Meteodyn will present 6 posters at the WindEurope Technology Workshop 2022
This year, the WindEurope Technology Workshop 2022 conferences will take place on June 23 and 24 at The Egg, a Congress and Meeting Center, in Brussels.
Meteodyn submitted 6 abstracts, and all were accepted. You’ll then be able to ask your questions to Ru LI and Gang HUANG, our Research Engineers. Don't hesitate to navigate between our posters to meet them.
Our Business Development Engineer, Anthony DIJON, will also be there to give you more information about our software and services.
Wind Resource Assessment
- PO060: Comparison of two methods in atmospheric stratification determination | See below for more details.
- PO069: Mesoscale-microscale coupling method for Wind Resource Assessment in Large Coastal Area in India | See below for more details.
- PO080: CFD-based LiDAR flow curvature correction in complex terrain | See below for more details.
Operation and maintenance
PO010: Performance Analysis of Wind Farm with SCADA data: from the methodology to a case study | See below for more details.
PO044: Wind power forecasting using different weather forecast data sources: a case study | See below for more details.
PO073: Power curve of wind farm fitting base on Gaussian mixture distribution and S-curve | See below for more details.
With a large number of wind farms operating over years, the performance analysis of those farms becomes an important subject. This study tries to propose a general methodology to address this issue and a case study to demonstrate its application.
- Firstly, the failure in measurement system must be detected at the first step. It could be separated into 2 categories: data missing and data erroneous. After being detected, those measurement failure datapoints will be excluded from further analysis.
- Secondly, the total dataset after measurement failure detection will be classified into 7 categories: normal, stop, curtailment, partial stop & partial curtailment (transition between stop/curtailment and normal), over production and under production: output power lower than expected
- The calculation of the real power curve is implemented on normal datapoint only, with the respect to the IEC-61400-12-1 standard. Air density correction is implemented on the wind speed if the measurement of temperature, humidity and pressure are available.
- Then, the losses are estimated by applying the measured speed (corrected if possible) to the real power curve determined in the previous step.
- At last, in order to have a general view on the performance of wind turbines, several KPIs are calculated.
A case study: La Haute Borne Wind Farm
Firstly, the measurement failure is detected and classified based on the cause. Currently, our algorithm is capable to detect 14 categories of measurement failure.
The development of wind energy is crucial to the building and development of a low-carbon and sustainable future worldwide. The intermittency of wind power output, related to the nature of atmospheric wind flow, remains a non-negligible aspect of wind farms, and the increasing share of wind energy in electricity generation and supply is adding more weight to the question of reliability.
Power forecasting based on machine learning methods is a technique to predict power output of a wind farm related to different horizons, as in short term (1 day ahead) or super short term (several hours ahead). The forecast is useful for wind farm operators and power grid management, and with the help of power storage technologies, results in better anticipation and regulation of variations in wind power output.
Two meso-scale weather forecast data sources have been used, the European Centre for Medium-Range Weather Forecasts (ECMWF) and the Global Forecast System (GFS) run by the U.S. National Weather Service. Both datasets provide data of wind speed, wind direction as well as pressure, humidity, and temperature at different heights above ground.
The wind farm production data used in this study derive from ENGIE's open-access SCADA dataset  from the La Haute Borne wind farm in France. The wind farm has an installed capacity of 8200 kW total. For this study, SCADA data between January 2014 and January 2018 have been selected for building and evaluating machine learning models.
Multi-Layer Perceptron neural networks based on TensorFlow-Keras framework is built to conduct a regression task that use meso-scale variables as features to predict wind farm power output at each 10-minute time step. The RMSE (Root Mean Square Error) is calculated daily (144 data points) and then averaged. The accuracy score is obtained from one minus the ratio between the average RMSE and the installed capacity.
To measure the model performance, a split of the whole dataset has been made to obtain a train dataset and test dataset, which represents a little longer than 1 year of data for the test dataset. In the search of hyperparameters, it is found that the accuracy score obtained based on cross-validation is the best indicator of model accuracy on unforeseen test data.
Based on power output data between January 2017 and January 2018 during which both GFS and ECMWF data are available (about 50,000 samples), a comparison has been made on the accuracy score of models trained exclusively with one of the mesoscale data sources. Preliminary results show that the ECMWF-based model has higher accuracy scores than the GFS-based one, whereas the hybrid model is slightly more accurate than ECMWF.
Using weather forecast data from ECMWF and GFS databases, this study focuses on the day-ahead prediction of wind farm power output. For wind farm and grid operators, forecast results from an appropriate mesoscale weather forecast data source can help anticipate wind farm power output and thus reduce the uncertainty in decision making related to integration of wind energy into power systems.
The atmospheric stratification, caused by the daily cycle of the surface heating and cooling, impacts a lot on the wind profile extrapolation to the hub heights in wind resource assessments. The atmospheric stratification could be an unstable stratification when the surface is heated during the day, or a stable stratification when the surface is cooled at night.
The atmospheric stratification is estimated with the Monin-Obukhov length or Richardson numbers etc. which depend on the heat flux. When the heat flux isn't available, it could be obtained by the temperature profile based on Monin-Obukhov similarity theory (MOST). Using the temperature profile and the wind speed, Meteodyn WT calculates the Monin-Obukhov length iteratively from a guessed Monin-Obukhov length. The algorithm searches iteratively the better approximation of the Monin-Obukhov length.
Svenningsen  proposes a novel method to quantify atmospheric stability in the absent of temperature measurement. Based on the standard wind measurements, the high wind speeds turbulence and wind shear are selected to estimate the neutral stratification in given direction. Then the wind with low turbulence and higher wind shear than the neutral wind is considered as the stable stratification. And the wind with high turbulence and lower wind shear than the neutral wind is considered as the unstable stratification.
The method of Meteodyn WT and the Svenningsen's method are compared using the Cabauw site data. At the Cabauw site, the air temperature and the wind speed are measured at heights between 10m and 200m for four years. The monthly and daily wind data analysis shows that the two methods can distinguish the seasonal and daily variation of the thermal stability. The method of Meteodyn WT demonstrates more stable stratification at night regarding the daily variation. And it presents more evident thermal stability in winter regarding seasonal variation.
 L.Svenningsen, René M.M.Slot and Morten L. Thogersen, A novel method to quantify atmospheric stability, Journal of Physics: Conf.Series (2018)
PO069: Mesoscale-microscale coupling method for Wind Resource Assessment in Large Coastal Area in India
In the recent year, the large wind farms on complex terrains are growing fast. The simulation of large wind farms on complex terrains becomes a great challenge. Considering the orography and roughness, the micro-scale wind farm simulation characterizes the local wind, via Reynolds-Averaged-Navier-Stokes equation and the turbulence model. The wind climate condition at the large wind farms, the essential thermal stability effect, was rarely considered but now draws more and more attention.
The wind resource assessment modeling on the large wind farm focus on three methods: the micro-scale method through solving the thermal effect in micro-scale model; the meso-scale method through refining the resolution to investigate the near ground wind flow in meso-scale model; and the meso-micro coupling method through introducing the meso-scale wind characters into the micro-scale models. The advantages of meso-micro coupling method are that the climate data which cannot be obtained easily in micro-scale model, like the temperature, the heat flux, and the humidity etc., can be resolved in meso-scale model. Then, the time-series or the statistic climate data from the meso-scale model are introduced into the micro-scale model as initial or boundary conditions. The meso-micro coupling method captures better the orography and roughness impacts on wind characters than the meso-scale method because of the fine grid resolution at the surface.
A meso-micro coupling method with volumetric forcing is developed in Meteodyn WT. The statistic wind profiles from meso-scale model are used not only as the initial and boundary conditions but also as forces in the computational domain to maintain the statistic meso-scale wind profiles in meso zone. In Meteodyn's mesoscale-microscale coupling method, the Monin-Obukhov length is obtained from the meso-scale data and is used to decide the choice of thermal stability classes in Meteodyn WT. The thermal stability classes range from the unstable to stable conditions. This stability classification is based on Monin-Obukhov Similarity Theory (MOST), in which the turbulent length scale LT, which is used to calculate the turbulence viscosity, depends on the Monin-Obukhov length.
A large coastal area of 120km at Rangewadi in the western part of India is studied. The site displays an altitude of 1034m above mean sea level. The dominant winds are the west wind blowing from the sea towards the land and the east wind blowing from the land towards the sea. The thermal effects of breezes on the coast are frequent. Three meteorological stations and a measurement mast were chosen to analyze wind measurements. The results show that the mesoscale-microscale coupling method improves the wind field. The wind speed ratios between the meteorological station and the mast with the meso-micro coupling method performs better than the microscale or the mesoscale modellings.
For a wind turbine, there is a corresponding curve to represent the relationship between the turbine output power and the wind speed (this relationship is similar to an S-shaped curve when the output power does not reach the rated power), which is called turbine power curve. For a wind farm, the relationship between the wind farm output power and the wind speed is also similar to an S-shaped curve.
In this study, the wind farm output power is calculated by the sum of the output power of all turbines in the wind farm and the wind speed of the wind farm is defined as the wind speed measured by the mast at the wind farm. This curve, which called the wind farm power curve, can be obtained by fitting the historical whole farm power and wind speed data.
Wind farm power curves have many applications. For example, the evolution of wind farm power generation performance can be analyzed by comparing the wind farm power curves over the years, so that the future power generation plan can be adjusted in time. Based on long-term wind speed forecast data, it is possible to predict future wind farm power forecast time series data by using the wind farm power curve and get the general trend of power generation in the future.
During the fitting of wind farm power curve, points that deviate from the trend of wind farm power curve may appear on the figure of power vs. wind speed due to different reasons such as turbines shutdown, power curtailment or data acquisition error in the wind farm, which brings many errors when fitting power curve. Of course, outliers can be filtered out manually, but if the data length is very long, it must be a very time-costly work. In the power vs. wind speed figure, it is assumed that the points in each power bin follow a mixture of Gaussian distributions, including outliers. The Gaussian mixture model can help to find the mean value (2-dimensions) of every distribution in each power bin. We use these mean values to represent the trend of wind farm power curve, since it reduces the influence of outliers on fitting wind farm power curve.
Finally fitting wind farm power curve using these mean values with 3-parameters S-curve (these 3 parameters are the main parameters for adjusting the S-shaped curve). The advantage of this method is that it can give a credible wind farm power curve without manually removing outlier data. Based on a dataset of wind power output of 10 wind farms, the performance of this method in analyzing the power generation yearly performance of wind farms and forecasting wind farm power can be tested.
The remote sensing devices, especially Doppler LiDAR, are widely used in wind resource assessment as supplementary measurement technique to mast-mounted cup anemometers. The LiDAR helps to reduce the wind measurement incertitude at hub height with low cost and high mobility. But the LiDAR measurement uncertainty in complex terrains is known. The Doppler LiDAR directly measures the wind speed of sent lasers into the atmosphere, called the radial velocity. Then, a flow curvature reconstruction process is applied to convert the radial velocities to the wind flow speed. This reconstruction process assumes a homogeneous wind flow or uses a simple wind flow model to estimate the flow variation, which challenges the performance of LiDAR in complex terrains.
Using Computational Fluid Dynamics (CFD)-based method to improve the LiDAR performance has been proposed. Meteodyn WT, a CFD-based wind simulation software, models the wind flow in complex terrains and provides the detailed wind flow characters, the inflow angles and wind deviation. A correction of raw LiDAR measurement technique taking into account the LiDAR geometry, plus the simulated inflow angles and wind deviations at LiDAR beam heights and in the directions of sent laser, is developed by Meteodyn to convert the raw LiDAR measurement to the wind flow speed.
Molas B300, a Doppler LiDAR manufactured by Movelaser is installed on the ridge of a hill, close to a met mast during a wind measurement campaign. The site displays an altitude of 220m above sea level. Considering the roughness and atmospheric stability, the LiDAR corrections from Meteodyn WT with 36 wind direction sectors being every 10-degree step, are applied to the LiDAR wind speed data to obtain the corrected wind speed series. The corrected LiDAR wind speed data is compared with the met mast wind speed data served as reference. The results show that the corrected LiDAR wind speed data allows to reduce the measurement errors less than 2% for 90% of the sectors, even less than 1% for 50% of the sectors. It can be concluded that the LiDAR correction is a promising solution to improve the LiDAR performance in complex terrain.
Discover our news
Homepage // Mise en avant actualités [EN]