Bingran You
Mirror SkillsBench v1.1 as a benchmark task-tree dataset
d03762b
Raw
History Blame Contribute Delete
31.4 kB
import json
import math
import os
import numpy as np
import pytest
from scipy.optimize import Bounds, LinearConstraint, milp
from scipy.sparse import coo_matrix
OUTPUT_FILE = "/root/report.json"
CASE_FILE = "/root/network.json"
TOL_POWER_BALANCE_MW = 1e-2
TOL_RESERVE_MW = 1e-2
TOL_GENERATOR_MW = 1e-3
TOL_RAMP_MW = 1e-3
TOL_BINARY = 1e-5
TOL_COST_REL = 1e-4
TOL_COST_ABS = 1e-2
TOL_OPTIMALITY_GAP_REL = 0.10
ACCEPTED_STATUSES = {
"optimal",
"feasible",
"time_limit_feasible",
"suboptimal_feasible",
"heuristic_feasible",
}
REQUIRED_CONSTRAINT_CHECKS = {
"demand_balance",
"spinning_reserve",
"reserve_deliverability",
"generator_limits",
"must_run",
"ramping",
"minimum_up_down",
"startup_shutdown_logic",
"initial_conditions",
"renewable_limits",
"cost_consistency",
}
def _load_json(path: str) -> dict:
with open(path, "r", encoding="utf-8") as f:
data = json.load(f)
assert isinstance(data, dict), f"{path} must contain a JSON object"
return data
def _require_keys(obj: dict, keys: set[str], label: str) -> None:
missing = keys - set(obj)
assert not missing, f"{label} missing keys: {sorted(missing)}"
def _numeric_array(values, length: int, label: str) -> np.ndarray:
arr = np.asarray(values, dtype=float)
assert arr.shape == (length,), f"{label} must have length {length}, got shape {arr.shape}"
assert np.all(np.isfinite(arr)), f"{label} contains non-finite values"
return arr
def _parse_pglib_uc_case(case: dict) -> dict:
_require_keys(case, {"time_periods", "demand", "reserves", "thermal_generators", "renewable_generators"}, "case")
time_periods = int(case["time_periods"])
assert time_periods > 0, "time_periods must be positive"
demand = _numeric_array(case["demand"], time_periods, "case demand")
reserves = _numeric_array(case["reserves"], time_periods, "case reserves")
assert np.all(demand >= 0), "case demand contains negative values"
assert np.all(reserves >= 0), "case reserves contains negative values"
thermal_items = list(case["thermal_generators"].items())
renewable_items = list(case["renewable_generators"].items())
assert thermal_items, "case must contain thermal generators"
thermal = []
thermal_names = []
for key, gen in thermal_items:
_require_keys(
gen,
{
"power_output_minimum",
"power_output_maximum",
"ramp_up_limit",
"ramp_down_limit",
"ramp_startup_limit",
"ramp_shutdown_limit",
"time_up_minimum",
"time_down_minimum",
"power_output_t0",
"unit_on_t0",
"time_down_t0",
"time_up_t0",
"startup",
"piecewise_production",
},
f"thermal generator {key}",
)
name = str(gen.get("name", key))
thermal_names.append(name)
pmin = float(gen["power_output_minimum"])
pmax = float(gen["power_output_maximum"])
assert pmax >= pmin >= 0, f"{name} has invalid output range"
curve = sorted(
[(float(point["mw"]), float(point["cost"])) for point in gen["piecewise_production"]],
key=lambda x: x[0],
)
assert len(curve) >= 2, f"{name} must have at least two piecewise cost points"
assert abs(curve[0][0] - pmin) <= TOL_GENERATOR_MW, (
f"{name} first piecewise MW {curve[0][0]} does not match Pmin {pmin}"
)
assert abs(curve[-1][0] - pmax) <= TOL_GENERATOR_MW, (
f"{name} last piecewise MW {curve[-1][0]} does not match Pmax {pmax}"
)
assert all(curve[i + 1][0] > curve[i][0] for i in range(len(curve) - 1)), (
f"{name} piecewise MW points must be strictly increasing"
)
startups = sorted(
[(int(item["lag"]), float(item["cost"])) for item in gen["startup"]],
key=lambda x: x[0],
)
assert startups, f"{name} must have at least one startup tier"
assert all(lag > 0 for lag, _ in startups), f"{name} startup lags must be positive"
thermal.append(
{
"name": name,
"pmin": pmin,
"pmax": pmax,
"cap": pmax - pmin,
"ru": float(gen["ramp_up_limit"]),
"rd": float(gen["ramp_down_limit"]),
"su": float(gen["ramp_startup_limit"]),
"sd": float(gen["ramp_shutdown_limit"]),
"min_up": int(gen["time_up_minimum"]),
"min_down": int(gen["time_down_minimum"]),
"p0": float(gen["power_output_t0"]),
"u0": int(round(float(gen["unit_on_t0"]))),
"time_down_t0": int(gen["time_down_t0"]),
"time_up_t0": int(gen["time_up_t0"]),
"must_run": int(gen.get("must_run", 0)),
"startup": startups,
"piecewise": curve,
}
)
assert len(set(thermal_names)) == len(thermal_names), "thermal generator names are not unique"
renewable = []
renewable_names = []
for key, gen in renewable_items:
_require_keys(gen, {"power_output_minimum", "power_output_maximum"}, f"renewable generator {key}")
name = str(gen.get("name", key))
renewable_names.append(name)
pmin = _numeric_array(gen["power_output_minimum"], time_periods, f"{name} renewable minimum")
pmax = _numeric_array(gen["power_output_maximum"], time_periods, f"{name} renewable maximum")
assert np.all(pmax + TOL_GENERATOR_MW >= pmin), f"{name} renewable max below min"
renewable.append({"name": name, "pmin": pmin, "pmax": pmax})
assert len(set(renewable_names)) == len(renewable_names), "renewable generator names are not unique"
return {
"T": time_periods,
"demand": demand,
"reserves": reserves,
"thermal": thermal,
"renewable": renewable,
"thermal_names": thermal_names,
"renewable_names": renewable_names,
}
def _as_binary_array(values, length: int, label: str) -> np.ndarray:
arr = _numeric_array(values, length, label)
rounded = np.rint(arr).astype(int)
assert np.all(np.abs(arr - rounded) <= TOL_BINARY), f"{label} must be binary within tolerance"
assert np.all((rounded == 0) | (rounded == 1)), f"{label} contains values other than 0/1"
return rounded
def _extract_report_arrays(report: dict, parsed_case: dict) -> dict:
T = parsed_case["T"]
_require_keys(
report,
{"case_name", "summary", "thermal_generators", "renewable_generators", "hourly_summary", "constraint_check"},
"report",
)
assert isinstance(report["summary"], dict), "summary must be an object"
assert isinstance(report["thermal_generators"], list), "thermal_generators must be a list"
assert isinstance(report["renewable_generators"], list), "renewable_generators must be a list"
assert isinstance(report["hourly_summary"], list), "hourly_summary must be a list"
assert isinstance(report["constraint_check"], dict), "constraint_check must be an object"
thermal_entries = {}
for entry in report["thermal_generators"]:
assert isinstance(entry, dict), "each thermal_generators entry must be an object"
name = entry.get("name")
assert isinstance(name, str), "thermal generator entry missing string name"
assert name not in thermal_entries, f"duplicate thermal generator {name}"
thermal_entries[name] = entry
assert set(thermal_entries) == set(parsed_case["thermal_names"]), (
"thermal generator names do not match case data"
)
renewable_entries = {}
for entry in report["renewable_generators"]:
assert isinstance(entry, dict), "each renewable_generators entry must be an object"
name = entry.get("name")
assert isinstance(name, str), "renewable generator entry missing string name"
assert name not in renewable_entries, f"duplicate renewable generator {name}"
renewable_entries[name] = entry
assert set(renewable_entries) == set(parsed_case["renewable_names"]), (
"renewable generator names do not match case data"
)
G = len(parsed_case["thermal"])
R = len(parsed_case["renewable"])
commitment = np.zeros((G, T), dtype=int)
startup = np.zeros((G, T), dtype=int)
shutdown = np.zeros((G, T), dtype=int)
production = np.zeros((G, T), dtype=float)
reserve = np.zeros((G, T), dtype=float)
for g, gen in enumerate(parsed_case["thermal"]):
entry = thermal_entries[gen["name"]]
_require_keys(entry, {"commitment", "production_MW", "reserve_MW", "startup", "shutdown"}, gen["name"])
commitment[g] = _as_binary_array(entry["commitment"], T, f"{gen['name']} commitment")
startup[g] = _as_binary_array(entry["startup"], T, f"{gen['name']} startup")
shutdown[g] = _as_binary_array(entry["shutdown"], T, f"{gen['name']} shutdown")
production[g] = _numeric_array(entry["production_MW"], T, f"{gen['name']} production_MW")
reserve[g] = _numeric_array(entry["reserve_MW"], T, f"{gen['name']} reserve_MW")
renewable_production = np.zeros((R, T), dtype=float)
for r, gen in enumerate(parsed_case["renewable"]):
entry = renewable_entries[gen["name"]]
_require_keys(entry, {"production_MW"}, gen["name"])
renewable_production[r] = _numeric_array(entry["production_MW"], T, f"{gen['name']} production_MW")
assert len(report["hourly_summary"]) == T, f"hourly_summary must have {T} records"
for t, row in enumerate(report["hourly_summary"]):
assert isinstance(row, dict), f"hourly_summary[{t}] must be an object"
_require_keys(
row,
{
"hour",
"demand_MW",
"thermal_generation_MW",
"renewable_generation_MW",
"reserve_requirement_MW",
"scheduled_spinning_reserve_MW",
},
f"hourly_summary[{t}]",
)
assert int(row["hour"]) == t + 1, f"hourly_summary[{t}] hour must be {t + 1}"
return {
"commitment": commitment,
"startup": startup,
"shutdown": shutdown,
"thermal_production": production,
"thermal_reserve": reserve,
"renewable_production": renewable_production,
}
def _recompute_transition_indicators(arrays: dict, parsed_case: dict) -> tuple[np.ndarray, np.ndarray]:
u = arrays["commitment"]
G, T = u.shape
expected_startup = np.zeros_like(u)
expected_shutdown = np.zeros_like(u)
for g, gen in enumerate(parsed_case["thermal"]):
prev = gen["u0"]
for t in range(T):
expected_startup[g, t] = max(u[g, t] - prev, 0)
expected_shutdown[g, t] = max(prev - u[g, t], 0)
prev = u[g, t]
return expected_startup, expected_shutdown
def _recompute_hourly_summary(arrays: dict, parsed_case: dict) -> dict:
thermal_generation = arrays["thermal_production"].sum(axis=0)
renewable_generation = arrays["renewable_production"].sum(axis=0)
scheduled_reserve = arrays["thermal_reserve"].sum(axis=0)
demand_balance_violation = np.abs(thermal_generation + renewable_generation - parsed_case["demand"])
reserve_shortfall = np.maximum(parsed_case["reserves"] - scheduled_reserve, 0.0)
return {
"thermal_generation_MW": thermal_generation,
"renewable_generation_MW": renewable_generation,
"scheduled_spinning_reserve_MW": scheduled_reserve,
"max_demand_balance_violation_MW": float(np.max(demand_balance_violation)),
"max_reserve_shortfall_MW": float(np.max(reserve_shortfall)),
}
def _startup_cost_for_duration(generator: dict, offline_duration: int) -> float:
chosen_cost = generator["startup"][0][1]
for lag, cost in generator["startup"]:
if lag <= offline_duration:
chosen_cost = cost
else:
break
return chosen_cost
def _piecewise_total_cost_at_output(generator: dict, production_mw: float) -> float:
curve = generator["piecewise"]
assert production_mw >= curve[0][0] - TOL_GENERATOR_MW, (
f"{generator['name']} production below first cost point"
)
assert production_mw <= curve[-1][0] + TOL_GENERATOR_MW, (
f"{generator['name']} production above last cost point"
)
if production_mw <= curve[0][0]:
return curve[0][1]
for (mw0, cost0), (mw1, cost1) in zip(curve, curve[1:]):
if production_mw <= mw1 + TOL_GENERATOR_MW:
if production_mw >= mw1:
return cost1
slope = (cost1 - cost0) / (mw1 - mw0)
return cost0 + slope * (production_mw - mw0)
return curve[-1][1]
def _recompute_total_cost(arrays: dict, parsed_case: dict) -> float:
u = arrays["commitment"]
v = arrays["startup"]
production = arrays["thermal_production"]
total_cost = 0.0
for g, gen in enumerate(parsed_case["thermal"]):
offline_duration = gen["time_down_t0"] if gen["u0"] == 0 else 0
for t in range(parsed_case["T"]):
if v[g, t] == 1:
total_cost += _startup_cost_for_duration(gen, offline_duration)
if u[g, t] == 1:
total_cost += _piecewise_total_cost_at_output(gen, production[g, t])
offline_duration = 0
else:
offline_duration += 1
return float(total_cost)
def _solve_uc_benchmark_cost(case: dict) -> float:
parsed = _parse_pglib_uc_case(case)
T = parsed["T"]
G = len(parsed["thermal"])
R = len(parsed["renewable"])
lb = []
ub = []
integrality = []
objective = []
def add_var(lower: float, upper: float, integer: bool, cost: float = 0.0) -> int:
idx = len(lb)
lb.append(float(lower))
ub.append(float(upper))
integrality.append(1 if integer else 0)
objective.append(float(cost))
return idx
u = np.empty((G, T), dtype=int)
v = np.empty((G, T), dtype=int)
w = np.empty((G, T), dtype=int)
p = np.empty((G, T), dtype=int)
r = np.empty((G, T), dtype=int)
y: list[list[list[int]]] = [[[] for _ in range(T)] for _ in range(G)]
q = np.empty((R, T), dtype=int)
for g, gen in enumerate(parsed["thermal"]):
first_cost = gen["piecewise"][0][1]
force_online_until = 0
force_offline_until = 0
if gen["u0"] == 1 and gen["time_up_t0"] < gen["min_up"]:
force_online_until = gen["min_up"] - gen["time_up_t0"]
if gen["u0"] == 0 and gen["time_down_t0"] < gen["min_down"]:
force_offline_until = gen["min_down"] - gen["time_down_t0"]
for t in range(T):
lower, upper = 0.0, 1.0
if gen["must_run"] == 1 or t < force_online_until:
lower = upper = 1.0
if t < force_offline_until:
lower = upper = 0.0
u[g, t] = add_var(lower, upper, True, first_cost)
startup_cost = min(cost for _, cost in gen["startup"])
v[g, t] = add_var(0.0, 1.0, True, startup_cost)
w[g, t] = add_var(0.0, 1.0, True, 0.0)
p[g, t] = add_var(0.0, gen["cap"], False, 0.0)
r[g, t] = add_var(0.0, gen["cap"], False, 0.0)
for (mw0, cost0), (mw1, cost1) in zip(gen["piecewise"], gen["piecewise"][1:]):
width = mw1 - mw0
slope = (cost1 - cost0) / width
y[g][t].append(add_var(0.0, width, False, slope))
for i, gen in enumerate(parsed["renewable"]):
for t in range(T):
q[i, t] = add_var(gen["pmin"][t], gen["pmax"][t], False, 0.0)
row_idx = []
col_idx = []
data = []
cons_lb = []
cons_ub = []
def add_constraint(entries: list[tuple[int, float]], lower: float, upper: float) -> None:
row = len(cons_lb)
for col, val in entries:
if abs(val) > 0:
row_idx.append(row)
col_idx.append(col)
data.append(float(val))
cons_lb.append(float(lower))
cons_ub.append(float(upper))
for g, gen in enumerate(parsed["thermal"]):
cap = gen["cap"]
startup_reduction = max(gen["pmax"] - gen["su"], 0.0)
shutdown_reduction = max(gen["pmax"] - gen["sd"], 0.0)
p0_above_min = gen["u0"] * (gen["p0"] - gen["pmin"])
for t in range(T):
prev_u = gen["u0"] if t == 0 else u[g, t - 1]
entries = [(u[g, t], 1.0), (v[g, t], -1.0), (w[g, t], 1.0)]
rhs = float(prev_u) if t == 0 else 0.0
if t > 0:
entries.append((u[g, t - 1], -1.0))
add_constraint(entries, rhs, rhs)
add_constraint([(v[g, t], 1.0), (w[g, t], 1.0)], -math.inf, 1.0)
add_constraint([(p[g, t], 1.0)] + [(seg, -1.0) for seg in y[g][t]], 0.0, 0.0)
for seg_idx, seg in enumerate(y[g][t]):
width = gen["piecewise"][seg_idx + 1][0] - gen["piecewise"][seg_idx][0]
add_constraint([(seg, 1.0), (u[g, t], -width)], -math.inf, 0.0)
add_constraint(
[(p[g, t], 1.0), (r[g, t], 1.0), (u[g, t], -cap), (v[g, t], startup_reduction)],
-math.inf,
0.0,
)
if t < T - 1:
add_constraint(
[(p[g, t], 1.0), (r[g, t], 1.0), (u[g, t], -cap), (w[g, t + 1], shutdown_reduction)],
-math.inf,
0.0,
)
if t == 0:
add_constraint([(p[g, t], 1.0), (r[g, t], 1.0)], -math.inf, gen["ru"] + p0_above_min)
add_constraint([(p[g, t], -1.0)], -math.inf, gen["rd"] - p0_above_min)
else:
add_constraint(
[(p[g, t], 1.0), (r[g, t], 1.0), (p[g, t - 1], -1.0)],
-math.inf,
gen["ru"],
)
add_constraint([(p[g, t - 1], 1.0), (p[g, t], -1.0)], -math.inf, gen["rd"])
up_span = min(gen["min_up"], T - t)
if up_span > 0:
add_constraint([(u[g, k], -1.0) for k in range(t, t + up_span)] + [(v[g, t], up_span)], -math.inf, 0.0)
down_span = min(gen["min_down"], T - t)
if down_span > 0:
add_constraint([(u[g, k], 1.0) for k in range(t, t + down_span)] + [(w[g, t], down_span)], -math.inf, down_span)
for t in range(T):
entries = []
for g, gen in enumerate(parsed["thermal"]):
entries.append((p[g, t], 1.0))
entries.append((u[g, t], gen["pmin"]))
for i in range(R):
entries.append((q[i, t], 1.0))
add_constraint(entries, parsed["demand"][t], parsed["demand"][t])
add_constraint([(r[g, t], 1.0) for g in range(G)], parsed["reserves"][t], math.inf)
matrix = coo_matrix((data, (row_idx, col_idx)), shape=(len(cons_lb), len(lb))).tocsr()
result = milp(
c=np.asarray(objective, dtype=float),
integrality=np.asarray(integrality, dtype=int),
bounds=Bounds(np.asarray(lb), np.asarray(ub)),
constraints=LinearConstraint(matrix, np.asarray(cons_lb), np.asarray(cons_ub)),
options={"time_limit": 240.0, "mip_rel_gap": 0.05, "disp": False},
)
assert result.x is not None, f"benchmark MILP did not return a feasible incumbent: {result.message}"
x = result.x
bench_arrays = {
"commitment": np.rint(x[u]).astype(int),
"startup": np.rint(x[v]).astype(int),
"shutdown": np.rint(x[w]).astype(int),
"thermal_production": x[p] + np.array([[gen["pmin"] for _ in range(T)] for gen in parsed["thermal"]]) * np.rint(x[u]),
"thermal_reserve": np.maximum(x[r], 0.0),
"renewable_production": x[q],
}
return _recompute_total_cost(bench_arrays, parsed)
@pytest.fixture(scope="module")
def report():
assert os.path.exists(OUTPUT_FILE), f"Output file {OUTPUT_FILE} does not exist"
return _load_json(OUTPUT_FILE)
@pytest.fixture(scope="module")
def case_data():
return _load_json(CASE_FILE)
@pytest.fixture(scope="module")
def parsed_case(case_data):
return _parse_pglib_uc_case(case_data)
@pytest.fixture(scope="module")
def arrays(report, parsed_case):
return _extract_report_arrays(report, parsed_case)
@pytest.fixture(scope="module")
def recomputed_cost(arrays, parsed_case):
return _recompute_total_cost(arrays, parsed_case)
class TestSchema:
def test_required_schema(self, report, parsed_case):
summary = report["summary"]
_require_keys(
summary,
{
"solver_status",
"objective_cost",
"reported_mip_gap",
"time_periods",
"num_thermal_generators",
"num_renewable_generators",
"total_startups",
"total_shutdowns",
"max_demand_balance_violation_MW",
"max_reserve_shortfall_MW",
},
"summary",
)
assert summary["solver_status"] in ACCEPTED_STATUSES, (
f"solver_status must be one of {sorted(ACCEPTED_STATUSES)}"
)
gap = summary["reported_mip_gap"]
if gap is not None:
gap = float(gap)
assert math.isfinite(gap) and gap >= 0.0, "reported_mip_gap must be null or finite nonnegative"
assert int(summary["time_periods"]) == parsed_case["T"], "summary time_periods does not match case"
assert int(summary["num_thermal_generators"]) == len(parsed_case["thermal"]), (
"summary num_thermal_generators does not match case"
)
assert int(summary["num_renewable_generators"]) == len(parsed_case["renewable"]), (
"summary num_renewable_generators does not match case"
)
_require_keys(report["constraint_check"], REQUIRED_CONSTRAINT_CHECKS, "constraint_check")
for key in REQUIRED_CONSTRAINT_CHECKS:
assert report["constraint_check"][key] == "pass", f"constraint_check.{key} must be 'pass'"
def test_arrays_align_with_case(self, arrays, parsed_case):
T = parsed_case["T"]
assert arrays["commitment"].shape == (len(parsed_case["thermal"]), T)
assert arrays["startup"].shape == (len(parsed_case["thermal"]), T)
assert arrays["shutdown"].shape == (len(parsed_case["thermal"]), T)
assert arrays["thermal_production"].shape == (len(parsed_case["thermal"]), T)
assert arrays["thermal_reserve"].shape == (len(parsed_case["thermal"]), T)
assert arrays["renewable_production"].shape == (len(parsed_case["renewable"]), T)
class TestBinaryLogic:
def test_startup_shutdown_match_commitment(self, arrays, parsed_case):
expected_startup, expected_shutdown = _recompute_transition_indicators(arrays, parsed_case)
np.testing.assert_array_equal(arrays["startup"], expected_startup)
np.testing.assert_array_equal(arrays["shutdown"], expected_shutdown)
def test_no_simultaneous_startup_shutdown(self, arrays):
assert np.all(arrays["startup"] + arrays["shutdown"] <= 1), "unit starts and shuts down in same hour"
def test_reported_transition_counts(self, report, arrays):
summary = report["summary"]
assert int(summary["total_startups"]) == int(arrays["startup"].sum()), "total_startups mismatch"
assert int(summary["total_shutdowns"]) == int(arrays["shutdown"].sum()), "total_shutdowns mismatch"
class TestThermalFeasibility:
def test_must_run_units_online(self, arrays, parsed_case):
for g, gen in enumerate(parsed_case["thermal"]):
if gen["must_run"] == 1:
assert np.all(arrays["commitment"][g] == 1), f"{gen['name']} is must-run but not always online"
def test_output_limits_and_offline_zeroes(self, arrays, parsed_case):
u = arrays["commitment"]
production = arrays["thermal_production"]
reserve = arrays["thermal_reserve"]
for g, gen in enumerate(parsed_case["thermal"]):
assert np.all(reserve[g] >= -TOL_GENERATOR_MW), f"{gen['name']} has negative reserve"
offline = u[g] == 0
assert np.all(np.abs(production[g, offline]) <= TOL_GENERATOR_MW), (
f"{gen['name']} has production while offline"
)
assert np.all(np.abs(reserve[g, offline]) <= TOL_GENERATOR_MW), (
f"{gen['name']} has reserve while offline"
)
assert np.all(production[g] + TOL_GENERATOR_MW >= gen["pmin"] * u[g]), (
f"{gen['name']} below minimum output"
)
assert np.all(production[g] <= gen["pmax"] * u[g] + TOL_GENERATOR_MW), (
f"{gen['name']} above maximum output"
)
def test_deliverable_reserve_and_ramping(self, arrays, parsed_case):
u = arrays["commitment"]
v = arrays["startup"]
w = arrays["shutdown"]
production = arrays["thermal_production"]
reserve = arrays["thermal_reserve"]
T = parsed_case["T"]
for g, gen in enumerate(parsed_case["thermal"]):
p_above_min = production[g] - gen["pmin"] * u[g]
p0 = gen["u0"] * (gen["p0"] - gen["pmin"])
cap = gen["cap"]
startup_reduction = max(gen["pmax"] - gen["su"], 0.0)
shutdown_reduction = max(gen["pmax"] - gen["sd"], 0.0)
assert np.all(p_above_min >= -TOL_GENERATOR_MW), f"{gen['name']} production above-min is negative"
for t in range(T):
lhs = p_above_min[t] + reserve[g, t]
cap_rhs = cap * u[g, t] - startup_reduction * v[g, t]
assert lhs <= cap_rhs + TOL_GENERATOR_MW, f"{gen['name']} violates startup capacity at hour {t + 1}"
if t < T - 1:
shut_rhs = cap * u[g, t] - shutdown_reduction * w[g, t + 1]
assert lhs <= shut_rhs + TOL_GENERATOR_MW, (
f"{gen['name']} violates pre-shutdown capacity at hour {t + 1}"
)
previous = p0 if t == 0 else p_above_min[t - 1]
assert lhs - previous <= gen["ru"] + TOL_RAMP_MW, (
f"{gen['name']} violates ramp-up reserve deliverability at hour {t + 1}"
)
assert previous - p_above_min[t] <= gen["rd"] + TOL_RAMP_MW, (
f"{gen['name']} violates ramp-down limit at hour {t + 1}"
)
class TestSystemFeasibility:
def test_demand_and_reserve(self, arrays, parsed_case):
thermal = arrays["thermal_production"].sum(axis=0)
renewable = arrays["renewable_production"].sum(axis=0)
reserve = arrays["thermal_reserve"].sum(axis=0)
assert np.max(np.abs(thermal + renewable - parsed_case["demand"])) <= TOL_POWER_BALANCE_MW, (
"thermal plus renewable generation does not match demand"
)
assert np.min(reserve - parsed_case["reserves"]) >= -TOL_RESERVE_MW, "spinning reserve shortfall"
def test_renewable_limits(self, arrays, parsed_case):
production = arrays["renewable_production"]
for i, gen in enumerate(parsed_case["renewable"]):
assert np.all(production[i] + TOL_GENERATOR_MW >= gen["pmin"]), (
f"{gen['name']} renewable production below minimum"
)
assert np.all(production[i] <= gen["pmax"] + TOL_GENERATOR_MW), (
f"{gen['name']} renewable production above maximum"
)
fixed = np.abs(gen["pmax"] - gen["pmin"]) <= TOL_GENERATOR_MW
assert np.all(np.abs(production[i, fixed] - gen["pmin"][fixed]) <= TOL_GENERATOR_MW), (
f"{gen['name']} fixed renewable output not respected"
)
def test_hourly_summary_matches_schedule(self, report, arrays, parsed_case):
recomputed = _recompute_hourly_summary(arrays, parsed_case)
for t, row in enumerate(report["hourly_summary"]):
assert abs(float(row["demand_MW"]) - parsed_case["demand"][t]) <= TOL_POWER_BALANCE_MW
assert abs(float(row["reserve_requirement_MW"]) - parsed_case["reserves"][t]) <= TOL_RESERVE_MW
assert abs(float(row["thermal_generation_MW"]) - recomputed["thermal_generation_MW"][t]) <= TOL_POWER_BALANCE_MW
assert abs(float(row["renewable_generation_MW"]) - recomputed["renewable_generation_MW"][t]) <= TOL_POWER_BALANCE_MW
assert (
abs(float(row["scheduled_spinning_reserve_MW"]) - recomputed["scheduled_spinning_reserve_MW"][t])
<= TOL_RESERVE_MW
)
summary = report["summary"]
assert (
abs(float(summary["max_demand_balance_violation_MW"]) - recomputed["max_demand_balance_violation_MW"])
<= TOL_POWER_BALANCE_MW
)
assert (
abs(float(summary["max_reserve_shortfall_MW"]) - recomputed["max_reserve_shortfall_MW"])
<= TOL_RESERVE_MW
)
class TestMinimumUpDown:
def test_initial_and_in_horizon_obligations(self, arrays, parsed_case):
u = arrays["commitment"]
v = arrays["startup"]
w = arrays["shutdown"]
T = parsed_case["T"]
for g, gen in enumerate(parsed_case["thermal"]):
if gen["u0"] == 1 and gen["time_up_t0"] < gen["min_up"]:
remaining = min(T, gen["min_up"] - gen["time_up_t0"])
assert np.all(u[g, :remaining] == 1), f"{gen['name']} violates initial min-up obligation"
if gen["u0"] == 0 and gen["time_down_t0"] < gen["min_down"]:
remaining = min(T, gen["min_down"] - gen["time_down_t0"])
assert np.all(u[g, :remaining] == 0), f"{gen['name']} violates initial min-down obligation"
for t in range(T):
if v[g, t] == 1:
end = min(T, t + gen["min_up"])
assert np.all(u[g, t:end] == 1), f"{gen['name']} violates min-up after hour {t + 1}"
if w[g, t] == 1:
end = min(T, t + gen["min_down"])
assert np.all(u[g, t:end] == 0), f"{gen['name']} violates min-down after hour {t + 1}"
class TestCostAndQuality:
def test_reported_cost_matches_schedule(self, report, recomputed_cost):
reported_cost = float(report["summary"]["objective_cost"])
assert math.isfinite(reported_cost) and reported_cost > 0, "reported objective_cost must be positive finite"
assert abs(reported_cost - recomputed_cost) <= max(
TOL_COST_ABS,
TOL_COST_REL * max(1.0, abs(recomputed_cost)),
), f"reported objective {reported_cost} does not match recomputed cost {recomputed_cost}"
def test_cost_within_benchmark_gap(self, case_data, recomputed_cost):
benchmark_cost = _solve_uc_benchmark_cost(case_data)
assert recomputed_cost <= benchmark_cost * (1.0 + TOL_OPTIMALITY_GAP_REL) + TOL_COST_ABS, (
f"recomputed cost {recomputed_cost} exceeds benchmark {benchmark_cost} by more than "
f"{TOL_OPTIMALITY_GAP_REL:.0%}"
)