Measure to calculate current value minus previous row for a specific device

[Anonymous]

Hey!

I have the same problem for some time now that I really can't solve. I am doing a calculation to calculate the total energy,

I have following as example:

I have a column with an accumulative TotalEnergy for every device, I have the devices as dirfferent name, Index and I made a Rank with following formula (Where EndUtc is Date and time)

Rank = RANKX(FILTER(Data, Data[Device] = EARLIER(Data[Device])), Data[EndUtc],,ASC)

I wanna make a calculation NEWENERGYPERDAY: as for device A, 365-364 = 1, 364-364=0 etc, for every row. 

I have tried many different calculations, as for example: 

NewEnergyPer300s_kWh =

VAR minIndx = CALCULATE (

MAX ( Data[Index] ),

ALLEXCEPT ( Data, Data[Device] ),

Data[Index] < EARLIER (Data[Index] )

)

RETURN

IF (

ISBLANK ( minIndx ),

BLANK (),

Data[TotalEnergy_kWh]

    - CALCULATE (

               SUM ( Data[TotalEnergy_kWh] ),

                   Data[Index] = minIndx,

                       ALL ( Data )

)

)

which worked fine, except that I am getting not enough memory capacity, which is weird cause I don't even have too much data, and when I find other formulas and using function EARLIER I get the same issue evrytime, is it possible to write this formula or similar as a measure and not a column?

And also if there are any better way to write the calculation that will be better for the memory capcity?

Thanks! Feeling a bit lost with this issue. 

Kind regards

[Anonymous · Accepted solution]

Hi again,

Can you share your file? I am sorry but I think that I have overlooked some important facts in original post, but if you have created a Index column that increases with every new entry per device you should be able to use this calculated column. And perhaps you do not have to use the device varible at all (i.e. if i Index contains unique values)

EnergyPerDay =
VAR device = Data[Device]    // could be skipped
VAR ind = Data[Index]
VAR preInd= ind-1
VAR energyPreDate  = 
  SUMX(
    FILTER(
      Data,
      Data[Index] = preInd 
      && Data[Device] = device   // could also be skipped
    ),
    Data[TotalEnergy_kWh]
)

RETURN
Data[TotalEnergy_kWh] - energyPreDate

Best regars,

Kristjan

[v-cherch-msft · Accepted solution]

Hi @Anonymous

You may try to create below measures:

NextValue =
CALCULATE (
    SUM ( Data[TotalEnergy_kWh] ),
    FILTER ( ALLEXCEPT ( Data, Data[Device] ), Data[Rank] = MAX ( Data[Rank] ) - 1 )
)

NEWENERGYPERDAY =
IF (
    ISBLANK ( [NextValue] ),
    BLANK (),
    SUM ( Data[TotalEnergy_kWh] ) - [NextValue]
)

Regards,

Cherie

[Anonymous]

Hi agnes_,

You can try this,

EnergyPerDay =
VAR device = Data[Device]
VAR d = Data[Date]
VAR tbl = 
  FILTER(
    Data,
    Data[Date] < d && Data[Device] = device
  )

VAR preDate = MAXX( tbl, Data[Date])
VAR energyPreDate = 
  SUMX(
    FILTER(
      tbl,
      Data[Date] = preDate
    ),
    Data[TotalEnergy_kWh]
)

RETURN
Data[TotalEnergy_kWh] - energyPreDate

  

If this works please LIKE and mark as solutions.

P.s. if there is no gap in the dates, you can change VAR preDate = MAXX( tbl, Data[Date]) to preDate = d-1 and then use preDate in the tbl varible right away.

EnergyPerDay =
VAR device = Data[Device]
VAR d = Data[Date]
VAR preDate = d-1
VAR tbl = 
  FILTER(
    Data,
    Data[Date] = preDate && Data[Device] = device
  )

VAR energyPreDate = 
  SUMX(
    tbl,
    Data[TotalEnergy_kWh]
)

RETURN
Data[TotalEnergy_kWh] - energyPreDate

Regards,

Kristjan

[Anonymous]

Thanks!

But everytime I try this as a measure I get the same problem, that PowerBI can't find  the column tables, and the names are correct: It find it at some places and some not. 

If I put max, it can find it though: 

[Anonymous]

And as a new column, same here not enough memory to complete the operation, which is weird cause I have other files that have more data then this one. 

[Anonymous]

Hello my question won't help but im curious : the same formula works on a bigger table than the one with memory error ? thank you

[Anonymous]

Yes, the other table uses this formula for the same type of calculation (someone else have made it though), but here are only one device!

Consumed Energy = 
VAR
    //Our Base Energy Consumption Value
    BaseValue = xlsx[ Stat_EnergyConsumptionSinceMantleChange]
    VAR
    //The date our BaseValue is based upon
    BaseDate=xlsx[DateTime]
    VAR
    //The date of the value we're planning on subtracting from our base value
    DateCheck=LOOKUPVALUE(xlsx[DateTime],xlsx[Index],xlsx[Index]-1)
VAR
    ValueToSubtract = IF(
        'xlsx'[Index]=0,xlsx[ Stat_EnergyConsumptionSinceMantleChange], //If index = 0 we're at the beginning of our list
        LOOKUPVALUE(
        xlsx[ Stat_EnergyConsumptionSinceMantleChange],
        xlsx[Index],
        xlsx[Index]-1)
        )  
                          
    RETURN
    IF(FORMAT(BaseDate,"YYYY-MM-DD")=FORMAT(DateCheck,"YYYY-MM-DD"), //If we're comparing values of the same date
    IF(ISBLANK(ValueToSubtract), //If value we're trying to subtract is blank
        BaseValue-BaseValue, // return basevalue - basevalue
    IF(ValueToSubtract>BaseValue, //If value we're trying to subtract is larger then our base value
        BaseValue-BaseValue, //return basevalue - basevalue
        BaseValue-ValueToSubtract) //Otherwise return the difference
        ),
        BaseValue-BaseValue) //If the dates are NOT the same - return 0
    
        

[Anonymous · Accepted solution]

Hi again,

Can you share your file? I am sorry but I think that I have overlooked some important facts in original post, but if you have created a Index column that increases with every new entry per device you should be able to use this calculated column. And perhaps you do not have to use the device varible at all (i.e. if i Index contains unique values)

EnergyPerDay =
VAR device = Data[Device]    // could be skipped
VAR ind = Data[Index]
VAR preInd= ind-1
VAR energyPreDate  = 
  SUMX(
    FILTER(
      Data,
      Data[Index] = preInd 
      && Data[Device] = device   // could also be skipped
    ),
    Data[TotalEnergy_kWh]
)

RETURN
Data[TotalEnergy_kWh] - energyPreDate

Best regars,

Kristjan

[Anonymous]

Hi again,

@Anonymous I am sorry but I do not understand what you mean.

@Anonymous, please let me know if the second calculated column worked.

Regards,

Kristjan

[Anonymous]

Yes, second formula worked without getting the memory capacity, but the results gets very high negative values, and the values should not be high at all, and positive, maybee I should use some LOOKUPVALUE, instead of the SUMX one?

[Anonymous]

My question wasn't for you 🙂
Agnes said that she got an memory error with the formula. But she had bigger tables. I wanres to be sure that she did run the actual formula on the big table as well.

If not clear, forget it, not a big deal ! 😉

[Anonymous]

Hi again,

If Data[EndUtc] only contains one row for each device and there are no missing dates you should try the second version of the calculated column. How many rows do you have in your table?

Regards,

kristjan

[Anonymous]

[EndUtc] have have data every 300s (every 5th minute of the day) but only one EndUtc for every device, so if there are five devices, there are five 2019-07-30 13:25:00 stamps for example.

I have 68,152 rows, 30 Columns, The memory capacity error occurs with Earlier function. 

Another PowerBI I have been working with the same type of calculation 

NewEnergyPerDay_kWh = 

var maxIndex =CALCULATE( MIN('Data'[MinIndex]), 
FILTER('Data','Data'[MinIndex]=EARLIER('Data'[MinIndex])))
VAR minVal='Data'[TotalEnergy_kWh]
var maxVal= IF('Data'[Index]=maxIndex,  
    LOOKUPVALUE('Data'[TotalEnergy_kWh],'Data'[Index],'Data'[Index]),
    LOOKUPVALUE('Data'[TotalEnergy_kWh],'Data'[Index],'Data'[Index]+1))
RETURN
CALCULATE(DIVIDE(minVal-maxVal,1,0),FILTER('Data','Data'[MinIndex]=EARLIER('Data'[MinIndex])), ALLEXCEPT(Data,Data[Device]))

 where: 

MaxIndex = CALCULATE(MAX('Data'[Index]),ALLEXCEPT('Data','Data'[Device]))

MinIndex = CALCULATE(MIN('Data'[Index]),ALLEXCEPT('Data','Data'[Device]))

 there I don't get any memory error, but still some error: