Forecasting ramps of wind power production with numerical weather prediction ensembles

Today, there is a growing interest in developing short‐term wind power forecasting tools able to provide reliable information about particular, so‐called ‘extreme’ situations. One of them is the large and sharp variation of the production a wind farm can experience within a few hours called ramp event. Developing forecast information specially dedicated to ramps is of primary interest because of both the difficulties that usual models have to predict and the potential risk they represent in the management of a power system. First, we propose a methodology to characterize ramps of wind power production with a derivative filtering approach derived from the edge detection literature. Then we investigate the skill of numerical weather prediction ensembles to make probabilistic forecasts of ramp occurrence. Through conditioning probability forecasts of ramp occurrence to the number of ensemble members forecasting a ramp in time intervals, we show how ensembles can be used to provide reliable forecasts of ramps with sharpness. Our study relies on 18 months of wind power measures from an 8 MW wind farm located in France and forecasts ensemble of 51 members from the Ensemble Prediction System of the European Center for Medium‐Range Weather Forecasts. Copyright © 2012 John Wiley & Sons, Ltd.     

Detecting,categorizing and forecasting large ramps in wind farm power output using meteorological observations and WPPT

The Wind Power Prediction Tool (WPPT) has been installed in Australia for the first time, to forecast the power output from the 65MW Roaring 40s Renewable Energy P/L Woolnorth Bluff Point wind farm. This article analyses the general performance of WPPT as well as its performance during large ramps (swings) in power output. In addition to this, detected large ramps are studied in detail and categorized. WPPT combines wind speed and direction forecasts from the Australian Bureau of Meteorology regional numerical weather prediction model, MesoLAPS, with real‐time wind power observations to make hourly forecasts of the wind farm power output. The general performances of MesoLAPS and WPPT are evaluated over 1 year using the root mean square error (RMSE). The errors are significantly lower than for basic benchmark forecasts but higher than for many other WPPT installations, where the site conditions are not as complicated as Woolnorth Bluff Point. Large ramps are considered critical events for a wind power forecast for energy trading as well as managing power system security. A methodology is developed to detect large ramp events in the wind farm power data. Forty‐one large ramp events are detected over 1 year and these are categorized according to their predictability by MesoLAPS, the mechanical behaviour of the wind turbine, the power change observed on the grid and the source weather event. During these events, MesoLAPS and WPPT are found to give an RMSE only roughly equivalent to just predicting the mean (climatology forecast). Copyright © 2007 John Wiley & Sons, Ltd.     

An edge model for the evaluation of wind power ramps characterization approaches

Wind power forecasting is a recognized means of facilitating large‐scale wind power integration into power systems. Recently, there has been focus on developing dedicated short‐term forecasting approaches for large and sharp wind power variations, so‐called ramps. Accurate forecasts of specific ramp characteristics (e.g., timing and probability of occurrence) are important, as related forecast errors may lead to potentially large power imbalances, with a high impact on the power system. Various works about ramps’ periodicity or predictability have led to the development of new characterization approaches. However, a thorough analysis of these approaches has not yet been carried out. Such an analysis is necessary to ensure the reliability of subsequent conclusions on ramps’ characteristics. In this paper, we propose a comprehensive framework for evaluating and comparing different characterization approaches of wind power ramps. As a first step, we introduce a theoretical model of a ramp inspired from edge‐detection literature. The proposed model incorporates some important aspects of the wind power production process so as to reflect its non‐stationary and bounded aspects, as well as the random nature of ramp occurrences. Then, we introduce adequate evaluation criteria from signal‐processing and statistical literature, in order to assess the ability of an approach for reliably estimating ramp characteristics (i.e., timing and intensity). On the basis of simulations from this model and using the evaluation criteria, we study the performance of different ramp detection filters and multi‐scale characterization approaches. Our results show that some practical choices in wind‐energy literature are inappropriate, while others, namely, from signal‐processing literature, are preferable. Copyright © 2014 John Wiley & Sons, Ltd.     

Dual‐Doppler measurements of a wind ramp event at an Oklahoma wind plant

An observational case study of a wind ramp event at Enel Green Power North America's wind plant in Oklahoma is presented. Using coordinated measurements collected by the Texas Tech University Ka‐band radars, dual‐Doppler‐synthesized wind fields are merged with data from a meteorological tower and 32 operational turbines to document the evolution and impact of the wind ramp on turbine behavior and performance over a 1 h period. During the event, average power output for turbines within the dual‐Doppler analysis domain increases from 18.3% of capacity to 98.9% of capacity, emphasizing the abrupt impact wind ramp events can have on the electrical grid. The presented measurements and analyses highlight the insights remote sensing technologies can offer towards documenting transient wind ramps and assisting modeling efforts used to forecast such events. Copyright © 2015 John Wiley & Sons, Ltd.     

Analysing wind power ramp events and improving very short-term wind power predictions by including wind speed observations

Though wind power predictions have been consistently improved in the last decade, persistent reasons for remaining uncertainties are sudden large changes in wind speed, so-called ramps. Here, we analyse the occurrence of ramp events in a wind farm in Eastern Germany and the performance of a wind power prediction tool in forecasting these events for forecasting horizons of 15 and 30 min. Results on the seasonality of ramp events and their diurnal cycle are presented for multiple ramp definition thresholds. Ramps were found to be most frequent in March and April and least frequent in November and December. For the analysis, the wind power prediction tool is fed by different wind velocity forecast products, for example, numerical weather prediction (NWP) model and measurement data. It is shown that including observational wind speed data for very short-term wind power forecasts improves the performance of the power prediction tool compared to the NWP reference, both in terms of ramp detection and in decreasing the mean absolute error between predicted and generated wind power. This improvement is enhanced during ramp events, highlighting the importance of wind observations for very short-term wind power prediction.     

Measuring the impact of additional instrumentation on the skill of numerical weather prediction models at forecasting wind ramp events during the first Wind Forecast Improvement Project (WFIP)

The first Wind Forecast Improvement Project (WFIP) was a DOE and NOAA‐funded 2‐year‐long observational, data assimilation, and modeling study with a 1‐year‐long field campaign aimed at demonstrating improvements in the accuracy of wind forecasts generated by the assimilation of additional observations for wind energy applications. In this paper, we present the results of applying a Ramp Tool and Metric (RT&M), developed during WFIP, to measure the skill of the 13‐km grid spacing National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESRL) Rapid Refresh (RAP) model at forecasting wind ramp events. To measure the impact on model skill generated by the additional observations, controlled data‐denial RAP simulations were run for six separate 7 to 12‐day periods (for a total of 55 days) over different seasons. The RT&M identifies ramp events in the time series of observed and forecast power, matches in time each forecast ramp event with the most appropriate observed ramp event, and computes the skill score of the forecast model penalizing both timing and amplitude errors. Because no unique definition of a ramp event exists (in terms of a single threshold of change in power over a single time duration), the RT&M computes integrated skill over a range of power change (Δp) and time period (Δt) values. A statistically significant improvement of the ramp event forecast skill is found through the assimilation of the special WFIP data in two different study areas, and variations in model skill between up‐ramp versus down‐ramp events are found.     

Production of the Finnish Wind Atlas

The Finnish Wind Atlas was prepared applying the mesoscale model AROME with 2.5 km horizontal resolution and the diagnostic downscaling method Wind Atlas Analysis and Application Programme (WAsP) with 250 m resolution. The latter was applied for areas most favourable for wind power production: a 30 km wide coastal/offshore zone, highlands, large lakes and large fields. The methodology included several novel aspects: (i) a climatologically representative period of real 48 months during 1989–2007 was simulated with the mesoscale model; (ii) in addition, the windiest and calmest months were simulated; (iii) the results were calculated separately for each month and for sectors 30° wide; (iv) the WAsP calculations were based on the mesoscale model outputs; (v) in addition to point measurements, also radar wind data were applied for the validation of the mesoscale model results; (vi) the parameterization method for gust factor was extended to be applicable at higher altitudes; and (vii) the dissemination of the Wind Atlas was based on new technical solutions. The AROME results were calculated for the heights of 50, 75, 100, 125, 150, 200, 300 and 400 m, and the WAsP results for the heights of 50, 75, 100, 125 and 150 m. In addition to the wind speed, the results included the values of the Weibull distribution parameters, the gust factor, wind power content and the potential power production, which was calculated for three turbine sizes. The Wind Atlas data are available for each grid point and can be downloaded free of charge from dynamic maps at www.windatlas.fi . Copyright © 2011 John Wiley & Sons, Ltd.     

Prediction of local wind climatology from Met Office models: Virtual Met Mast techniques

The Met Office has developed the Virtual Met Mast? (VMM) tool for assessing the feasibility of potential wind farm sites. It provides site‐specific climatological wind information for both onshore and offshore locations. The VMM relies on existing data from past forecasts from regional‐scale numerical weather prediction (NWP) models, to which corrections are applied to account for local site complexity. The techniques include corrections to account for the enhanced roughness lengths used in NWP models to represent drag due to sub‐grid orography and downscaling methods that predict local wind acceleration over small‐scale terrain. The corrected NWP data are extended to cover long periods (decades) using a technique in which the data are related to alternative long‐term datasets. For locations in the UK, the VMM currently relies on operational mesoscale model forecast data at 4 km horizontal resolution. Predictions have been verified against observations made at typical wind turbine hub heights at over 80 sites across the UK. In general, the predictions compare well with the observations. The techniques provide an efficient method for screening potential wind resource sites. Examples of how the VMM techniques can be used to produce local wind maps are also presented. © 2016 Crown copyright. Wind Energy © 2016 John Wiley & Sons, Ltd     

Leveraging stochastic differential equations for probabilistic forecasting of wind power using a dynamic power curve

Short‐term (hours to days) probabilistic forecasts of wind power generation provide useful information about the associated uncertainty of these forecasts. Standard probabilistic forecasts are usually issued on a per‐horizon‐basis, meaning that they lack information about the development of the uncertainty over time or the inter‐temporal correlation of forecast errors for different horizons. This information is very important for forecast end‐users optimizing time‐dependent variables or dealing with multi‐period decision‐making problems, such as the management and operation of power systems with a high penetration of renewable generation. This paper provides input to these problems by proposing a model based on stochastic differential equations that allows generating predictive densities as well as scenarios for wind power. We build upon a probabilistic model for wind speed and introduce a dynamic power curve. The model thus decomposes the dynamics of wind power prediction errors into wind speed forecast errors and errors related to the conversion from wind speed to wind power. We test the proposed model on an out‐of‐sample period of 1year for a wind farm with a rated capacity of 21MW. The model outperforms simple as well as advanced benchmarks on horizons ranging from 1 to 24h. Copyright © 2016 John Wiley & Sons, Ltd.     

Regime‐switching models and multiple causal factors in forecasting wind speed

This study evaluates two types of models for wind speed forecasting. The first is models with multiple causal factors, such as offsite readings of wind speed and meteorological variables. These can be estimated using either regressions or neural networks. The second is state transition and the closely related class of regime‐switching transition models. These are attractive in that they can be used to predict outlying fluctuations or large ramp events. The regime‐switching model uses a persistence forecast during periods of high wind speed, and regressions for low and intermediate speeds. These techniques are tested on three databases. Two main criteria are used to evaluate the outcomes, the number of high and low states than can be predicted correctly and the mean absolute percent error of the forecast. Neural nets are found to predict the state transitions somewhat better than logistic regressions, although the regressions do not do badly. Three methods all achieve about the same degree of forecast accuracy: multivariate regressions, state transition and regime‐switching models. If the states could be predicted perfectly, the regime‐switching model would improve forecast accuracy by an additional 2.5 to 3 percentage points. Analysis of the density functions of wind speed and the forecasting models finds that the regime‐switching method more closely approximates the distribution of the actual data. Copyright © 2009 John Wiley & Sons, Ltd.     

Using temperature and state transitions to forecast wind speed

A major issue in forecasting wind speed is non‐linear variability. The probability distribution of wind speed series shows heavy tails, while there are frequent state transitions, in which wind speed changes by large magnitudes, over relatively short time periods. These so‐called large ramp events are one of the critical issues currently facing the wind energy community. Two forecasting algorithms are analyzed here. The first is a regression on lags, including temperature as a causal factor, with time‐varying parameters. The second augments the first using state transition terms. The main innovation in state transition models is that the cumulative density function from regressions on the states is used as a right‐hand side variable in the regressions for wind speed. These two methods are tested against a persistence forecast and several non‐linear models, using eight hourly wind speed series. On average, these two models produce the best results. The state transition model improves slightly over the regression. However, the improvement achieved by both models relative to the persistence forecast is fairly small. These results argue that there are limits to the accuracy that can be achieved in forecasting wind speed data. Copyright © 2007 John Wiley & Sons, Ltd.     

Adaptive post‐processing of short‐term wind forecasts for energy applications

We present a new method of reducing the error in predicted wind speed, thus enabling better management of wind energy facilities. A numerical weather prediction model, COSMO, was used to produce 48 h forecast data every day in 2008 at horizontal resolutions of 10 and 3 km. A new adaptive statistical method was applied to the model output to improve the forecast skill. The method applied corrective weights to a set of forecasts generated using several post‐processing methods. The weights were calculated based on the recent skill of the different forecasts. The resulting forecast data were compared with observed data, and skill scores were calculated to allow comparison between different post‐processing methods. The total root mean square error performance of the composite forecast is superior to that of any of the individual methods. Copyright © 2010 John Wiley & Sons, Ltd.     

Estimating the offshore wind resources of the State of Ceará in Brazil

The onshore wind power is consolidated; the challenge is to reach the same level of maturity for offshore exploitation. Brazil has no offshore wind power plants and there are few studies in this direction. This paper aims to estimate the offshore wind resources in the State of Ceará, in Brazil. The investigation uses a mesoscale atmospheric computer model, the Regional Atmospheric Modeling System (RAMS), with horizontal resolution of 2 km, which estimates the offshore average wind speed, average wind direction, power density and turbulence taking into account the bathymetry data and navigation routes along the coast of Ceará. The wind potential was evaluated in three representative periods, La Niña, El Niño and Neutral year, analyzing the dry and rainy season for each period. The results indicate an average wind speed above 8 m/s and power density above 720 W/m² no matter the period evaluated, in the dry season. The predominant wind direction in the observed dry periods was from East to West and the turbulence intensity is smaller during dry season of El Niño. Besides, the bathymetry of the State of Ceará is shallow and the large ships route is far beyond the coast, offering no danger to future endeavors.     

Realistic simulations of the July 1, 2011 severe wind event over the Buffalo Ridge Wind Farm

Severe winds from thunderstorm outflows pose a challenge to wind turbine arrays. They can cause significant power ramps and disruption in energy production. They can also cause extreme structural damage to turbines as was seen in the severe storm event over the Buffalo Ridge Wind Farm on July 1, 2011. At this southwestern Minnesota site, blades from multiple turbines broke away and a tower buckled in the intense winds. In this study, we attempt to characterize meteorological conditions over the Buffalo Ridge Wind Farm area during this event. The observational network included NEXRAD radars, automated surface observation stations and a wind profiler. Storm reports from the Storm Prediction Center and damage surveys provided additional insight to the in situ measurements. Even with these datasets, assessing wind speeds around turbine rotors is difficult. Thus, Weather Research and Forecasting model simulations of the event are carried out that consider current and anticipated future operational model setups. This work addresses model spatial resolution versus parameterization complexity. Parameterizations of the planetary boundary layer and microphysics processes are evaluated based on their impact on storm dynamics and the low‐level wind field. Results are also compared with the Wind Integration National Dataset, which utilizes data assimilation and an extensive continental domain. Enhanced horizontal resolution with simplistic parameterization helps increase resolved wind speeds and ramp intensity. Enhanced sophistication of microphysics parameterizations also helps increase resolved wind speeds, improve storm timing and structure and resolve higher values of turbulent kinetic energy in the lowest 1 km of the atmosphere. Copyright © 2017 John Wiley & Sons, Ltd.     

A wind field downscaling strategy based on domain segmentation and transfer functions

This paper presents a novel methodology for mesoscale‐to‐microscale downscaling of near‐surface wind fields. The model chain consists on the Weather Research and Forecast mesoscale model and the Alya‐CFDWind microscale model (assuming neutral stability). The downscaling methodology combines precomputed microscale simulations with a mesoscale forecast using a domain segmentation technique and transfer functions. As a result, the downscaled wind field preserves the mesoscale pattern but, at the same time, incorporates local mesoscale subgrid terrain effects, particularly at valleys and channelling zones. The methodology has been validated over a 9‐month period on a very complex terrain site instrumented with a dense observational network of meteorological masts. With respect to mesoscale results, the global skills of the downscaled wind at masts improve for wind direction and remain similar for wind velocity. However, a substantial improvement occurs under stable and neutral conditions and for high wind velocity regimes.     

Influence of neural network training parameters on short-term wind forecasting

This paper investigates factors which can affect the accuracy of short-term wind speed prediction when done over long periods spanning different seasons. Two types of neural networks (NNs) are used to forecast power generated via specific horizontal axis wind turbines. Meteorological data used are for a specific Western Australian location. Results reveal that seasonal variations affect the prediction accuracy of the wind resource, but the magnitude of this influence strongly depends on the details of the NN deployed. Factors investigated include the span of the time series needed to initially train the networks, the temporal resolution of these data, the length of training pattern within the overall span which is used to implement the predictions and whether the inclusion of solar irradiance data can appreciably affect wind speed prediction accuracy. There appears to be a relatively complex relationship between these factors and the accuracy of wind speed prediction via NNs. Predicting wind speed based on NNs trained using wind speed and solar irradiance data also increases the prediction accuracy of wind power generated, as can the type of network selected.     

Characteristics of day‐ahead wind power forecast errors in Nordic countries and benefits of aggregation

The growing proportion of wind power in the Nordic power system increases day‐ahead forecasting errors, which have a link to the rising need for balancing power. However, having a large interconnected synchronous power system has its benefits, because it enables to aggregate imbalances from large geographical areas. In this paper, day‐ahead forecast errors from four Nordic countries and the impacts of wind power plant dispersion on forecast errors in areas of different sizes are studied. The forecast accuracy in different regions depends on the amount of the total wind power capacity in the region, how dispersed the capacity is and the forecast model applied. Further, there is a saturation effect involved, after which the reduction in the relative forecast error is not very large anymore. The correlations of day‐ahead forecast errors between areas decline rapidly when the distance increases. All error statistics show a strong decreasing trend up to the area sizes of 50,000 km². The average mean absolute error (MAE) in different regions is 5.7% of installed capacity. However, MAE of a smaller area can be over 8% of the capacity, but when all the Nordic regions are aggregated together, the capacity‐normalized MAE decreases to 2.5%. The average of the largest errors for different regions is 39.8% and when looking at the largest forecast errors for smaller areas, the largest errors can exceed 80% of the installed capacity, whereas at the Nordic level, the maximum forecast error is only 13.5% of the installed capacity. Copyright © 2016 John Wiley & Sons, Ltd.     

A surface-layer wind speed correction: A case-study of Darling station

In previous study, the vertical wind speed extrapolation from measurement station to modern turbine hubs over an open homogenous terrain was considered. It was presented that an assumption of wind shear exponent under different stability conditions was an inaccurate representation of the actual wind climates as the precise knowledge of the site's wind characteristics at different levels and seasons are essential for planning and implementation of a proposed energy project. In this study, the surface-layer wind speed correction at Darling using the WRF modeling with mesoscale terrain corrections is presented. An hourly mesoscale modeled winds at 3 km grid spacing obtained for one month are postprocessed for estimation of local wind speed profiles at 10 and 50 m height AGL. The sensitivity of the modeled winds to surface terrain corrections is investigated using mesoscale topography parameterizations. Furthermore, 6-hourly mesoscale modeled and satellite observed winds as well as measurements from Darling station are utilized for validation of the statistical downscaling method utilized for the postprocessing of the boundary layer winds over land. It is presented that the precision of the mesoscale modeled winds for local wind speed estimates at potential site without historical measurements can be significantly improved. The confidence in the validity of this methodology for local wind speed correction is estimated at 96–98%.     

On the tails of the wind ramp distributions

We analyzed several multiyear wind speed datasets from 4 different geographical locations. The probability density functions of wind ramps from all these sites revealed remarkably similar shapes. The tails of the probability density functions are much heavier than a Gaussian distribution, and they also systematically depend on time increments. Quite interestingly, from a purely statistical standpoint, the characteristics of the extreme ramp‐up and ramp‐down events are found to be almost identical. With the aid of extreme value theory, we describe several other inherent features of extreme wind ramps in this paper.