Regression and correlation#

The Regression Analysis module allows:

using correlation analyses to find dependencies in data and thus promote process understanding and transparency,
determine the nature and strength of the dependencies as a basis for optimisation and investment decisions,
Build regression models based on the knowledge of dependencies,
Perform regression analyses based on the models created for arbitrary time periods,
cover the requirements for the normalisation of ratios according to ISO 50 006,
visualise the impact of optimisation measures in energy and process.

Optimisation cycle:

Identify and evaluate dependencies in time series data,
Make number-based decisions for optimization,
Optimisation of energy management and processes,
Creating Regression Models Over Reference Periods,
Documentation and monitoring of the efficiency of optimization measures based on regression analyses.
…

Correlation analysis#

Correlation analysis is a statistical procedure that provides information about the type and strength of the relationship between variables.

Entry point#

Click path: menu > regression analysis > correlation analyses. All existing correlation analyses are displayed here.

Interaction possibilities#

Es öffnet die Korrelationsübersichtsseite. Diese Seite ist die zentrale Verwaltungsstelle für Ihre Korrelationsanalysen. Hier werden alle bestehenden Korrelationsanalysen angezeigt.

Am Ende jeder Zeile befindet sich ein Dreipunktmenü, das die Bearbeitung von Metadaten und das Löschen der Korrelationsanalyse ermöglicht.

Clicking on a line takes you to the detail page of the selected correlation analysis.

Create correlation analyses#

Entry point#

A new correlation analysis can be created via the correlation overview page.

Click path: Menu > Regression analysis > Correlation analysis

To create a new correlation analysis, click the + Add correlation button on the overview page. A dialogue opens.

Function#

Enter a name for the new correlation analysis. After confirming the dialogue, you are automatically redirected to the detail page of the newly created correlation analysis.

Copy correlation analyses#

You can create a new correlation analysis by copying an existing analysis.You will find entry points for this on the overview and detail pages in the respective menus.

When copying, you can assign the name for the new correlation analysis. You will then be redirected directly to the detail page of the newly created correlation.

Perform correlation analyses#

Add measuring points#

On the detail page of a correlation analysis, measuring points can be added from the tree by drag & drop. A new chip then appears with the language-dependent display name of the measuring point.

Remove measuring points#

The x-icon at the end of a chip can be used to remove measuring points.

Calculate correlation#

Fügen Sie zunächst alle Messpunkte hinzu, deren Korrelationen Sie untereinander ermitteln möchten.

Then click on the Start Calculation button.

Es öffnet ein Dialog, über den die Berechnungseinstellungen vorgenommen werden können.

Aggregationsintervall wird für alle beteiligten Messpunkte gesetzt. Wählen Sie das Aggregationsintervall Kein aus, um Rohdaten zu nutzen.

Then select the aggregation type for each measuring point. If you calculate with raw data due to the aggregation interval setting, the configuration of the aggregation types has no influence on the calculation.

Klicken Sie auf den Button Berechnung starten, um die Korrelationsmatrix zu ermitteln.

Evaluate correlation analyses#

The correlation matrix shows you whether the correlation between two measuring points in the evaluation period is positive or negative and how strong the correlation is.For this purpose, the correlation matrix uses a colour indication for positive / negative as well as a colour weighting regarding the strength of the correlation. In addition, the correlation coefficients are displayed in the corresponding tiles.

If you want to know more details about a particular correlation, click on the corresponding tile. A dialogue opens in which a dot diagram graphically displays the details of the correlation.

Delete correlation analyses#

Entry point: Entry points for deleting correlation analyses can be found in the overview and detail pages, each in the three-point menu.

Regression models#

Regression models are statistical models for investigating the nature of the relationships between a target variable and one or more influencing variables. Based on a regression model, forecasts, i.e. predictions, can be made for the target variable.

In the field of normalisation of key figures according to ISO 50 006, regression is used in the context of energy management to prove the effectiveness of optimisation measures. In this process, disturbance factors and their influence are first determined and then eliminated in the evaluation.

For example, when comparing the heating load after optimisations in the heating system, the influence of a cold or warm winter can be determined and thus eliminated in the evaluation.

In the area of process optimisation, regression can be used to serve as a virtual sensor, e.g. in the collection of quality data.

Entry point#

Click path: Menu > Regression analysis > Regression models

It opens the regression model overview page. This page is the central administration point for your regression models.All existing regression models are displayed here.

Interaction possibilities#

Am Ende jeder Zeile befindet sich ein Dreipunktmenü, das die Bearbeitung von Metadaten und das Löschen der Regressionsmodelle ermöglicht. Durch Klick in eine Zeile wird in die Detailseite des gewählten Regressionsmodells gewechselt.

Create regression models#

Entry point: A new regression model can be created via the regression model overview page.

Click path: Menu > Regression analysis > Regression models

To create a new regression model, click the + Add model button on the overview page. A dialogue opens.

Function:

Enter a name for the new regression model. After confirming the dialogue, you will automatically be redirected to the detail page of the newly created regression model.

Copy regression models#

You can create a new regression model by copying an existing model.You will find entry points for this on the overview and detail pages in the respective menus.

When copying, you can assign the name for the new regression model. You will then be redirected directly to the detail page of the newly created regression model.

Calculate regression models#

Add measuring points:

On the detail page of a regression model, measurement points can be added from the tree by drag & drop in the edit tab of the page. The first measuring point is set as the target variable. All further measurement points are set as influence variables.

Turn influence variable into target variable:

Click on the three-dot menu at the end of the line of the influence variable you want to set as a target. Select the option Set as target variable. The influence variable is set as the target variable. The previous target variable, on the other hand, is included in the influence variables.

Remove measuring points:

Via the three-point menu at the end of the line of an influence variable, it can be removed from the view via the Delete option. Target variables cannot be removed.

Calculate regression model:

First add all the measurement points that are to be included in the regression model and determine the target variable.

Then use the time selection to choose the time range that is to be taken into account for determining the regression model between the measuring points.

Then click on the Calculate Model button.

A dialogue opens in which the calculation settings can be made. It also contains info icons that show further information about this step when the mouse is placed over it.

Select the aggregation interval is. The aggregation interval is set for all measuring points involved. Select the aggregation interval None to use raw data.

Click the Start Calculation button to start the model calculation.

You will then be redirected to the model tab within the page. If the model creation takes longer, you will be shown a loading animation. You can leave or close the page at any time and return later. The calculation will continue to take place in the background and will be available the next time you call up the page.

Once the calculation is complete, the cross-validated model is displayed.

The upper chart shows you two series of values. On the one hand, the actual values that were used to determine the model according to your calculation settings. On the other hand, the values determined on the basis of the model for this period.

The second chart shows you the deviation between model prediction and actual value. This deviation can be switched between absolute and percentage representation.

Below you will find a table showing statistical measures for linear regression that say something about the quality of the model.

Delete regression models#

Entry point:

Entry points for deleting regression models can be found in the overview and detail pages, each in the three-point menu.

Regression analyses#

The regression analysis itself is the application of a previously created regression model to an evaluation period. The model, on the other hand, is based on a reference period.

The regression analysis compares how a value actually presents itself and how it presents itself in calculated form on the basis of the regression model.

This representation can be used, for example, after an optimisation measure in the energy sector. The regression model was created on the basis of a reference period before the optimisation. Subsequently, a regression analysis is carried out for the evaluation period after the optimisation. The comparison of measured values and values calculated on the basis of the model shows the effect of the optimisation measure.

Entry point:

Click path: Menu > Regression analysis > Regression analyses

It opens the regression analysis overview page. This page is the central administration point for your regression analyses. All existing regression analyses are displayed here.

**Interaction possibilities:**At the end of each row is a three-item menu that allows editing metadata, copying and deleting the regression analysis.Clicking on a line takes you to the detail page of the selected regression analysis.

Create regression analyses#

Entry point:

A new regression analysis can be created via the regression analysis overview page.

Click path: Menu > Regression analysis > Regression analysis

To create a new regression analysis, click the + Add Analysis button on the overview page. A dialogue opens.

**Function:**Enter a name for the new regression analysis and select the regression model to be used for the analysis. After confirming the dialogue, you will automatically be redirected to the detail page of the newly created regression analysis.

Copy regression analyses#

You can create a new regression analysis by copying an existing analysis.

You will find entry points for this on the overview and detail pages in the respective menus.

When copying, you can assign the name for the new regression analysis. You will then be redirected directly to the detail page of the newly created regression analysis.

Perform regression analyses#

Call up an existing regression analysis from the overview or create a new regression analysis. Both options take you to the regression analysis detail page.

Here you can adjust the evaluation period to apply the underlying model to a desired period.

Delete regression analyses#

Entry point: Entry points for deleting regression analyses can be found in the overview and detail pages, each in the three-point menu.