In the Forecast Data interface, there are three possibilities:
You can select different locations, different time and date frames and you can download data needed.
See the following questions for more detail information.
In ELFI we have setup an artifical neural network (ANN) tool for optimisation tasks of advanced users. The ANN is based on the FANN libraries with general public licencing.
There is plenty of documentation available for the FANN library. Therefore, we will focus in this user guide only on the specific setup and usage of the FANN library in ELFI. Documentation of the FANN library can be found by following this link to the developer http://leenissen.dk/fann/wp/
With the ELFI ANN interface you can create and train an optimised forecast with an Artificial Neural Network (ANN) algorithm to predict the future of a variable based on different input forecast curves from the MSEPS system. In other words, the user can let the ANN search for pattern of the end products and find an appropriate function of these pattern in a training of the ANN. The end result is an optimised forecast with the features that the user defines and that are important for the user.
The special feature of the ELFI-ANN is that you do not have to create an ANN from scratch, but use already pre-computed end products and mix uncertainty estimates in form of minimum, maximum, mean, percentiles and other optimised or weighted combinations of forecasts of the end product parameter together and form an ANN that is optimised on the parameter(s) of your choice and requirements.
Details on how to create an ANN can be found in the FAQ section Artificial Neural Network or by following this link: ANN
In the interface “Curve Calculation” you can copy timeseries of forecast data and/or measured data into other timeseries variables for data manipulation such as optimisation to a specific target.
In practice this means that you can copy a timeseries of data from a selected data curve into another data curve. All data is scaled by comparing the maximum value of the source curve with the maximum value of the target curve.
In the menu curve-calculation we address specific questions and processes of how to manipulate data and use its results for your data analysis.
With the Verification interface a set of standard statistical error measures are calculated for all forecast time series in relation to a measurement time series or a defined true value time series.
The statistical error measures are calculated in the unit of the true value, as percentual values of the installed capacity or as percentual values of the production, where the values will fist be converted to the percentage of the average measured production value. The following error measures are computed and displayed in this interface: - Mean of the timeseries
More details and how to carry out a verification can be found in verification-details
In the forecast tools menu there are 4 applications:
Creating a new curve with a artificial neural network training by searching for patterns within given setup parameters and production values
The Observation Monitoring interface can be used to retrieve statistics about data errors that have been found in measurement data.
To use the interface, please make sure you are logged into the ELFI Interface.
In the main menu, select the box called “Observation Monitoring”.
You will be redirected to a page where you can select for which locations (1) an variables (2) you would like to receive data error statistics.
Please note that you can make a range of locations/variables using the shift key when clicking on a location/variable. This will select all locations/variables starting from the previously selected location/variable.
You may also use the “select all” (1) or “select none” (2) buttons to reset the selection.
You can also define the time span that should be used to accumulate the result.
Lastly you should make sure that the time frame is set correctly. You can click and drag the beginning (1) or the end (2) of the selection to modify the time frame duration, or you can click and drag centrally (3), to shift the time span to another period.
Once your selections are complete, please press the “generate” button and wait a moment for the results to appear.
Once the result page appears, you will have an overview table of the accumulated error counts, sorted by location/variable, and error code.
At the top of the table you may again see you selected time frame. In the very left column of the table you may see your selected locations (bold) and the forecast variable (grey) to which the counts regard.
The error code number is displayed as the heading for each column.
A more descriptive label for the error code can be found at the bottom of the page.
The number in each cell indicates how often that specific error occurred for the location/variable within the selected time frame. The colour shifts from green to red according to the count of errors. A legend for the colours can be found at the bottom of the page.
Clicking on one of the error numbers will open a detail window where each of the errors is listed individually with further information such as the time of occurrence.
Note
The details table is designed to be easily copyable into a Microsoft Excel spreadsheet.