Model Template Requirements

The Custom MetrixND model assignment requires a properly constructed and formatted MetrixND project file. This section discusses the template requirements.

Data Table External Links

MetrixIDR is a multi-threaded system, and as such, does not support MetrixND project file templates that contain links to single-user data sources (e.g. Microsoft Excel). The MetrixND project file templates may contain Data Tables with deactivated links. If the user chooses to use data contained in a single-user data source, it is recommended the user link to the single-user data source, populate the data table, and then deactivate the link prior to importing the template into the automated system.

Data Tables for Models Assigned a Weather Zone or Weather Station

  1. Calendar. The Calendar Data Table stores a binary representation of holiday variables. The Calendar Data Table is populated through the assignment of a Holiday set, which is managed within the MetrixIDR system. To use Day of Week (DOW) variables in the template models, the user should configure a transformation table that contains the DOW variables. To apply the holiday set:

    • The Data Table must be named "Calendar."

    • The data variable names contained in the Calendar Data Table must agree with the names of the holidays contained within the holiday set. The alignment of the names is case sensitive.

  2. DLSav. The DLSav Data Table stores a binary representation of daylight savings time. The DLSav Data Table is populated through the assignment of a Time Zone and by checking the Use Daylight Saving option on the Model List. To apply the variable:

    • The Data Table must be named "DLSav."

    • The Data Variable name must also be "DLSav."

    • The alignment of the table and variable names is case sensitive.

  3. Load. When MetrixND projects are accessed by the IDR system, the Load Data table is updated with the load data from the Data Source assigned to the model. Start and End dates for the record set are specified by the MetrixIDR application. The Load Data table:

    • must be present in the .ndm file.

    • must be named "Load".

    • should not have a check in the Active Link check box for the variables.

    • For hourly data, 24 variables must be present on the data table. It is recommended that they must be named "HR01" – "HR24" or “HR00” – “HR23”. The variables should be on the table in the correct order as the data is updated by position rather than by name.

    • For half-hourly data, it is recommended that variables be named "Int0030" – "Int2400."

    • For fifteen-minute data, it is recommended that variables be named "Int0015" – "Int2400."

    • For five-minute data, it is recommended that variables be named "Int0005" – "Int2400."

  4. Weather. The weather data tables contain hourly weather data. When MetrixND projects are accessed by the IDR system, these tables are updated with the weather data from the assigned weather station or weather zone. Data tables can exist for each weather concept defined in the system (i.e. DryBulb data table, DewPoint data table, etc.). Start and End dates for the record set are specified by the MetrixIDR application. The weather data tables:

    • must be named as defined in your implementation specific documentation. These names are also defined for each weather concept in the MetrixIDR database.

    • should not have a check in the Active Link check box for the variables.

    • it is required that there are 24 variables present on the data table. It is recommended that the variables be named "HR00" – "HR23". The variables should be on the table in the correct order as the data is updated by position rather than by name.

  5. Bad. If present, the Bad table contains a field called Bad. A 1 in the Bad column indicates the data has been marked as bad for the day and should not be used for model-estimation. The Bad table:

    • must be named "Bad"

    • must contain only the Bad variable

    • should not have a check in the Active Link check box for the variables

  6. BadWeather. If present, the BadWeather tables contain a field called Bad. A 1 in the Bad column indicates the data has been marked as bad and should not be used for model-estimation. BadWeather data tables can exist for each weather concept defined in the system (i.e. BadDryBulb data table, BadDewPoint data table, etc.). The BadWeather data tables are updated with a record set compiled from the WthrReading table in the project database. The BadWeather tables:

    • must be named as defined in your implementation specific documentation. These names are also defined for each weather concept in the MetrixIDR database.

    • must contain only the Bad variable.

    • should not have a check in the Active Link check box for the variables.

Data Tables for Models Assigned a Weather Group

When a model is assigned a Weather Group, there is a table for each region/concept combination defined by the Weather Group.

  1. RegionIDWeather. The region weather data tables contain hourly weather data. When MetrixND projects are accessed by the IDR system, these tables are updated with the weather data from the assigned weather group. Start and End dates for the record set are specified by the MetrixIDR application. Region data tables can exist for each weather concept defined in the system (i.e. RegionIDDryBulb data table, RegionIDDewPoint data table, etc.). Data tables must exist for each Region in the Weather Group. The region weather data tables:

    • Must be named as defined in your implementation specific documentation. These names are also defined for each weather concept in the MetrixIDR database.

    • Should not have a check in the Active Link check box for the variables.

    • It is required that there are 24 variables present on the data table. It is recommended that the variables be named "HR00" – "HR23". The variables should be on the table in the correct order as the data is updated by position rather than by name.

Models

For the MetrixND project to interact effectively with the MetrixIDR application, all models must be locked. Additionally, we recommend checking Autobad for regression and neural network models.

MStat Compare Object

The MStat Compare object calculates the statistics displayed on the Statistics tabs. The MStat Compare object:

  1. must be named "Stats"

  2. must have the following statistics checked:

    • Adjusted R-squared

    • Regression MAD

    • Regression MAPE

    • Regression Std. Error

GroupForecast Object

The GroupForecast object generates the forecast information that is ultimately passed into MetrixIDR. Forecasts generated by the GroupForecast object are adjusted with a transformation table called AdjPred. The AdjPred transform compensates for any possible negative forecast values by substituting 0 for any negative values. The FcstPred report organizes the transform data for availability to MetrixIDR. The GroupForecast Object:

  1. must be named "Forecast"
  2. must be set to use "Predicted"

For more information on the GroupForecast process, please refer to the MetrixND

Reports

Reports format data generated by prior MetrixND objects for compatibility and integration into the MetrixIDR system. Report names in the model must correspond exactly with the report names shown below. The following reports are required.

  1. 1FcstPred. The FcstPred report contains the GroupForecast results as adjusted with the AdjPred transform table.
  2. EstPred. The EstPred report contains the individual model predicted values for each of the intervals modeled.
  3. EstResid. The EstResid report contains the individual model residual values for each of the intervals modeled.