Requirements on submitted time series
- Christian Böhm
Analytics
Interpretation
Reservoir telemetry
Values are interpreted as being valid instantaneously, and only at the timestamp they refer to. If the HYDROGRID System does not get regular reservoir reading, it will lineraly interpolate between two readings, up to 24h gap. It is therefore assuming the change was constant over time getting between two readings.
Turbine and Gate telemetry
In a timeseries like below, every value is interpreted as being the average power over the timespan between timestamp ts and ts+1. (Similar for throughput/opening on gates.)
Example:
[
{
"turbineId": "Turbi_1",
"timeseries":
[
{
"timestamp": 1700053200000, # 15. November 2023 13:00:00 (UTC)
"value": 4.0
},
{
"timestamp": 1700056800000, # 14:00:00 (UTC)
"value": 4.2
},
# notice span of 2 hour here between the data points
{
"timestamp": 1700064000000, # 16:00:00 (UTC)
"value": 3.8
},
{
"timestamp": 1700067600000, # 17:00:00 (UTC)
"value": null
}
]
}
]
The value 4.0 with the timestamp 1pm is followed by another timestamp at 2pm. So this is interpreted in the way
“The average production over this timespan 1pm and 2pm (one hour) was 4.0MW”.
Analog the value 4.2 with the timestamp 2pm does not have a following timestamp until 4pm. So the HYDROGRID System interprets this data as:
“The average production over this timespan 2pm and 4pm (two hours) was 4.2 MW”.
Rules
Timestamps interval
for gates and turbine
should be at least in market granularity (or finer) e.g. one value every 15 minutes
for reservoirs
best in plant granularity (e.g. 15min), but please provide at least one sensor reading a day
Continuously provide data in one time-series in a 48 hour rolling window
Each time data is sent, provide telemetry of actual time and telemetry for the last 48 hours. (see below)
Timestamps
Cannot be older than 2 weeks (otherwise the entire dataset will be rejected)
Cannot be in the future (otherwise the entire dataset will be rejected)
Data
Should only be positive values (>= 0)
Format
The format is a time series. A time series is a list of pairs of timestamp and an actual value (no planned data).
Reservoir: Unix timestamp in ms, value in masl
→ The value represents the instantaneous measurement at the point in time
Turbine: Unix timestamp in ms + value in MW
→ The value represents the average power which was produced over the timespan between the timestamp and the next timestamp that is contained in the provided dataset
Gate: Unix timestamp in ms + value in m³/s (throughput, preferred) or cm (opening) or % (opening)
→ The value represents the average throughput (or opening) that was achieved over the timespan between the timestamp and the next timestamp that is contained in the provided dataset
Submit sensor data as time series in 48 hours rolling window with intervals in at least market granularity. i.E: Market granularity is hourly → data is sent hourly
Variants of submitted time series
HYDROGRID Insight processes each submitted time series as one consistent time series. It can process continous and split time-series.
Continuous time series
Split time series
Reservoir
Turbine / Gate
Data processing
On the turbine and gate endpoints (i.e. turbine power, gate discharge, gate opening) the terminating timestamp should provide the value NULL to indicate the end marker for the previous datapoint. If it is missing, the last datapoint will be discarded. To ensure data consistency, please provide an extra timestamp with a value of NULL. This indicates for how long of a duration the previous value is valid for. On reservoir level endpoints you send your telemetry as a current state, so there is no end marker necessary.