feat(replay): execute periodic diagnosis and hydraulic checks

This commit is contained in:
2026-11-16 15:13:22 +08:00
parent 4c441fcbf2
commit 389892cb3f
22 changed files with 1934 additions and 12 deletions
+167
View File
@@ -0,0 +1,167 @@
"""Rolling hydraulic checks using immutable INP and observed input only.
Each observed timestamp is a steady hydraulic snapshot, matching the online
recalculation's semantics. This module never opens a business/telemetry database.
"""
from datetime import datetime
from math import isfinite
from pathlib import Path
from statistics import mean
from tempfile import TemporaryDirectory
from zoneinfo import ZoneInfo
import re
def _sampling_seconds(device: dict) -> int:
frequency = str(device.get("transmission_frequency") or "")
clock = re.fullmatch(r"(\d+):(\d+):(\d+)", frequency)
minutes = re.fullmatch(r"(\d+)\s*(?:min|分钟)", frequency)
return max(1, int(clock[1]) * 3600 + int(clock[2]) * 60 + int(clock[3]) if clock
else int(minutes[1]) * 60 if minutes else 900)
def pressure_metres(value: float, unit: str) -> float | None:
factors = {"m": 1.0, "mH2O": 1.0, "MPa": 101.9716213,
"kPa": .1019716213, "bar": 10.19716213}
return value * factors[unit] if unit in factors else None
def flow_cms(value: float, unit: str) -> float | None:
factors = {"m3/s": 1.0, "m³/s": 1.0, "m3/h": 1 / 3600,
"m³/h": 1 / 3600, "L/s": .001, "l/s": .001,
"LPS": .001, "CMH": 1 / 3600, "CMS": 1.0}
return value * factors[unit] if unit in factors else None
def run_periodic_hydraulics(model_input: str, devices: list[dict], rows: list[dict],
period_start: str, period_end: str) -> dict:
import wntr
start, end = datetime.fromisoformat(period_start), datetime.fromisoformat(period_end)
usable = [row for row in rows if row.get("quality") in {"valid", "corrected"}
and row.get("cleaned_value") is not None and isfinite(row["cleaned_value"])
and start <= datetime.fromisoformat(row["time"]) <= end]
sample_times = sorted({row["time"] for row in usable}, key=datetime.fromisoformat)
if len(sample_times) > 100:
raise ValueError("PERIOD_SAMPLE_LIMIT_EXCEEDED")
if not sample_times:
return {"conclusion": "insufficient_data", "comparisons": [], "snapshots": [],
"data_gaps": ["本周期没有有效观测,无法开展模拟与实测复核。"], "metrics": []}
comparisons, snapshots, gaps = [], [], set()
directory_devices = {device["device_id"]: device for device in devices}
with TemporaryDirectory(prefix="tjwater-periodic-hydraulics-") as directory:
source = Path(directory) / "frozen.inp"
source.write_text(model_input, encoding="utf-8")
for index, at in enumerate(sample_times):
model = wntr.network.WaterNetworkModel(str(source))
clock = datetime.fromisoformat(at).astimezone(ZoneInfo("Asia/Shanghai"))
model.options.time.duration = 0
model.options.time.pattern_start = clock.hour * 3600 + clock.minute * 60 + clock.second
model.options.time.start_clocktime = model.options.time.pattern_start
frame = {row["device_id"]: row for row in usable if row["time"] == at}
# Only exact-timestamp boundary observations can override the frozen
# model. Missing boundaries retain explicit model assumptions.
applied = []
for device_id, row in frame.items():
device = directory_devices.get(device_id, {})
node_id, link_id = device.get("node_id"), device.get("link_id")
value, kind = row["cleaned_value"], device.get("device_type")
if kind == "tank_liquid_level" and node_id in model.tank_name_list:
if device.get("measurement_unit") != "m":
gaps.add(f"{device_id} 水位单位未适配。")
continue
model.get_node(node_id).init_level = value
applied.append(device_id)
elif kind == "fixed_pump" and link_id in model.pump_name_list:
if value not in (0, 1):
gaps.add(f"{device_id} 工频泵状态不是 0/1。")
continue
model.get_link(link_id).initial_status = "OPEN" if value else "CLOSED"
applied.append(device_id)
elif kind == "demand" and node_id in model.junction_name_list:
demand = flow_cms(value, device.get("measurement_unit", ""))
if demand is None:
gaps.add(f"{device_id} 需水量单位未适配。")
continue
node = model.get_node(node_id)
node.demand_timeseries_list[0].base_value = demand
node.demand_timeseries_list[0].pattern_name = None
applied.append(device_id)
elif kind in {"variable_pump", "reservoir_liquid_level"}:
gaps.add(f"{device_id} 缺少已核实的频率基准或水位高程转换,沿用冻结模型边界。")
result = wntr.sim.EpanetSimulator(model).run_sim(
file_prefix=str(Path(directory) / f"snapshot-{index}"), convergence_error=True)
pressure = result.node["pressure"].iloc[-1]
flow = result.link["flowrate"].iloc[-1]
if not all(isfinite(float(value)) for value in pressure) or not all(isfinite(float(value)) for value in flow):
raise ValueError("NONFINITE_HYDRAULIC_RESULT")
snapshots.append({"time": at, "node_count": len(pressure), "link_count": len(flow),
"minimum_pressure_m": float(pressure.min()),
"maximum_pressure_m": float(pressure.max()),
"applied_boundary_device_ids": applied})
for device_id, row in frame.items():
device = directory_devices.get(device_id, {})
kind, node_id, link_id = device.get("device_type"), device.get("node_id"), device.get("link_id")
if kind == "pressure" and node_id in pressure:
observed = pressure_metres(row["cleaned_value"], device.get("measurement_unit", ""))
predicted, unit = float(pressure[node_id]), "m"
elif kind in {"flow", "pipe_flow"} and link_id in flow:
observed = flow_cms(row["cleaned_value"], device.get("measurement_unit", ""))
predicted, unit = float(flow[link_id]), "m3/s"
else:
if kind in {"pressure", "flow", "pipe_flow"}:
gaps.add(f"{device_id} 缺少有效模型关联。")
continue
if observed is None:
gaps.add(f"{device_id} 观测单位未适配。")
continue
delta = abs(predicted - observed)
comparisons.append({"time": at, "device_id": device_id, "kind": kind,
"element_id": node_id if kind == "pressure" else link_id,
"observed": observed, "predicted": predicted, "unit": unit,
"absolute_deviation": delta,
"relative_deviation_percent": delta / abs(observed) * 100
if abs(observed) > 1e-8 else None})
metrics = []
for kind, label, attention, critical in (("pressure", "模拟-实测压力偏差", 8, 12),
("flow", "模拟-实测流量偏差", 10, 15)):
values = [item["relative_deviation_percent"] for item in comparisons
if (item["kind"] == "pressure") == (kind == "pressure")
and item["relative_deviation_percent"] is not None]
if values:
value = mean(values)
metrics.append({"id": kind + "-deviation", "label": label, "value": f"{value:.2f}",
"unit": "%", "status": "critical" if value >= critical else "attention" if value >= attention else "normal",
"detail": f"有效比较 {len(values)} 项;近零实测只报告绝对偏差。"})
else:
gaps.add(f"缺少可计算相对偏差的{label}数据。")
pressure_deltas = [item["absolute_deviation"] for item in comparisons if item["kind"] == "pressure"]
if pressure_deltas:
maximum = max(pressure_deltas)
metrics.append({"id": "max-pressure-delta", "label": "最大节点压差", "value": f"{maximum:.2f}",
"unit": "m", "status": "critical" if maximum >= 8 else "attention" if maximum >= 5 else "normal",
"detail": "有实测映射节点的最大绝对压力偏差。"})
compared_ids = {item["device_id"] for item in comparisons}
valid_ids = {item["device_id"] for item in comparisons
if (end - datetime.fromisoformat(item["time"])).total_seconds()
< _sampling_seconds(directory_devices.get(item["device_id"], {}))}
expected = {device["device_id"] for device in devices if device.get("device_type") in {"pressure", "flow", "pipe_flow"}}
missing = sorted(expected - valid_ids)
if missing:
gaps.add("缺少周期终点可用观测或模型映射的测点:" + "、".join(missing))
gaps.add("未取得有效边界遥测的要素采用本轮冻结模型设定;本结果为逐时刻稳态水力复核。")
anomaly = any(metric["status"] != "normal" for metric in metrics)
complete = not missing and all(any(metric["id"] == kind + "-deviation" for metric in metrics) for kind in ("pressure", "flow"))
deviations = sorted(
[item for item in comparisons if (item["relative_deviation_percent"] is not None
and item["relative_deviation_percent"] >= (8 if item["kind"] == "pressure" else 10))
or (item["kind"] == "pressure" and item["absolute_deviation"] >= 5)],
key=lambda item: item["relative_deviation_percent"] or 0, reverse=True,
)
causes = (["先核查偏差测点的数据质量与模型关联,再复核冻结模型中的泵阀状态、供水边界和需水模式。",
"当前证据仅定位偏差测点和关联要素,尚不能单独确定边界滞后或模型参数错误。"] if deviations else [])
return {"conclusion": "attention" if anomaly else "normal" if complete else "insufficient_data",
"metrics": metrics, "comparisons": comparisons, "snapshots": snapshots,
"data_gaps": sorted(gaps), "compared_device_count": len(compared_ids), "current_device_count": len(valid_ids),
"missing_device_ids": missing, "model_scope": "all_nodes_and_links",
"deviation_targets": deviations[:50], "cause_review": causes}