diff --git a/app/database.py b/app/database.py
index bffaac4..9ab9777 100644
--- a/app/database.py
+++ b/app/database.py
@@ -445,7 +445,10 @@ class DatabaseManager:
             mysql_type = self.TYPE_MAPPING.get(col_type, 'VARCHAR(255)')
             column_defs.append(f'`{col}` {mysql_type}')
         
-        sql = f"CREATE TABLE IF NOT EXISTS `{table_name}` ({', '.join(column_defs)}) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4"
+        sql = (
+            f"CREATE TABLE IF NOT EXISTS `{table_name}` ({', '.join(column_defs)}) "
+            "ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci"
+        )
         
         with self.get_connection() as conn:
             with conn.cursor() as cursor:
diff --git a/app/services/cell_data.py b/app/services/cell_data.py
index 24eea67..484c964 100644
--- a/app/services/cell_data.py
+++ b/app/services/cell_data.py
@@ -396,7 +396,7 @@ class CellDataProcessor:
               `功率` varchar(100) DEFAULT NULL,
               `网络` varchar(20) DEFAULT NULL,
               KEY `CGI` (`CGI`)
-            ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4
+            ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci
             """
         )
 
@@ -548,9 +548,17 @@ def copy_celldata_tables_to_capacity(app_config: AppConfig, logger=None) -> int:
                 logger.info("CellData 数据库中没有表，跳过复制")
             return 0
 
+        target_charset, target_collation = _database_charset_and_collation(cap_conn)
         copied = 0
         for table in tables:
-            rows = _copy_one_table(cd_conn, cap_conn, table, logger)
+            rows = _copy_one_table(
+                cd_conn,
+                cap_conn,
+                table,
+                target_charset,
+                target_collation,
+                logger,
+            )
             if rows >= 0:
                 copied += 1
         cap_conn.commit()
@@ -571,7 +579,28 @@ def _list_tables(conn) -> list[str]:
         return [list(row.values())[0] for row in cur.fetchall()]
 
 
-def _copy_one_table(src_conn, dst_conn, table: str, logger=None) -> int:
+def _database_charset_and_collation(conn) -> tuple[str, str]:
+    with conn.cursor() as cur:
+        cur.execute(
+            "SELECT @@character_set_database AS `charset_name`, "
+            "@@collation_database AS `collation_name`"
+        )
+        row = cur.fetchone()
+    charset = str(row["charset_name"])
+    collation = str(row["collation_name"])
+    if not re.fullmatch(r"[A-Za-z0-9_]+", charset) or not re.fullmatch(r"[A-Za-z0-9_]+", collation):
+        raise RuntimeError("目标数据库字符集或排序规则名称无效")
+    return charset, collation
+
+
+def _copy_one_table(
+    src_conn,
+    dst_conn,
+    table: str,
+    target_charset: str,
+    target_collation: str,
+    logger=None,
+) -> int:
     with src_conn.cursor() as cur:
         cur.execute(f"SHOW CREATE TABLE `{table}`")
         row = cur.fetchone()
@@ -584,6 +613,10 @@ def _copy_one_table(src_conn, dst_conn, table: str, logger=None) -> int:
     with dst_conn.cursor() as cur:
         cur.execute(f"DROP TABLE IF EXISTS `{table}`")
         cur.execute(create_sql)
+        cur.execute(
+            f"ALTER TABLE `{table}` CONVERT TO CHARACTER SET {target_charset} "
+            f"COLLATE {target_collation}"
+        )
 
     if count == 0:
         if logger:
diff --git a/db_init/celldata.cellinfo.sql b/db_init/celldata.cellinfo.sql
index 683ecdf..0f03ba8 100644
--- a/db_init/celldata.cellinfo.sql
+++ b/db_init/celldata.cellinfo.sql
@@ -14,4 +14,4 @@ CREATE TABLE IF NOT EXISTS `cellinfo` (
   `功率` varchar(100) DEFAULT NULL,
   `网络` varchar(20) DEFAULT NULL,
   KEY `CGI` (`CGI`)
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
diff --git a/db_init/celldata.sector.sql b/db_init/celldata.sector.sql
index e133f87..a32baec 100644
--- a/db_init/celldata.sector.sql
+++ b/db_init/celldata.sector.sql
@@ -11,4 +11,4 @@ CREATE TABLE IF NOT EXISTS `sector` (
   `站型` varchar(50) DEFAULT NULL,
   `网络` varchar(20) DEFAULT NULL,
   KEY `CGI` (`CGI`)
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
diff --git a/db_init/celldata.sector_band_ref.sql b/db_init/celldata.sector_band_ref.sql
index cfe3791..3573976 100644
--- a/db_init/celldata.sector_band_ref.sql
+++ b/db_init/celldata.sector_band_ref.sql
@@ -9,7 +9,7 @@ CREATE TABLE IF NOT EXISTS `sector_band_ref` (
   `PLMN` varchar(16) DEFAULT NULL,
   `制式` varchar(20) DEFAULT NULL,
   `频段` varchar(20) DEFAULT NULL
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
 
 INSERT INTO `sector_band_ref` (`网络`,`频点下限`,`频点上限`,`PLMN`,`制式`,`频段`) VALUES
 ('5G',4000,99999,NULL,'4.9G','4.9G'),
diff --git a/docs/project_context.md b/docs/project_context.md
index 3852cad..9eff29a 100644
--- a/docs/project_context.md
+++ b/docs/project_context.md
@@ -1112,3 +1112,10 @@ CellData 处理日志细化（`app/services/cell_data.py`）：
 - 全脚本审计覆盖 32 个表标识：配置输入、脚本内建表和线上实际表之间无剩余大小写错配/冲突，且按 157 条语句执行顺序检查无“表尚未创建便引用”和未加反引号的表名。
 - 修复镜像 `capacityrepost-web:c054b48` 已部署并重启 Web；MySQL 和旧实例未重启。线上以同一数据库执行第 126、127 条核心查询通过，移动/广电临时校验结果分别为 23,227/23,107 行，临时表已删除。
 - 自动调度保持启用，正式任务 `20260806_170846` 已启动；因完整处理耗时较长，后续结果由用户自行观察。
+
+## 2026-08-06：统一 CellData 与容量库排序规则
+
+- 第 136 条 SQL 的 1267 错误来自跨表 CGI 关联：容量结果表使用 `utf8mb4_unicode_ci`，CellData 复制来的 `sector/cellinfo` 保留源表 `utf8mb4_0900_ai_ci`；第 136-139 条四个 CGI/NCGI 关联均有同类风险。
+- `copy_celldata_tables_to_capacity` 在目标空表创建后、插入前读取目标数据库的 `@@character_set_database/@@collation_database` 并执行 `ALTER TABLE ... CONVERT`，复制表不再继承不兼容的源排序规则；字符集和排序规则名称限定为字母数字下划线。
+- `CellDataProcessor`、`DatabaseManager` 和三份 `db_init` 建表 SQL 均显式使用 `utf8mb4_unicode_ci`，避免 MySQL 8 在只写 `DEFAULT CHARSET=utf8mb4` 时回落到 `utf8mb4_0900_ai_ci`。
+- 线上 `celldata` 与 `CapacityReport` 两库的 `sector/cellinfo` 已迁移为 `utf8mb4_unicode_ci`；4G/5G 结果表分别关联 sector/cellinfo 的四组真实查询均通过。自动重试任务 `20260806_180436` 保持运行，未重启容器。