...
- A union merge function for the PredictionTimes is applied. This leads to a result that ignores if some elements that participate in an aggregation are not valid in the prediction time dimension at some points in time, which is probably what the user wants. An intersection could be useful in some situations, but is not the default option in the implementation.
- Most aggregation functions are incremental functions. Currently, only those are supported to work with temporal attributes. They are translated to a
TemporalIncrementalAggregationFunction, which creates a temporal output for the function.
Examples
Text
Energy Consumption
In the following, a few examples that use the temporal feature are presented.
Energy Consumption
Imagine you have a few smart meters that send you the current total energy value in an interval of 15 minutes. You need to know the consumption in-between these measurements or you want to know which households are consuming a high amount of energy right now. The following query does that for you. It takes the energy consumption and converts it into temporal attributes. Then the prediction time is set to the next 15 minutes. The derivative function can be used to get the wh per minute. Finally, the select operator selects the points in time where the consumption is "high".Text
| Code Block |
|---|
#PARSER PQL
#DOREWRITE false
#DEFINE PATH_LOCAL '/media/mydata/dev/odysseus_workspace/phd-workspace/Moving Object/Evaluation/scenarios/energy/energy_data.csv'
#ADDQUERY
input = ACCESS({
source='households',
wrapper='GenericPull',
schema=[
['id','Integer'],
['wh','Integer'],
['BaseDateTime','StartTimeStamp']
],
inputschema=['Integer','Integer','Integer'],
transport='File',
protocol='csv',
datahandler='Tuple',
metaattribute = ['TimeInterval', 'PredictionTimes'],
options=[
['filename', ${PATH_LOCAL}],
['Delimiter',','],
['TextDelimiter','"'],
['delay','1000'],
['readfirstline','false'],
['BaseTimeUnit','MINUTES']
]
}
)
/// Only use the last two measured values
time = TIMEWINDOW({
size = [20, 'minutes']
},
input
)
/// Convert the wh-attribute to a temporal double
temporalize = AGGREGATION({
aggregations = [
['function' = 'ToTemporalDouble', 'input_attributes' = 'wh', 'output_attributes' = 'temp_wh']
],
group_by = ['id'],
eval_at_outdating = false
},
time
)
predTime = PREDICTIONTIME({
addtostartvalue = [0, 'MINUTES'],
addtoendvalue = [15, 'MINUTEs']
},
temporalize
)
/// Energy consumption per household per minute in the next 15 minutes
derivative = MAP({
expressions = [
['derivative(temp_wh, PredictionTimes)','whPerMinute'],
['id','id']
]
},
predTime
)
/// Energy consumption per household per minute in the next 15 minutes
watts = MAP({
expressions = [
['whPerMinute * 60','watt'],
['id','id']
]
},
derivative
)
/// Filter out elements with a low energy consumption
highConsumption = SELECT({
predicate = 'watt > 300'
},
watts
) |
TextAnd here's an example file with energy consumption data:
| Code Block | ||
|---|---|---|
| ||
id,wh,time 1,0,0 2,0,3 3,0,6 1,115,15 2,50,18 3,250,21 1,200,30 2,60,33 3,500,36 1,210,45 2,100,48 3,600,51 |