diff --git a/app/native/wndb/connection.py b/app/native/wndb/connection.py
index b42b481..7dd5de2 100644
--- a/app/native/wndb/connection.py
+++ b/app/native/wndb/connection.py
@@ -1,3 +1,66 @@
+from collections.abc import Iterator
+from contextlib import contextmanager
+from threading import RLock
+
 import psycopg as pg
 
-g_conn_dict : dict[str, pg.Connection] = {}
\ No newline at end of file
+from app.core.config import get_pgconn_string
+
+g_conn_dict: dict[str, pg.Connection] = {}
+_registry_lock = RLock()
+_project_locks: dict[str, RLock] = {}
+
+
+def _is_closed(connection: pg.Connection) -> bool:
+    return bool(getattr(connection, "closed", False))
+
+
+def _close_connection(connection: pg.Connection) -> None:
+    if not _is_closed(connection):
+        connection.close()
+
+
+def _get_project_lock(name: str) -> RLock:
+    with _registry_lock:
+        lock = _project_locks.get(name)
+        if lock is None:
+            lock = RLock()
+            _project_locks[name] = lock
+        return lock
+
+
+def open_connection(name: str) -> pg.Connection:
+    with _get_project_lock(name):
+        connection = g_conn_dict.get(name)
+        if connection is None or _is_closed(connection):
+            if connection is not None:
+                _close_connection(connection)
+            connection = pg.connect(
+                conninfo=get_pgconn_string(db_name=name), autocommit=True
+            )
+            g_conn_dict[name] = connection
+        return connection
+
+
+def is_connection_open(name: str) -> bool:
+    with _get_project_lock(name):
+        connection = g_conn_dict.get(name)
+        if connection is None:
+            return False
+        if _is_closed(connection):
+            del g_conn_dict[name]
+            return False
+        return True
+
+
+def close_connection(name: str) -> None:
+    with _get_project_lock(name):
+        connection = g_conn_dict.pop(name, None)
+        if connection is not None:
+            _close_connection(connection)
+
+
+@contextmanager
+def project_connection(name: str) -> Iterator[pg.Connection]:
+    with _get_project_lock(name):
+        yield open_connection(name)
diff --git a/app/native/wndb/database.py b/app/native/wndb/database.py
index 248b4a3..6d2893b 100644
--- a/app/native/wndb/database.py
+++ b/app/native/wndb/database.py
@@ -1,6 +1,6 @@
 from typing import Any
 from psycopg.rows import dict_row, Row
-from .connection import g_conn_dict as conn
+from .connection import project_connection
 
 API_ADD = 'add'
 API_UPDATE = 'update'
@@ -83,29 +83,33 @@ class DbChangeSet:
 
 
 def read(name: str, sql: str) -> Row:
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(sql)
-        row = cur.fetchone()
-        if row == None:
-            raise Exception(sql)
-        return row
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(sql)
+            row = cur.fetchone()
+            if row == None:
+                raise Exception(sql)
+            return row
 
 
 def read_all(name: str, sql: str) -> list[Row]:
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(sql)
-        return cur.fetchall()
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(sql)
+            return cur.fetchall()
 
 
 def try_read(name: str, sql: str) -> Row | None:
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(sql)
-        return cur.fetchone()
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(sql)
+            return cur.fetchone()
 
 
 def write(name: str, sql: str) -> None:
-    with conn[name].cursor() as cur:
-        cur.execute(sql)
+    with project_connection(name) as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql)
 
 
 def get_current_operation(name: str) -> int:
diff --git a/app/native/wndb/project.py b/app/native/wndb/project.py
index 25b9328..5403d0f 100644
--- a/app/native/wndb/project.py
+++ b/app/native/wndb/project.py
@@ -2,7 +2,11 @@ import os
 import psycopg as pg
 from psycopg import sql
 from psycopg.rows import dict_row
-from .connection import g_conn_dict as conn
+from .connection import (
+    close_connection,
+    is_connection_open,
+    open_connection,
+)
 from app.core.config import get_pgconn_string, get_pg_config, get_pg_password
 
 # no undo/redo
@@ -31,9 +35,7 @@ def have_project(name: str) -> bool:
 
 
 def copy_project(source: str, new: str) -> None:
-    if source in conn:
-        conn[source].close()
-        del conn[source]
+    close_connection(source)
 
     with pg.connect(
         conninfo=get_pgconn_string(db_name="postgres"), autocommit=True
@@ -176,17 +178,12 @@ def clean_project(excluded: list[str] = []) -> None:
 
 
 def open_project(name: str) -> None:
-    if name not in conn:
-        conn[name] = pg.connect(
-            conninfo=get_pgconn_string(db_name=name), autocommit=True
-        )
+    open_connection(name)
 
 
 def is_project_open(name: str) -> bool:
-    return name in conn
+    return is_connection_open(name)
 
 
 def close_project(name: str) -> None:
-    if name in conn:
-        conn[name].close()
-        del conn[name]
+    close_connection(name)
diff --git a/app/native/wndb/s0_base.py b/app/native/wndb/s0_base.py
index 65882ab..2986af2 100644
--- a/app/native/wndb/s0_base.py
+++ b/app/native/wndb/s0_base.py
@@ -1,5 +1,5 @@
 from psycopg.rows import dict_row, Row
-from .connection import g_conn_dict as conn
+from .connection import project_connection
 from .database import read
 from typing import Any
 
@@ -47,9 +47,10 @@ ELEMENT_TYPES : dict[str, int] = {
 }
 
 def _get_from(name: str, id: str, base_type: str) -> Row | None:
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select * from {base_type} where id = '{id}'")
-        return cur.fetchone()
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select * from {base_type} where id = '{id}'")
+            return cur.fetchone()
 
 
 def is_node(name: str, id: str) -> bool:
@@ -125,10 +126,11 @@ def is_region(name: str, id: str) -> bool:
 
 def _get_all(name: str, base_type: str) -> list[str]:
     ids : list[str] = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id from {base_type} order by id")
-        for record in cur:
-            ids.append(record['id'])
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id from {base_type} order by id")
+            for record in cur:
+                ids.append(record['id'])
     return ids
 
 
@@ -138,29 +140,32 @@ def get_nodes(name: str) -> list[str]:
 # DingZQ
 def _get_nodes_by_type(name: str, type: str) -> list[str]:
     ids : list[str] = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id from {_NODE} where type = '{type}' order by id")
-        for record in cur:
-            ids.append(record['id'])
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id from {_NODE} where type = '{type}' order by id")
+            for record in cur:
+                ids.append(record['id'])
     return ids 
 
 # DingZQ
 def get_nodes_id_and_type(name: str) -> dict[str, str]:
     nodes_id_and_type: dict[str, str] = {}
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id, type from {_NODE} order by id")
-        for record in cur:
-            nodes_id_and_type[record['id']] = record['type']
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id, type from {_NODE} order by id")
+            for record in cur:
+                nodes_id_and_type[record['id']] = record['type']
     return nodes_id_and_type
 
 # DingZQ 2024-12-31
 def get_major_nodes(name: str, diameter: int) -> list[str]:
     major_nodes_set = set()
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select node1, node2 from pipes where diameter > {diameter}")
-        for record in cur:
-            major_nodes_set.add(record['node1'])
-            major_nodes_set.add(record['node2'])
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select node1, node2 from pipes where diameter > {diameter}")
+            for record in cur:
+                major_nodes_set.add(record['node1'])
+                major_nodes_set.add(record['node2'])
 
     return list(major_nodes_set)
 
@@ -183,29 +188,32 @@ def get_links(name: str) -> list[str]:
 # DingZQ
 def _get_links_by_type(name: str, type: str) -> list[str]:
     ids : list[str] = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id from {_LINK} where type = '{type}' order by id")
-        for record in cur:
-            ids.append(record['id'])
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id from {_LINK} where type = '{type}' order by id")
+            for record in cur:
+                ids.append(record['id'])
     return ids
 
 # DingZQ
 def get_links_id_and_type(name: str) -> dict[str, str]:
     links_id_and_type: dict[str, str] = {}
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id, type from {_LINK} order by id")
-        for record in cur:
-            links_id_and_type[record['id']] = record['type']
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id, type from {_LINK} order by id")
+            for record in cur:
+                links_id_and_type[record['id']] = record['type']
     return links_id_and_type
 
 # DingZQ 2024-12-31
 # 获取直径大于800的管道
 def get_major_pipes(name: str, diameter: int) -> list[str]:
     major_pipe_ids: list[str] = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id from pipes where diameter > {diameter} order by id")
-        for record in cur:
-            major_pipe_ids.append(record['id'])
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id from pipes where diameter > {diameter} order by id")
+            for record in cur:
+                major_pipe_ids.append(record['id'])
     return major_pipe_ids
     
 # DingZQ
@@ -232,15 +240,16 @@ def get_regions(name: str) -> list[str]:
     return _get_all(name, _REGION)
 
 def get_node_links(name: str, id: str) -> list[str]:
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        links: list[str] = []
-        for p in cur.execute(f"select id from pipes where node1 = '{id}' or node2 = '{id}'").fetchall():
-            links.append(p['id'])
-        for p in cur.execute(f"select id from pumps where node1 = '{id}' or node2 = '{id}'").fetchall():
-            links.append(p['id'])
-        for p in cur.execute(f"select id from valves where node1 = '{id}' or node2 = '{id}'").fetchall():
-            links.append(p['id'])
-        return links
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            links: list[str] = []
+            for p in cur.execute(f"select id from pipes where node1 = '{id}' or node2 = '{id}'").fetchall():
+                links.append(p['id'])
+            for p in cur.execute(f"select id from pumps where node1 = '{id}' or node2 = '{id}'").fetchall():
+                links.append(p['id'])
+            for p in cur.execute(f"select id from valves where node1 = '{id}' or node2 = '{id}'").fetchall():
+                links.append(p['id'])
+            return links
 
 
 def get_link_nodes(name: str, id: str) -> list[str]:
@@ -259,4 +268,3 @@ def get_region_type(name: str, id: str)->str:
         return type
 
 
-
diff --git a/app/native/wndb/s24_coordinates.py b/app/native/wndb/s24_coordinates.py
index d038a96..df0fcc6 100644
--- a/app/native/wndb/s24_coordinates.py
+++ b/app/native/wndb/s24_coordinates.py
@@ -1,5 +1,7 @@
 from .database import *
+from .connection import project_connection
 from .s0_base import get_link_nodes
+from psycopg.rows import dict_row
 
 def sql_update_coord(node: str, x: float, y: float) -> str:
     coord = f"st_geomfromtext('point({x} {y})')"
@@ -49,10 +51,11 @@ def get_links_in_extent(name: str, x1: float, y1: float, x2: float, y2: float) -
     node_ids = set([s.split(':')[0] for s in get_nodes_in_extent(name, x1, y1, x2, y2)])
 
     all_link_ids = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select id from pipes")
-        for record in cur:
-            all_link_ids.append(record['id'])
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select id from pipes")
+            for record in cur:
+                all_link_ids.append(record['id'])
 
     links = []
     for link_id in all_link_ids:
diff --git a/app/native/wndb/s41_pipe_risk_probability.py b/app/native/wndb/s41_pipe_risk_probability.py
index 33f0fe9..b441305 100644
--- a/app/native/wndb/s41_pipe_risk_probability.py
+++ b/app/native/wndb/s41_pipe_risk_probability.py
@@ -1,5 +1,7 @@
 from .database import *
+from .connection import project_connection
 from .s0_base import *
+from psycopg.rows import dict_row
 import json
 
 def get_pipe_risk_probability_now(name: str, pipe_id: str) -> dict[str, Any]:
@@ -28,29 +30,31 @@ def get_pipe_risk_probability(name: str, pipe_id: str) -> dict[str, Any]:
 
 def get_network_pipe_risk_probability_now(name: str) -> list[dict[str, Any]]:
     pipe_risk_probability_list = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select * from pipe_risk_probability")
-        for record in cur:
-            #pipe_risk_probability_list.append(record)
-            t = {}
-            t['pipeid'] = record['pipeid']
-            t['pipeage'] = record['pipeage']
-            t['risk_probability_now'] = record['risk_probability_now']
-            pipe_risk_probability_list.append(t)
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select * from pipe_risk_probability")
+            for record in cur:
+                #pipe_risk_probability_list.append(record)
+                t = {}
+                t['pipeid'] = record['pipeid']
+                t['pipeage'] = record['pipeage']
+                t['risk_probability_now'] = record['risk_probability_now']
+                pipe_risk_probability_list.append(t)
 
     return pipe_risk_probability_list
 
 def get_pipes_risk_probability(name: str, pipe_ids: list[str]) -> list[dict[str, Any]]:
     pipe_risk_probability_list = []
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select * from pipe_risk_probability")
-        for record in cur:
-            if record['pipeid'] in pipe_ids:
-                t = {}
-                t['pipeid'] = record['pipeid']
-                t['x'] = record['x']
-                t['y'] = record['y']
-                pipe_risk_probability_list.append(t)
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select * from pipe_risk_probability")
+            for record in cur:
+                if record['pipeid'] in pipe_ids:
+                    t = {}
+                    t['pipeid'] = record['pipeid']
+                    t['x'] = record['x']
+                    t['y'] = record['y']
+                    pipe_risk_probability_list.append(t)
 
     return pipe_risk_probability_list
 
@@ -67,21 +71,22 @@ def get_pipe_risk_probability_geometries(name: str) -> dict[str, Any]:
     # key_endnode = '下游节点'
     key_geometry = 'geometry'
 
-    with conn[name].cursor(row_factory=dict_row) as cur:
-        cur.execute(f"select *, ST_AsGeoJSON(geometry) AS {key_geometry} from gis_pipe")
+    with project_connection(name) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(f"select *, ST_AsGeoJSON(geometry) AS {key_geometry} from gis_pipe")
 
-        for record in cur:
-            id = record[key_pipeId]
-            geom = json.loads(record[key_geometry])
+            for record in cur:
+                id = record[key_pipeId]
+                geom = json.loads(record[key_geometry])
 
-            pipe_risk_probability_geometries[id] = {
-                'points': geom['coordinates']
-            }
+                pipe_risk_probability_geometries[id] = {
+                    'points': geom['coordinates']
+                }
 
-            for col in record:
-                if col != key_geometry:
-                    pipe_risk_probability_geometries[id][col] = record[col]
+                for col in record:
+                    if col != key_geometry:
+                        pipe_risk_probability_geometries[id][col] = record[col]
 
     # print(len(pipe_risk_probability_geometries))
 
-    return pipe_risk_probability_geometries
\ No newline at end of file
+    return pipe_risk_probability_geometries
diff --git a/tests/unit/test_wndb_connection.py b/tests/unit/test_wndb_connection.py
new file mode 100644
index 0000000..ea68d90
--- /dev/null
+++ b/tests/unit/test_wndb_connection.py
@@ -0,0 +1,78 @@
+import pytest
+
+from app.native.wndb import connection
+from app.native.wndb import database
+from app.native.wndb import project
+
+
+class _FakeCursor:
+    def __init__(self, rows):
+        self.rows = rows
+        self.executed = []
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, exc_type, exc, tb):
+        return False
+
+    def execute(self, sql):
+        self.executed.append(sql)
+
+    def fetchall(self):
+        return self.rows
+
+
+class _FakeConnection:
+    def __init__(self, rows=None, *, closed=False):
+        self.rows = list(rows or [])
+        self.closed = closed
+        self.close_calls = 0
+
+    def cursor(self, row_factory=None):
+        if self.closed:
+            raise RuntimeError("the connection is closed")
+        return _FakeCursor(self.rows)
+
+    def close(self):
+        self.close_calls += 1
+        self.closed = True
+
+
+@pytest.fixture(autouse=True)
+def clear_native_connections():
+    connection.g_conn_dict.clear()
+    connection._project_locks.clear()
+    yield
+    connection.g_conn_dict.clear()
+    connection._project_locks.clear()
+
+
+def test_is_project_open_drops_closed_cached_connection():
+    connection.g_conn_dict["fengyang"] = _FakeConnection(closed=True)
+
+    assert project.is_project_open("fengyang") is False
+    assert "fengyang" not in connection.g_conn_dict
+
+
+def test_read_all_reopens_closed_cached_connection(monkeypatch):
+    stale = _FakeConnection(closed=True)
+    fresh = _FakeConnection(rows=[{"key": "DURATION", "value": "01:00:00"}])
+    connection.g_conn_dict["fengyang"] = stale
+
+    opened = []
+
+    def fake_connect(*, conninfo, autocommit):
+        opened.append((conninfo, autocommit))
+        return fresh
+
+    monkeypatch.setattr(connection.pg, "connect", fake_connect)
+    monkeypatch.setattr(
+        connection, "get_pgconn_string", lambda db_name: f"dbname={db_name}"
+    )
+
+    rows = database.read_all("fengyang", "select * from times")
+
+    assert rows == [{"key": "DURATION", "value": "01:00:00"}]
+    assert opened == [("dbname=fengyang", True)]
+    assert connection.g_conn_dict["fengyang"] is fresh