Form Based Regression Model

Both the Level and Ramp Rate Model have Form-Based models specifications. The Form-Based modeling method allows the user to implement a linear regression model by selecting from a list of predefined model specifications. These specifications work well under most conditions. If more complex model specifications are required, the user may use a MetrixND Custom model selection and develop a custom model using MetrixND.

The following fourteen (14) form-based regression models are available in the MetrixIDR. These model specifications may be applied to both the Level Model and Ramp Rate Models. The difference between the Level and Ramp Rate models is the different dependent variable.

  • CONSTANT.
  • CONSTANT_LAG.
  • DOW.
  • DOWDBT.
  • DOW_LAG.
  • DOWDBT_LAG.
  • DAYTYPE.
  • DAYTYPEDBT.
  • DAYTYPE_LAG.
  • DAYTYPEDBT_LAG.
  • EXDAYTYPE.
  • EXDAYTYPEDBT.
  • EXDAYTYPE_LAG.
  • EXDAYTYPEDBT_LAG.

The predefined regression model specifications are comprised of the following five groups of independent variables.

  • Day of Week. This independent variable group contains Day of the Week binary variables used to explain variation in the usage profile due to weekly usage patterns.
  • Day Type. This group contains Monthly binary and Monthly Weekend binary variables used to explain variation in usage due to seasonality.
  • Extended Day Type. This group contains five (5) daytype binary variables by month. The daytypes are Sunday, Monday, Friday, Saturday, and Tuesday-Wednesday-Thursday.
  • Weather. This group contains Cooling Degree-Day (CDD) and Heating-Degree-Day (HDD) variables that are computed from user defined cut points. A maximum of three cut points can be specified for each set of HDD and CDD variables, respectively (see the Degree-Day Cut Point section below for details regarding the cut point specification)
  • Lags. This group contains six (6) lagged values of the dependent variable. For example, the model for 07:00 would include the loads at 06:55, 06:50, 06:45, 06:40, 06:35, and 06:30. Models with lag structures are not recommended for modeling Ramp Rates.

A detailed description of each Form-Based model specification is presented below.

Constant (CONSTANT)

  • The simplest model specification, which includes a lone Constant Term.
  • This specification is used to model a flat load/ramp rate pattern, which is independent of weekly operational patterns, seasonality, weather etc.
  • For example, manufacturing plants that operate 24x7.

Autoregressive (CONSTANT_LAG)

  • The simplest autoregressive model specification, which includes a Constant Term and a six-period lag.
  • This specification is used to model regime shift load patterns that are independent of weekly operational patterns, seasonality, weather etc.
  • For example, compressor stations.

Day of Week (DOW)

  • Includes a Constant Term and Day of the Week variables.
  • This specification is used to model loads/ramp rates that have a distinct weekly operational pattern but are not weather sensitive.
  • For example, manufacturing plants that are either closed or ramp down on the weekends.

Day of Week Dry Bulb Temperatures (DOWDBT)

  • Includes a Constant Term, Day of the Week, and Weather variables.
  • This specification is used to model loads/ramp rates that have a distinct weekly operational pattern and are sensitive to weather.
  • For example, office buildings.

Day of Week Lags (DOW_LAG)

  • Includes a Constant Term, Day of the Week, and Lag variables.
  • This specification is used to model loads that have a distinct weekly operational pattern but are not weather sensitive. The lag structure will help sync the forecast to the most recent data.
  • For example, manufacturing plants that are either closed or ramp down on the weekends.

Day of Week Dry Bulb Temperatures Lags (DOWDBT_LAG)

  • Includes a Constant Term, Day of the Week, Weather, and Lag variables.
  • This specification is used to model loads that have a distinct weekly operational pattern and are sensitive to weather.
  • For example, office buildings.

Day Type (DAYTYPE)

  • Includes a Constant Term and Day Type variables.
  • This specification is used to model loads/ramp rates that have a clear seasonal pattern but are not sensitive to weather.
  • For example, irrigation customers.

Day Type Dry Bulb Temperature (DAYTYPEDBT)

  • Includes a Constant Term, Day Type, and Weather variables.
  • This specification is used to model loads/ramp rates that have a discrete seasonal pattern and sensitivity to weather.
  • For example, restaurants.

Day Type Lags (DAYTYPE_LAG)

  • Includes a Constant Term, Day Type, and Lag variables.
  • This specification is used to model loads that have a clear seasonal pattern, but are not sensitive to weather.
  • For example, irrigation customers.

Day Type Dry Bulb Temperature Lags (DAYTYPEDBT_LAG)

  • Includes a Constant Term, Day Type, Weather, and Lag variables.
  • This specification is used to model loads that have a discrete seasonal pattern and sensitivity to weather.
  • For example, restaurants.

Extended Day Type (EXDAYTYPE)

  • Includes a Constant Term and Extended Day Type variables.
  • This specification is used to model loads/ramp rates that have a clear seasonal and day-of-the-week pattern but are not sensitive to weather.
  • For example, one and two shift manufacturing customers.

Extended Day Type Dry Bulb Temperature (EXDAYTYPEDBT)

  • Includes a Constant Term, Extended Day Type variables, and Weather variables.
  • This specification is used to model loads/ramp rates that have a clear seasonal and day-of-the-week pattern and are sensitive to weather.
  • For example, hospitals.

Extended Day Type Lags (EXDAYTYPE_LAG)

  • Includes a Constant Term, Extended Day Type, and Lag variables.
  • This specification is used to model loads that have a clear seasonal and day-of-the-week pattern but are not sensitive to weather.
  • For example, one and two shift manufacturing customers.

Extended Day Type Dry Bulb Temperature Lags (EXDAYTYPEDBT_LAG)

  • Includes a Constant Term, Extended Day Type variables, Weather, and Lag variables.
  • This specification is used to model loads that have a clear seasonal and day-of-the-week pattern and are sensitive to weather.
  • For example, hospitals.

Extend Form Based Regressions to Include Solar Irradiance

When the box "Include Solar Irr." is checked for both level and ramp regression properties the following variables are included in the regression specification. All six variables are included in each interval equation (e.g. for a five minute model, all six variables are included in all 288 interval equations).

  • SolIrrMorn: sum of cloud adjusted irradiance between 6:00 am – 10:00 am

  • SolIrrMornWkEnd: SolIrrMorn interacted with weekend.

  • SolIrrMid: sum of cloud adjusted irradiance between 10:00 am – 02:00 pm

  • SolIrrMidWkEnd: SolIrrMid interacted with weekend.

  • SolIrrAfter: sum of cloud adjusted irradiance between 02:00 pm – 06:00 pm

  • SolIrrAfterWkEnd: SolIrrAfter interacted with weekend.

The option requires a SolarIrradiance weather concept configured in the system, defined by the SolarRadiation Stock Concept ID.

The following two figures represent the collection of independent variables included for each form-based regression model specification. The first figure maps composite groups (Constant, Lag, DOW, Daytype, EXDayType, and DBT) to each form-based regression model specifications. The second figure lists the variables included in each composite group.

Composite Groups Assigned to Form-Based Regression Models

Variables Include in Each Composite Group