Introduction to multivariable regression

Multivariable regression puts a parameter (KPI, variable) in the context of other parameters and maps it as a linear regression model. This means the value of the output parameter can be determined via the input parameters and the regression model. (model result)

The energy consumption of a plant is influenced by certain factors such as ambient temperature, quantity produced and quantity filled. When the regression model and the input variables are known, you can determine the theoretical energy consumption.

In addition to the model, parameters are exported that determine the quality of the model. If an input quantity is used that has no relation to the output dimension, the correlation coefficient is 0. This means that the model is not trustworthy.

Different models can be calculated and saved for each asset and parameter. The saved models are available to the user in an automatically generated MVR dashboard. Here, the result of the model is compared with the actual value. To be able to detect the deviation, the deviation and the cumulative deviation are also displayed in addition to the model result and the actual measured values.

After the KPIs, the model result and the deviation have been created for each model, a limit can be defined by means of a standard function. The limit value determines when the user will receive a notification, for example, from the Notifier.

You can create multiple models for each output parameter.