),
+            (0.86, 0.01, 0, 3_500),
+            (0.86, 0.01, 5_000, 2_000),
+        ]
+        slots: list[PlanningSlot] = []
+        for i, (buy, sell, pv, load) in enumerate(raw):
+            slots.append(
+                PlanningSlot(
+                    interval_start=base + timedelta(minutes=15 * i),
+                    buy_price=buy,
+                    sell_price=sell,
+                    pv_a_forecast_w=0,
+                    pv_b_forecast_w=pv,
+                    load_baseline_w=load,
+                    ev1_connected=False,
+                    ev2_connected=False,
+                    is_predicted_price=False,
+                )
+            )
+        mb = mask_battery(uc_wh=64_000.0)
+        soc0 = 0.31 * mb.usable_capacity_wh
+        charge = _select_charge_slots(slots, mb, soc0)
+        discharge = _select_discharge_export_slots(slots, mb, soc0)
+        for t, s in enumerate(slots):
+            s.allow_charge = t in charge
+            s.allow_discharge_export = t in discharge
+
+        battery = _battery(uc_wh=64_000.0, terminal_soc_value_factor=0.9)
+        hp = SimpleNamespace(rated_heating_power_w=0, tuv_min_temp_c=45.0, tuv_target_temp_c=55.0)
+        grid = SimpleNamespace(max_import_power_w=20_000, max_export_power_w=20_000)
+        vehicles = [
+            SimpleNamespace(max_charge_power_w=0, battery_capacity_kwh=1.0, default_target_soc_pct=80.0),
+            SimpleNamespace(max_charge_power_w=0, battery_capacity_kwh=1.0, default_target_soc_pct=80.0),
+        ]
+        results, _ms, _ = solve_dispatch(
+            slots,
+            battery,
+            hp,
+            grid,
+            [None, None],
+            vehicles,
+            soc0,
+            50.0,
+            operating_mode="AUTO",
+        )
+        self.assertEqual(len(results), len(slots))
+        morning = results[0]
+        vt_before_nt = results[2]
+        self.assertLessEqual(morning.grid_setpoint_w, slots[0].load_baseline_w + 500)
+        self.assertNotEqual(morning.export_mode, "PV_SURPLUS")
+        self.assertLessEqual(vt_before_nt.grid_setpoint_w, 4_000)
+        self.assertLessEqual(vt_before_nt.battery_setpoint_w, 2_000)
+
+
 if __name__ == "__main__":
     unittest.main()