Business Intelligence Solutions for Wind Power Plants Operation

The European Union energy strategies imply significant changes in the power systems. These should contribute to sustainable development and protection of the environment by enabling the EU to achieve its targets set in the third package of legislative proposals for electricity and gas markets. Renewable energy sources (RES) in Romania have been encouraged since 2007 and this lead to a large volume of projects. It took several years to have the first MW installed in wind power plants (WPP). Nowadays, the installed wind power in Romania is about 2642 MW, most of them (about 85%) being concentrated in the south-eastern part of the country. Based on recorded data during four years in Romania, a couple of analyses have been performed. They contribute to a better integration of wind energy into power systems. This paper will describe how business intelligence solutions are applied on data regarding wind power plants operation and main conclusions that could be drawn.


Introduction
The EU's energy and climate policy objectives consist in completing the internal market in energy, guaranteeing security of supply, notably for gas and oil, reducing greenhouse gas emissions by 20%, increasing the share of renewable energy in the final energy consumption to 20% and achieving a 20% increase in energy efficiency by 2020 [1].Romania as one of the State Members has to fulfil its obligations related to EU's targets in terms of RES integration.The incentive support scheme for RES has been enacted by Law no.220/2008 for establishing the promoting scheme for energy produced out of RES, Law no.139/2010 (modifying Law 220/2008) and a series of four governmental orders dated November 2011 [2].As a result of the supporting scheme mainly based on green certificates, since 2007, a large number of projects have been noticed.Most of them are located in Dobrogea, Moldova and Banat areas as in Figure 1 [3].

Fig. 1. Areas with large projects in Romania 1
This concentration of interest from the private investors coincides with the wind poten-tial map as in Figure 2 [4].

Fig. 2. Wind potential in Romania
Starting from 2010, installed power increased from 13 MW to about 400 MW by the end of the year.In 2011, the installed power was almost double (700 MW) compared with the previous year.The maximum installed power was recorded in 2012 (800 MW), then in 2013 it decreased up to 500 MW and in 2014 as forecast it will decrease even more (270 MW).This evolution is well-related to the specific legislation that incentives RES development.These figures are given in Table 1.Due to the fact that the installing process is very dynamic, the total figures are approximate.Its trend is ascending, but in the last two years, it has been moderated by legislative means.For 2014 the increase for the first half of the year was double so that to get an approximate value for entire year.

Data Integration and Analysis Models
In the research project [5] we aim to develop a set of templates for data integration in a central database within the online platform, to define a set of performance indicators at macro level and to develop analytical and interactive reports for monitoring these indicators intended for supporting decisions.We'll propose simulation models of the operation of power plants at regional and national level, models that are based on data mining techniques and developed with geospatial elements for tracking indicators through interactive maps.A particularly important indicator on which to base an accurate forecast of the produced energy from renewable sources is the degree of simultaneity of operation of wind power plants located in different geographical areas.Wind energy production is conditioned by several factors factors such as: slipstream effect, soil orography, power characteristics, losses up to the connection point of etc.These factors are identified and detailed in the fundamental works [6], [7].
The analytical component developed for national authorities will contain a model for determining the degree of simultaneity which will allow a more accurate dimensioning of power reserves in the system.Thus, if some of the production companies in a given area will have an accurate prediction system, based on the degree of simultaneity of the model we will be able to determine and correct the estimation of production for those power plants without efficient prediction systems (for e.g.undispatchable units or units that have systems with big errors) as described in [8].
The proposed model will have advanced data analysis capabilities and it can be used to improve decision-making and ensure knowledge management.The component for national operators will allow the streamlining of the information flow, required statements and reports being obtained automatically via the online platform.The prototype's interfaces will be developed so as to allow users single access to the system via mobile devices, and the use of the Cloud Computing platform will allow the connecting of servers, services and applications necessary for the prototype, thus streamlining access to information to decision makers and reducing infrastructure costs.The system will enable effective and real-time analysis of the operation of renewable power plants.Also, using an integrated platform, through which there are monitored and analyzed in real time all the renewable power plants included in the system provides a competitive advantage when integrating with similar networks in the European Union.The first phase of the project involves identifying and analyzing the data sources, by designing the conceptual database diagrams and mappings between data.The conceptual data model will be designed.The system must implement the features of an integrated decision support system, using multidimensional models through which we can implement technological and business workflows.We will define the Business Intelligence methods and technologies used for analysis and data presentation and we'll define the main components of the system based on the following levels: the data level, the model analysis level and the presentation level.But first of all, an analyses of the WPP operation over time is needed, based on the data series recorded in the last 4 years.

Business intelligence analyses of WPP operation
Taking into account the large available recorded data set that describes the global operation of WPP between 2010 and 2013 (over 200000 records), business intelligence solutions will be used.No business intelligence technique has been applied for wind power plants operation until now.Out of data set some interesting results are found such as hourly average WPP output grouped by studied years, comparison among curves that describe hourly average WPP output, relation between WPP output and installed power in WPP in terms of maximum and average values and seasonal analyses on each studied year.Table 5 shows average and maximum values recorded in June.-wind blows more at night that is not helpful for system operation; -summer months were less windy; -if we compare WPP output in the same month in two consecutive years, the difference could be significant; -average WPP output could be considered no more than 30% and maximum WPP output could be considered around 80%.Although in this paper several analysis regarding WPP operation are presented for different consecutive years, it is obvious that more data is required in order to obtain better correlations and more significant conclusion that could be useful in power systems opera-tion.

Fig. 3 .
Fig. 3. Installed power evolution from 2010 up to 2014 Figure 4 depicts average WPP output hour by hour in January.The first three studied years WPP output was almost flat, but in 2013 lower values were recorded around 5 and 12 o'clock and higher values were recorded at 21.Some similarities are identified between 2012 and 2013 curves up to 12 o'clock.

Table 2 .Figure 5 Fig. 5 .Figure 6
Figure 5 depicts average WPP output hour by hour in February.The middle two studied years (2011 and 2012) WPP output is quite similar.

Fig. 12 .
Fig. 12. Hourly average WPP output in September 2010-2013 Figure 13 depicts average WPP output hour by hour in October.This month the level of

Fig. 15 .
Fig. 15.Hourly average WPP output in December 2010-2013 Table 10 indicates average and maximum percentage of installed power recorded in December 2012 and 2013.

Fig. 16 .
Fig. 16.Hourly average WPP output each month in 2013 Figure 17 depicts average WPP output hour by hour each month in 2012.At the beginning of the year about 1100 MW have been installed.By the end of the year about 1900 MW have been installed.In this figure WPP monthly output is compared.In winter time the level of WPP output is much higher than WPP output in summer time.The lowest level is about 120 MW recorded in June and the highest level is about 600 MW recorded in December.As for the rest of the months, the curves are quite close and compact.

Table 3 .
Average and maximum values in Figure 7 depicts average WPP output hour by hour in April.WPP output in 2011 and 2013 was slightly similar.WPP output in 2013 has many windings.WPP output in 2011 and 2012 are quite different, but with little wind-ings.As for WPP output in 2010 is quite flat up to June-July when significant power was installed.Fig.7. Hourly average WPP output in April 2010-2013Table 4 shows average and maximum values recorded in April.

Table 4 .
Average and maximum values in These values are significant because in 2013 the installed power increased over 2000 MW.Average values of WPP output is about 30% and maximum value is almost 88% of installed power.Figure 8 depicts average WPP output hour by hour in May.WPP output in 2013 is opposite with load curve and it does not help balance of the power system.WPP output in 2011 and 2012 are similar, but with little windings.

Table 5 .
Average and maximum values in Average value of WPP output is about 13% and maximum value is almost 65% of installed power.Figure10depicts average WPP output hour by hour in July.All three important curves are similar and again they are opposite with load curve and do not help balance of the power system.This month the level of output is much smaller than in winter and spring time.WPP output in July is similar with WPP output in June.

Table 6
indicates average and maximum percentage of installed power recorded in July 2013.Figure 11 depicts average WPP output hour by hour in August.

Table 6 .
Average and maximum values in

Table 7
indicates average and maximum percentage of installed power recorded in August 2011.Figure 12 depicts average WPP output hour by hour in September.This month the level of WPP output is increasing.

Table 7 .
Average and maximum values in In 2011 and 2012 the curves are similar, but different from 2013 curve.

Table 10 .
Average and maximum values in December 2012 and 2013 [9]winter time the level of WPP output is higher than WPP output in summer time.The lowest level is about 30 MW recorded in July and the highest level is about 220 MW recorded in December[9].