我将把比较的细节留给您，但是您可以使用CROSS APPLY来收集用于比较的数据。
比如：

SELECT *
FROM TemperatureData T
CROSS APPLY (
    SELECT AVG(T2.Temp * 1.0) AS PriorAvgTemp, COUNT(*) As PriorCount
    FROM TemperatureData T2
    WHERE T2.id = T.id
    AND T2.datetime >= DATEADD(second, -60, T.datetime)
    AND T2.datetime < T.datetime
) P
WHERE T.Temp BETWEEN P.PriorAvgTemp - 10 AND P.PriorAvgTemp + 10
--OR P.PriorCount < 3 -- Should we allow if there is insufficient prior data
--AND P.PriorCount >= 3 -- Should we omit if there is insufficient prior data

请确保在TemperatureData(id, datetime)上有索引。
如果您愿意接受最后N个值而不是时间范围，则窗口聚合计算可能更有效。

SELECT *
FROM (
    SELECT *,
        AVG(T.Temp * 1.0)
            OVER(PARTITION BY id ORDER BY datetime
                 ROWS BETWEEN 60 PRECEDING AND 1 PRECEDING)
            AS PriorAvgTemp,
        COUNT(*) 
            OVER(PARTITION BY id ORDER BY datetime
                 ROWS BETWEEN 60 PRECEDING AND 1 PRECEDING)
            AS PriorCount
    FROM TemperatureData T
) TT
WHERE TT.Temp BETWEEN TT.PriorAvgTemp - 10 AND TT.PriorAvgTemp + 10
--OR TT.PriorCount < 3 -- Should we allow if there is insufficient prior data
--AND TT.PriorCount >= 3 -- Should we omit if there is insufficient prior data

请注意：以上是未经测试的代码，可能需要一些语法修正和调试。如果你发现错误，请评论，我会更正后。