diff --git a/database_test_python/Loops_Power_Valid_2.py b/database_test_python/Loops_Power_Valid_2.py
new file mode 100644
index 0000000..b4a5f86
--- /dev/null
+++ b/database_test_python/Loops_Power_Valid_2.py
@@ -0,0 +1,140 @@
+import os
+
+import pandas as pd
+from sqlalchemy import create_engine, text
+
+from db_config import get_db_url
+
+
+LOOPS_COLUMNS = [
+    "id",
+    "loop_name",
+    "room_type_id",
+    "loop_address",
+    "loop_type",
+    "type",
+    "name",
+    "power",
+    "rate",
+    "temperature",
+    "air_type",
+    "air_brand",
+    "air_model",
+    "height",
+    "area",
+    "heat_loss",
+    "remark",
+]
+
+
+def create_target_table(target_conn, target_schema: str, target_table: str):
+    """
+    在目标库中创建目标表。
+    与 Valid_1 不同，此表用于存储跨房间的一对多摊平(Flattened)后的回路与对应明细 (单个回路对应多个房会生成多行)。
+    """
+    target_conn.execute(text(f'CREATE SCHEMA IF NOT EXISTS "{target_schema}"'))
+    target_conn.execute(text(f'DROP TABLE IF EXISTS "{target_schema}"."{target_table}"'))
+
+    # room_type_id 已经在 LOOPS_COLUMNS 中包含，所以不重复添加，只需要单独加 hotel_id 和 room_id
+    create_sql = f'''
+        CREATE TABLE "{target_schema}"."{target_table}" (
+          "id" int4 NOT NULL,
+          "loop_name" varchar(255),
+          "room_type_id" int4 NOT NULL,
+          "loop_address" varchar(255),
+          "loop_type" varchar(50),
+          "type" varchar(254),
+          "name" varchar(254),
+          "power" float8,
+          "rate" float8,
+          "temperature" float8,
+          "air_type" varchar(254),
+          "air_brand" varchar(254),
+          "air_model" varchar(254),
+          "height" float8,
+          "area" float8,
+          "heat_loss" float8,
+          "remark" varchar(254),
+          "hotel_id" int4,
+          "room_id" varchar(255)
+        )
+    '''
+    target_conn.execute(text(create_sql))
+
+
+def main():
+    """
+    主函数：执行数据的抽取、转换与加载 (ETL)。
+    此脚本主要使用 JOIN 将 loops 表与 rooms 关联，
+    如果某个 circuit (loop) 对应的是多个 room_id，则会直接平铺 (Flattened) 输出为多行记录分别保存。
+    """
+    source_schema = os.getenv("SOURCE_SCHEMA", "temporary_project")
+    target_schema = os.getenv("TARGET_SCHEMA", "wh_test")
+    target_table = os.getenv("TARGET_TABLE", "loops_type01_flattened")
+
+    # 此处利用 db_config 中统一配置创建源和目标的 SQLAlchemy 连接引擎
+    # 默认源库：log_platform；默认目标库：test
+    source_engine = create_engine(get_db_url("SOURCE"))
+    target_engine = create_engine(get_db_url("TARGET"))
+
+    with source_engine.connect() as source_conn:
+        stats_sql = f'''
+            SELECT
+                count(*) AS total_rows,
+                count(*) FILTER (WHERE "type" IN ('0', '1')) AS exact_match_rows,
+                count(*) FILTER (WHERE trim(coalesce("type", '')) IN ('0', '1')) AS normalized_match_rows
+            FROM "{source_schema}"."loops"
+        '''
+        stats = source_conn.execute(text(stats_sql)).mappings().first()
+
+        dist_sql = f'''
+            SELECT
+                "type",
+                count(*) AS cnt
+            FROM "{source_schema}"."loops"
+            GROUP BY "type"
+            ORDER BY cnt DESC, "type" NULLS FIRST
+            LIMIT 20
+        '''
+        type_dist = source_conn.execute(text(dist_sql)).mappings().all()
+
+        loops_select = ",\n                ".join([f'l."{c}"' for c in LOOPS_COLUMNS])
+
+        # 核心 SQL:
+        # 使用标准的 LEFT JOIN 自然展开。
+        # 取 type 为 '0' 或 '1' 的回路，一行 loop 如果通过房型关联对应多个 room_id，则会自动分裂成多行 (平铺存储)。
+        data_sql = f'''
+            SELECT
+                {loops_select},
+                h."hotel_id" AS hotel_id,
+                r."room_id" AS room_id
+            FROM "{source_schema}"."loops" l
+            LEFT JOIN "{source_schema}"."room_type" rt
+                ON rt."id" = l."room_type_id"
+            LEFT JOIN "{source_schema}"."hotels" h
+                ON h."id" = rt."hotel_id"
+            LEFT JOIN "{source_schema}"."rooms" r
+                ON r."hotel_id" = rt."hotel_id" AND r."room_type_id" = l."room_type_id"
+            WHERE trim(coalesce(l."type", '')) IN ('0', '1')
+        '''
+        df = pd.read_sql(text(data_sql), source_conn)
+
+    with target_engine.begin() as target_conn:
+        create_target_table(target_conn, target_schema, target_table)
+        df.to_sql(target_table, target_conn, schema=target_schema, if_exists="append", index=False)
+
+    print(f"Inserted rows: {len(df)}")
+    print(f"Target table: {target_schema}.{target_table}")
+    print(
+        "Source counts: "
+        f"total={stats['total_rows']}, "
+        f"exact_type_0_1={stats['exact_match_rows']}, "
+        f"normalized_type_0_1={stats['normalized_match_rows']}"
+    )
+    print("Top 20 type distribution:")
+    for row in type_dist:
+        print(f"type={row['type']!r}, count={row['cnt']}")
+
+
+if __name__ == "__main__":
+    main()