In this article you will learn how to use the Automatic Baseline Calculator, available in the Apps Market inside the Platform.
Contents of this article:
This application allows you to calculate baseline formulas via linear regression that you can use in Measure and Verification projects. You can find more information about calculating baseline formulas in this article.
In order to be able to use this app, you will need first to install it through the platform's Apps market.
You will also need to have data for the consumption device you want to calculate, and any variables that you would like to include in the linear regression analysis.
How does it work?
Once the app is installed, you can access it in the Analysis section, under Automatic Baseline Calculator. The app is divided in two tabs: Baseline calculator and Measure and Verification (M&V) export.
- Baseline calculator
- Analysing the results
- Create a Measure and Verification (M&V) project with the calculated baseline formula
This tab allows the calculation of the baseline formula. In the first section, the target for the baseline calculation is selected, as well as any variables that want to be included in the analysis.
Selecting the variables
Any devices available on the account can be selected here.
Filter by location and select the device and its parameter to use as a target for the baseline calculation. Then repeat the process for all variables that have to be taken into account for the calculation. A variable name will be assigned to them and they will be listed at the bottom of this section.
Select the reference period and frequency for the baseline.
If you do not have weather data yet, you can test weather variables (temperature, degree days) using an external virtual weather stations directly in the app and use them to calculate a baseline formula. Please note that you will not be able to export this formula in the M&V module, as it needs accepted devices in the account.
Using an external virtual weather station
If you want to use external virtual weather stations, check the box "Add external weather".
Select the country, postal code and virtual station that you want to use. It will default to the post code configured in the current location.
You can choose which external parameters to include in the analysis: Cooling Degree Days (CDD), Heating Degree Days (HDD) and Temperature.
If you click on advanced options, you can choose the optimisation method for degree days (Mono pivot or poly pivot) and choose the step and minimum and maximum thresholds.
If you already have weather data in your account (because you imported it and calculated Degree Days, or because you are using a virtual weather gateway), don't check the "add external weather" box and continue on to the next section.
The method for the calculation of the degree days will vary between UK Met Office and COSTIC, depending on which method is selected in the location configuration to which the target device for the formula calculation is assigned.
Once the target variable, the variables to be included in the analysis, the reference period and the frequency are selected, the baseline calculation settings can be customised.
Under "Time parameters and interaction", different parameters and interactions can be added to the baseline calculation:
- Weekend and week days can be added as parameters only if the frequency selected under "Select frequency" is set to Daily, as shown in the example below.
- Months can be added as parameter
- Squared exponents of the parameters, for instance HDD * HDD
- Interactions between parameters for instance HDD * Occupation
The key code for all selected parameters as well as any operation and interaction will be listed under "All selected parameters"
The "Calculate baseline" button will calculate the formula with the current configuration.
For more advanced settings, select "Advanced options" which will display other customisation options:
- Select parameters metric: Choose wether you define P-value or T-value as significance parameter for the regression analysis. Both will be calculated, but the threshold is defined using only one of the methods.
- Parameters significance threshold: Define the threshold for the chosen parameter. For P-Value, 0.05 is the default threshold value for the variables in the calculation, for T-Value, it is 2.
- Select the optimisation metric for the baseline calculation between the following options:
- Coeffient of multiple determination - Adjusted (Ra2)
- Coeffient of multiple determination (R2)
- Bayesian information criterion (BIC)
- Test sample on reference period: Select if you want to set aside a % of your data, to check the model with that data. Only recommended for analysis in a daily frequency or higher, or if several years of data are available.
- Compare baseline with degreedays [18ºC]: This add a comparison of the baseline using degree day data calculated with 18ºC as reference for both heating and cooling degree days.
- Discard period: If there is data that should not be included in the calculation for given periods, they can be selected here. Any periods selected will be listed on the bar on the right and can be removed if necessary.
Once the configuration is set, hit the Calculate baseline button, and the algorithm will run to calculate a baseline formula.
Analysing the results
Once the baseline calculation has run, the app will present the results obtained. In this example, daily frequency was selected, and the weekend was selected as a variable, along with the Heating Degree Days and the Cooling Degree Days, in order to create a baseline consumption that models the consumption of our main load. For the Advanced Options, the default settings were used (0.05 p-value, and Coeffient of multiple determination - Adjusted (Ra2)).
The obtained formula, which uses all the variables introduced for this example, can be seen under "Automatic baseline". The coefficients for the statistical tests are listed below the formula.
Further information on the statistic tests for each variable can be obtained by hitting "Parameters details" tab.
The tabs below display different information regarding the baseline calculation:
- Comparison of the calculated baseline comsumption and the real consumption
- Acumulated error in absolute and relative (%) values
- Metrics and statistics parameters of the regression analysis
- Regression residual charts
- Regression details: this is the direct console output obtained from the regression calculation
You can dowload all the baseline data in a csv file by clicking on the "Download baseline data" button.
Compare with a manual baseline
Once the baseline is calculated, you can add a manual baseline formula to compare it to the calculated baseline. Write the baseline formula in the textbox and click the "Add manual baseline" button. You will be able to see the comparison in the different analysis tabs.
Create a Measure and Verification (M&V) project with the calculated baseline formula
If no external weather stations were used as variables for the analysis, the baseline obtained can be used to create a Measure and Verification project in the platform.
Click on the Measure and verification (M&V) tab on the top menu to access the export section:
Select the evaluation period for the project, a savings target and the baseline maximum error in percentage. The preview button allows to previsualise the resulting measure and verification project, compared to the real consumption during the evaluation period, as long as there is data for all the baseline variables and the real consumption device for that period.
The charts displayed here have the same behaviour as the charts in the Measure and Verification section. You can alternate between the accumuluated savings and the target comparison charts by selecting the corresponding tab. Click here to know more about this powerful analysis module.
Click on the "Create new M&V project with these details" button in order to create a project with this configuration. Once created you can click on "View M&V project" button that will appear in order to analyse the project in the M&V section.
This application is free with professional licenses.