Time Series Data

Categories of Time Series

The bulk of the data in many power system models is time series data. Given the potential complexity, PowerSystems.jl has a set of definitions to organize this data and enable consistent modeling.

PowerSystems.jl supports two categories of time series data depending on the process to obtain the data and its interpretation:

• Static Time Series Data
• Forecasts

These categories are are all subtypes of TimeSeriesData and fall within this time series type hierarchy:

TimeSeriesData
 ├─ Forecast
 │   ├─ AbstractDeterministic
 │   │   ├─ Deterministic
 │   │   └─ DeterministicSingleTimeSeries
 │   ├─ Probabilistic
 │   └─ Scenarios
 └─ StaticTimeSeries
     └─ SingleTimeSeries

Static Time Series Data

A static time series data is a single column of data where each time period has a single value assigned to a component field, such as its maximum active power. This data commonly is obtained from historical information or the realization of a time-varying quantity.

Static time series usually comes in the following format, with a set resolution between the time-stamps:

This example is a 1-hour resolution static time-series.

In PowerSystems, a static time series is represented using SingleTimeSeries.

Forecasts

A forecast time series includes predicted values of a time-varying quantity that commonly includes a look-ahead window and can have multiple data values representing each time period. This data is used in simulation with receding horizons or data generated from forecasting algorithms.

Key forecast format parameters are the forecast resolution, the interval of time between forecast initial times, and the number of forecast windows (or forecasted values) in the forecast horizon.

Forecast data usually comes in the following format, where a column represents the time stamp associated with the initial time of the forecast, and the remaining columns represent the forecasted values at each step in the forecast horizon.

This example forecast has a interval of 1 hour and a horizon of 8.

PowerSystems defines the following Julia structs to represent forecasts:

• Deterministic: Point forecast without any uncertainty representation.
• Probabilistic: Stores a discretized cumulative distribution functions (CDFs) or probability distribution functions (PDFs) at each time step in the look-ahead window.
• Scenarios: Stores a set of probable trajectories for forecasted quantity with equal probability.

Data Storage

By default PowerSystems stores time series data in an HDF5 file. This prevents large datasets from overwhelming system memory. Refer to this page for details on how the time series data is stored in HDF5 files.

Time series data can be stored actual component values (for instance MW) or scaling factors intended to be multiplied by a scalar to generate the component values. By default PowerSystems treats the values in the time series data as physical units. In order to specify them as scaling factors, you must pass the accessor function that provides the multiplier value (e.g., get_time_series_array). The scaling factor multiplier must be passed into the forecast when you create it to use this option.

The time series contains fields for scaling_factor_multiplier and data to identify the details of th Component field that the time series describes, and the time series data. For example: we commonly want to use a time series to describe the maximum active power capability of a renewable generator. In this case, we can create a SingleTimeSeries with a TimeArray and an accessor function to the maximum active power field in the struct describing the generator. In this way, we can store a scaling factor time series that will get multiplied by the maximum active power rather than the magnitudes of the maximum active power time series.

Examples of how to create and add time series to system can be found in the Add Time Series Example