Aggregate values of 15 minute steps to values of h

问题:

I have a data frame that looks like this:

     Timedate              TotalSolar_MW
20  2013-06-01 04:45:00     13.0
21  2013-06-01 05:00:00     41.7
22  2013-06-01 05:15:00     81.8
23  2013-06-01 05:30:00    153.0
24  2013-06-01 05:45:00    270.7
25  2013-06-01 06:00:00    429.3
26  2013-06-01 06:15:00    535.4

"Timedate" is POSIXlt, and "Total_Solar" is numeric. The time steps are in 15 minute intervals from June 1, 0:00 to June 24, 24:00.

Now I want to aggregate the quarter hourly data to hourly steps e.g. 2013-06-01 06:00:00 934.8MW (81.8MW + 153.0MW + 270.7MW + 429.3MW; from 05:15 to 06:00)

I tried this with:

 Sum <-aggregate(Total_Solar_Gesamt$TotalSolar_MW, 
          list(as.POSIXlt(Total_Solar_Gesamt$Timedate)$hour), FUN=sum)

But it returns the aggregated hourly data of the whole data frame and gives me a new data.frame with 24 rows and the summed up MW for every hour.

How can I change the structure, only to reduce from a quarter hourly to a hourly interval? I tried a for loop but this also didn't work. Also subset didn't work for me.

Thanks for the help!

回答1:

When working with time series, I suggest you work with xts package for this, and for example hourly.apply:

 library(xts)
 dat.xts <- xts(Total_Solar_Gesamt$TotalSolar_MW,
                as.POSIXct(otal_Solar_Gesamt$Timedate))
 hourly.apply(dat.xts,sum)

More general you can use period.apply which is (lapply equivalent) , for example to aggregate your data each 2 hours you can do the following:

 ends <- endpoints(zoo.data,'hours',2) 
 period.apply(dat.xts,ends ,sum)

回答2:

Total_Solar_Gesamt <- read.table(header=TRUE, sep=",", text="
Timedate, TotalSolar_MW
2013-06-01 04:45:00, 13.0
2013-06-01 05:00:00, 41.7
2013-06-01 05:15:00, 81.8
2013-06-01 05:30:00, 153.0
2013-06-01 05:45:00, 270.7
2013-06-01 06:00:00, 429.3
2013-06-01 06:15:00, 535.4
")

Use cut.POSIXt to divide the dates into hourly intervals:

Sum <- aggregate(Total_Solar_Gesamt["TotalSolar_MW"], 
                 list(hour=cut(as.POSIXct(Total_Solar_Gesamt$Timedate), "hour")),
                 sum)
Sum
                 hour TotalSolar_MW
1 2013-06-01 04:00:00          13.0
2 2013-06-01 05:00:00         547.2
3 2013-06-01 06:00:00         964.7

Note that the above will group 06:00:00 with the other 06 times. If you want to group the top of the hour with the previous hour, just subtract one second from each timestamp:

Sum2 <- aggregate(Total_Solar_Gesamt["TotalSolar_MW"], 
                 list(hour=cut(as.POSIXct(Total_Solar_Gesamt$Timedate)-1, "hour")),
                 sum)
Sum2
                 hour TotalSolar_MW
1 2013-06-01 04:00:00          54.7
2 2013-06-01 05:00:00         934.8
3 2013-06-01 06:00:00         535.4

And, if you want to report your dates one hour forward, like in your question:

Sum2$adjustedHour <- as.POSIXct(Sum2$hour) + 3600
Sum2
                 hour TotalSolar_MW        adjustedHour
1 2013-06-01 04:00:00          54.7 2013-06-01 05:00:00
2 2013-06-01 05:00:00         934.8 2013-06-01 06:00:00
3 2013-06-01 06:00:00         535.4 2013-06-01 07:00:00

---

Using xts:

library(xts)
data.xts <- xts(Total_Solar_Gesamt$TotalSolar_MW, 
                as.POSIXct(Total_Solar_Gesamt$Timedate)-1)  
                # subtract 1 second, as discussed above
Sum.xts <- period.apply(data.xts, INDEX=endpoints(data.xts, "hours"), FUN=sum)
Sum.xts
                     [,1]
2013-06-01 04:59:59  54.7
2013-06-01 05:59:59 934.8
2013-06-01 06:14:59 535.4

Note how with xts, the timestamps in Sum.xts are the last timestamps of each hour. xts makes it easy to align them though:

Sum.xts <- align.time(Sum.xts, 3600)  # round up to next hour

Sum.xts
                     [,1]
2013-06-01 05:00:00  54.7
2013-06-01 06:00:00 934.8
2013-06-01 07:00:00 535.4