This table has a special indexing which is related to the time step be-tween subsequent predictions and the time step bebe-tween each retrieval of new predictions. If the prediction horizon is 36 hours and the time step between each prediction is 3 hours, then this table will contain 13 columns. This table also has another interpretation, which is based on only two columns, a timestamp column and a data column, where the timestamp column contains the valid time for the predictions contained in the data column. This representation is obtained by using the predic-tion table as input to a filter table, which transforms the input table into a table only containing two columns, a timestamp column and a data column. This filter table is e.g. used to construct prediction plots.
5.6 Clients and user interface
The system architecture as presented in the previous sections, is ideally suited for a multitude of small clients, each dedicated to a particular task. E.g. depending on which information is wanted by a particular user, or a group of users, dedicated clients can be developed. Some users might only be interested in knowing the predictions for one larger area, while other users might be interested in knowing the current running conditions at a particular farm, e.g. the maintenance crew. When a new wind farm is constructed, this farm has to be added to the system and the system has to be re-configured, this task can also be handle by a specific client.
The idea behind the client and the client user interface which is presented here, is to provide a way for the user to access and perform all the above mentioned tasks. The user can configure the client to his or hers particular needs.
5.6.1 Main user interface
Figure 5.3 shows a screen shot of the graphical user interface (GUI) implemented in the Zephyr system. The GUI layout is project oriented, where a project is illustrated by a project tree. The system operates
with two project definitions, and these two projects are illustrated by graphical tree components on the left hand side of the GUI. The lower left tree illustrates the project defining the system setup of a particular utility, and the tree in the upper left corner illustrates the project of a particular user.
The system setup of a particular utility is organized hierarchically. The basic objects in the system setup tree are regions, wind farms and tables of data, where data related to the regions and the wind farms are stored.
These objects or the tree nodes are directly related to the business ob-jects described previously. The region object defines the properties of a larger region, containing one or more wind farms. The properties of a region are e.g. - the wind farms in the region, the model which is used to calculate total power predictions for the region and tables of measured and predicted total power production for the region. Similarly, the wind farm object defines the properties of a wind farm, like the model for cal-culating the wind farm power predictions, and tables of data containing measured variables at the wind farm. Furthermore, the wind farm object contains the state of the wind farm, i.e. - if the farm is currently running or not and the current wind farm power production. The system setup is customizable via the system setup tree and node specific popup menus, allowing super-users to customize the system setup, i.e. wind farms can be added or removed, regions can be redefined, the models connected to the wind farms and the regions can be changed and so on. These changes are applied at runtime, and the system does not need to be restarted.
Such a change will in general correspond to adding, removing or chang-ing a business object in an object store. Therefore, these changes will be visible to all clients, and, as this is the system setup or utility wind farm setup model, only super-users are allowed to perform such changes.
The user project tree holds the structure and definition of the windows which are shown in the right hand side of the GUI, each window is il-lustrated by a node in this tree. The user has the option to customize how the windows are organized to fit his or her’s particular needs, by inserting, moving and deleting the position of the window nodes in the user project tree. Apart from the default windows defined by the sys-tem, like the wind farm window, windows showing plots of predictions and measurements, the user can create customized windows either using scripts or via dialog windows. When a new window is created, the user
5.6 Clients and user interface 65
Figure 5.3: Screen-shot of the Zephyr graphical user interface
has to select the node in the tree where the new window will be placed as a sub-node. Each user is not limited to having only one project tree, the user project tree can be saved and reloaded, allowing the users to op-erate with several trees, i.e. the user can define several user trees, where each tree is fitted to one particular operational situation. Each window node in this tree links to one or more business objects, illustrating the information contained in the business objects.
5.6.2 Internal windows
The right hand side of the GUI screen shot in Figure 5.3 shows some of the windows in the system, such windows are referred to as internal windows. The windows which are shown in Figure5.3are: the map win-dow, a table winwin-dow, a prediction plot, the wind farm status window and the script window. The map window shows the area which the system is operation on. This map contains wind farm symbols, which illustrate the current running conditions at the wind farm, i.e. the wind farm sta-tus. Furthermore, the map window servers as background for weather animations created from the numerical weather predictions. The cur-rent implementation is capable of showing the predicted meteorological variables as contour animations.
Some of the internal windows in the Zephyr client are shown below.
Figure 5.4 shows a time series plot window, and Figure 5.5 shows a combined scatter and line plot of a power curve.
The slider on the top and the scroll bar on the bottom of the time series plot can be used to specify the visible time period of the time series. The slider specifies the time window width, and the scroll bar specifies the start and end times.
Figure 5.6shows the dialog window for adding a layer to a plot window.
A layer defines the x and y data, the axis settings and the plot type, i.e.
scatter or line. The data is selected by dragging a node, which represents a column, from the tree to the fields in this window.
Data can be edited via the table window, an example of a table window is shown in Figure 5.7. Editing a number in a table from a client
auto-5.6 Clients and user interface 67
Figure 5.4: Time series plot window.
Figure 5.5: Power curve plot window.
Figure 5.6: Plot layer window.
matically updates all relevant plots, tables, etc. in all connected clients.
Such an actions does not automatically force a model update. If e.g.
data related to a wind farm is modified then a wind farm model update has to be requested via the wind farm node popup menu.
Figure 5.7: Table window.
Figure5.8 shows the wind farm status window. This window illustrates graphically the current state of a wind farm. The state is defined as the current production, shown as the dark fraction of the wind turbine tower, the direction, shown by the dark gray turbine blade and other variables shown by the meters in the right hand side of the window.
5.6 Clients and user interface 69
Figure 5.8: Farm window.
Chapter
6
Conclusions
This thesis deals with a number of aspects related to short-term predic-tion of wind power. The thesis is structured as a summary report and a collection of ten research papers. In the summary report the background and the motivation for the Ph.D. study are outlined. Bibliographic notes to previous research within the field of short-term wind power prediction are provided. The meteorological theory, which is particularly relevant for wind power prediction is described and the physical approach used in the system developed by Risø National Laboratory is analysed and described. A short introduction to the statistical models that have been considered is provided together with bibliographic notes to the litera-ture for a more detailed description of the models and in particular the estimation methods for the models. Some general consideration with regard to statistical versus physical modelling is outlined, and possible areas of combined statistical and physical models for short-term wind power prediction are described. Finally, the summary report describes the client/server software application that has been developed. The in-cluded papers can be categorized into two groups, development of general statistical models and methods, and development of dedicated short-term wind power prediction models.
The conclusions will be structured according to the main areas consid-ered in this thesis and in the Ph.D. study. In short, these areas are the development of general statistical models and methods for on-line estima-tion and predicestima-tion, development of models and methods for short-term prediction of wind power and implementation of the prediction models in a client/server software application.