Wind power definition with forecasting system

Wind power forecasting systems

 

The accurate knowledge of wind production and its evolution is fundamental to achieve effective integration of wind energy in the electrical system.

As this information is necessary to match production and demand, to prevent grid failures and to operate in the electrical market.

In this sense, the accuracy in wind power predictions helps to optimize plans of wind energy producers and to reduce the reserved capacity and, consequently, the electricity cost.

The main tools thought to obtain the necessary wind energy information for the electrical system participants are wind power forecasting systems.

However,

wind control centres are becoming another critical source of information for efficient wind energy integration.

Both elements are complementary as wind power forecasting systems requires information about wind farms production to perform accurate predictions, being this information provided by wind control centres.

In the same way, wind farms are operated and scheduled according to the wind predictions. Thus, wind control centres generally receive data from one more wind power forecasting systems.

 

Wind control centres

The information about the current wind energy production and the state of the wind farms is obtained via the wind control centres. Wind control centres not only monitor but also control the production and configuration of the wind farm.

The monitor/control chain starts in each individual wind turbine and, in some cases, ends in regional control centres of the system operator. This is the case of the Spanish electrical system which will be the reference in this subsection.

Manufacturers generally include a supervisory control and data acquisition (SCADA) system as a part of the wind farm supply. The SCADA communicates with the sensors and actuators of the wind turbine, collecting the acquired data and allowing to send commands to configure the wind farm properly.

Other article: Gas turbine

Most important wind energy companies manage wind farms distributed over large regions, even different countries.

In order to make economically viable the surveillance and management of their wind farms, they have installed wind control centres to assemble the data from the different plants and/or their local SCADA systems. Thus, the personnel of the wind control centre is able to quickly detect incidents and operate the wind farms under control.

Each control centre is designed according to the amount of data which have to be managed. The basic elements in the control centre are the communication systems, the computer systems and the wind farms and wind data models.

The communication system is critical in the control centre performance. It must assure stability and low latency to obtain and send information.

Another article: Tariff

Typically, communications are performed via the internet by using virtual private networks, but in some cases fibre optics dedicated lines or satellite links are used. The computer system generally comprises one front end unit for every wind farm linked to the control centre.

Each unit is associated to a console to show the received information. The information is generally summarized and represented in a video wall to have an overview of the state of the associated wind farms. The computer system also includes a cluster to store historical data and a web server to provide connections to the web clients.

Managing different wind farms requires a standardization of the transmitted data, as each of them could implement different technologies, protocols or configurations. Wind farms models are thereby required to allow an easy and uniform management from the control centre.

Wind forecasting models are also necessary to perform an efficient management of the wind farms. Fundamentally focused on production control and market bids, but also for maintenance operations in which calm periods of several days could be necessary.

Another article: Diesel power plant

Spain is a reference in regulating the renewable energy production via control centres. In 2007, it was established that:

all facilities of renewable energy whose rated power exceeds 10 MW must be assigned to a generation control centre. Which acts as an interlocutor to the grid operator, transmitting information in real time and enforcing to execute the instructions received in order to ensure the reliability of the electric system.

According to this directive, the Spanish transmission system operator (Red Electrica de Espana) created the Control Centre of Renewable Energy (CECRE). The CECRE homologates and receives information from other control centres which, in turn, are connected to renewable energy plants.

Currently, there are 47 homologated control centres which act as interlocutor with CECRE. And through which the system operator can send commands to renewable energy installations to assure a maximum integration of renewable energies while maintaining grid stability.

##

During these years, CECRE has demonstrated its benefits for integrating renewable energies and, thus, from 2015, its operation range has been expanded to facilities of 5 MW.

Focusing on the wind farms, CECRE demands information about the output voltage and active and the reactive power, as well as other meteorological information of the emplacement such as wind speed, wind direction and temperature.

These data must be updated every 12 s. According to this information, and via the homologated control centres, the system operator sends curtailment commands to the wind farms which must be fulfilled in 15 min.

If levels are continuously infringed, the system operator could even shut down the wind farm. If connection with CECRE is lost the wind farm is automatically curtailed at the 85% of the last power level.

 

Description of wind power forecasting systems

A wind power forecasting system is a tool which processes all the relevant data in order to generate estimations of the wind energy production. The consolidated wind power forecasting systems are run by companies which sell their results to system operators and wind energy producers.

However, the most important wind energy companies and system operators with high penetration of wind energy generally run their own wind power forecasting systems.

Figure 3.3 shows a typical scheme of a wind power forecasting system. The energy production of a wind farm depends on the local wind conditions of the emplacement. But these conditions are driven by the weather at the regional scale. Hence, accurate wind power forecasts should include information about regional weather conditions.

Numerical weather prediction (NWP) models represent the main tool to provide global and mesoscale descriptions of the atmospheric conditions and their evolution. Typically, these models generate gridded predictions of atmospheric variables in steps of 1–3 h with a final forecasting horizon of some days.

##

However, the coarse resolution of the output grid and the absence of local considerations in the estimations advise against the direct application of NWP data for accurate wind energy estimations. Thus, a downscaling process is required in order to include local features in wind predictions. As a result of this process, the NWP data is refined and adapted to the topography surrounding the wind farm.

In this refinement, it is desirable the inclusion of in situ weather stations to help in modelling local phenomena which could not be inferred from the mesoscale data. Once local wind forecasts are obtained, they are processed along with wind farm data to obtain wind power forecasts, which is the final goal of a wind power forecasting system.

In this point, wind farm models, information about the current production and historical data are critical elements for accurate estimations. Hence, wind control centres play a fundamental role in this stage.



Figure 3.3 Typical scheme of a wind power forecasting system

 

The output of the wind power forecasting system is configured depending on the objective.

Since not all actors in the electricity system are interested in the same information. Thus, wind farm owners demand a local prediction with horizons from several hours to days.

##

The information allows to establish schedules and to supply bids in the day-ahead electricity market. And, in the same way, schedules adjustments and re-bidding in the intraday market. Transmission system operators require similar information for reserve requirements and wind power projections but adapted to a regional framework.

They also demand additional information related to shorter forecasting horizons (seconds to minutes) for real-time management of the grid. In this sense, NWP data become more relevant as forecasting horizons increases.

On the contrary, wind farm and weather stations data prevail in very short-term forecasts. Errors in predictions of wind power forecasting systems will be lower for shorter forecasting horizons. But also for simple terrains where local conditions are weaker.

In the same way and as commented along the text, wind power predictions at regional scales will be more accurate as the aggregated errors tend to cancel out, i.e., the local conditions are smoothed.

Following the description of the Spanish case in the previous subsection, the Spanish transmission system operator (REE) has developed its own wind power forecasting system: SIPREOLICO.

SIPREOLICO processes NWP data from the European Centre for Medium-Range Weather Forecasts along with historical information about each wind farm to generate hourly wind power predictions for the following 48 h.

The combination of CECRE and SIPREOLICO brings an excellent framework for efficient integration of wind energy in the Spanish power system.

##

Foley et al. compile a set of wind power forecasting systems, detailing the countries in which they are applied. And the institutions and companies in charge of their management and development.

In this compilation, it is shown how these wind power forecasting systems are mainly based on hybrid approaches.

There are wind power forecasting systems which have been developed for a specific area or electrical system and exploited by a unique organization.

However, as a result of the increasing demand for this information. Some wind power forecasting systems have become more adaptable in order to offer their results as commercial products according to customer requirements.

In this sense, several of these power forecastings have become commercial spin-off of research projects. For instance, ANEMOS Wind Power Predictions is the result of collaboration among numerous European organizations since 2002.

 

Wind power forecasting system results: representation and validation

As said, wind power forecasts must be adapted to the necessity of the different actors of the electrical systems. In first steps of wind power forecasting, single estimations were the main option to provide this information. These single estimations can be assimilated to deterministic predictions as calculations lead to a unique solution.

Improving single estimates is a direct way to enhance wind power forecasting system results. But, the single estimations of the wind energy production do not reflect the uncertainties derived from the different elements in wind power forecasting systems.

Thereby these results are not fully useful for decision-making strategies in the electrical system. In this sense, wind energy bids or reserve requirements can be benefited by a probabilistic or risk representation rather than a single scenario which, in a high percentage of cases, does not actually occur.

##

Thus, in recent years, wind power forecasts are expressed via wind power uncertainty analysis. Evaluating the uncertainty inherent in the results of wind power forecasting systems.

There are three main ways to express the results in wind power forecasting uncertainty analysis:

Probabilistic forecasting, scenario forecasting and risk index (Figure 3.4).

[caption id="attachment_4069" align="alignnone" width="206"] Figure 3.4 Wind power forecasts uncertainty analysis: interval representation (left) and scenarios representation (right)[/caption]

##

In probabilistic forecasting, wind power is treated as a random variable. Results are mainly expressed as probability density functions or quantile/interval forms.

In the case of probability density functions, the problem lies in determining the type of distribution and adjusting its parameters. This representation is appropriated for very short-term prediction, and due to its simplicity requires a low computational cost.

The quantile/interval form is focused on providing upper and lower limits for the energy production generally based on empirical distributions.

In horizons from hours to days, these approaches perform better than the ones based on probability density functions. But large datasets and high computational costs are required.

##

Probabilistic forecasting generates predictions for each look-ahead time independently, without considerations of its evolution in time. However, decisions about market and grid operations would be benefited by a time-dependent description of the wind power uncertainty.

For this reason, scenario descriptions are built as a set of single forecasts over a period. Thereby providing a temporal evolution of the wind power uncertainty.

The level of uncertainty is derived from the dispersion of possible scenarios. In this sense, larger numbers of scenarios involve a better description of the uncertainty but requiring more computation.

Market and dispatching decisions are made based on the most likely scenario. And which is assumed to be the closest scenario to the actual situation. Thus, scenario descriptions allow real-time updating as new wind power and meteorological data is received.

Another article: Solar panel

The risk index is the simplest way to express wind power uncertainty. It basically consists of a single-valued forecasting with additional information about its reliability. Generally, it represented by a number or a colour code. Two examples of risk indexes are the Meteo-Risk Index and the Normalized Prediction Risk Index.

Wind power forecasting systems are generally focused on providing estimations for individual wind farms. But the system operator is interested in regional wind power forecasts. The aggregated power output of the wind farms in a certain area.

Obtaining specific forecasts to compute the total power in a region is more difficult than a direct estimation. Furthermore, the direct regional forecast is faster and more accurate; due to the commented smoothing effect of the aggregated wind production over an area.

##

Regional forecasts are generally obtained upscaling online measurements from the concerned wind farms along with NWP data. Uncertainties in regional forecasts are better described by the scenario representation. As it is suitable for stochastic optimization problems with temporal and spatial correlations.

The evaluation of the wind power forecasting system results depends on the manner in which results are expressed. Single wind power forecasts are generally assessed by comparison with a reference model. The persistence model is the most used reference model in these cases.

The persistence assumes that the predicted wind speed in a look-ahead time will be similar to the last measured value. This naive approach establishes a lower limit from which the contribution of other models can be measured. An adaptation of the persistence method for regional forecasts is described in.

Common statistics are used to quantify the accuracy of predictions as mean absolute error, mean squared error or correlation coefficient. A compilation of the most frequent statistics used in wind forecasting and their formulas can be consulted in.

##

If results are expressed using probabilistic descriptions, the validation is more complex. Pinson et al. propose some properties to evaluate probabilistic forecasts: reliability, sharpness and skill score. Reliability informs about how the empirical proportion derived from actual values match the predictive quantiles.

In a model with the highest reliability, the actual values would be distributed according to the predictive quantiles. Sharpness gives a measure of the concentration of the probability distribution.

Good models should have a high sharpness, thereby being the probability concentrated in a short-range of possible values.

Skill score has different implementations, as it is defined depending on the uncertainty representation. The scoring rules selected to test the results.

 

আমাদের অন্যান্য পোষ্টসমূহ:

BCS Preliminary: All Bcs Question, Answer and Solve.

IT solutions: Google Adsense Approval পাওয়ার করণীয়, কম্পিউটারের  Speed  বাড়ানোর কিছু উপায়

Engineering: Solar cell, solar plant for 100 kW Installation, এসিতে বিদ্যুৎ খরচ কমানো যায় তার বিস্তারিত উপায়

Education: Importance of Reading Newspaper, Duties of a Student Full Composition, ‍School Library

Biographies of famous people: Rabindranath Tagore, Prophet Muhammad (Sm), APJ Abdul Kala