Dissertations / Theses on the topic 'Wind Speed Estimation'

This thesis includes the results of a PhD study about methods to compare Sonic Detection And Ranging (SODAR) measurements to measurements from other instruments. The study focuses on theoretical analysis, the design of a transponder system for simulating winds and the measurement of the acoustic radiation patterns of SODARs. These methods are integrated to reduce uncertainty in SODAR measurements. Through theoretical analysis it is shown that the effective measurement volume of a range gate is 15% of a cone section based on the SODAR's Full Width Half Maximum (FWHM). Models of the beam pattern are used to calculate the ratio of air passing a turbine to that measured by a SODAR over 10 minutes with values of 3-5% found at 10ms-1. The model is used to find angles where significant Sound Pressure Levels (SPLs) occur close to a SODARs baffle giving the highest chance of fixed echoes. This is converted into an orientation guide for SODAR set-up. The design of a transponder system is detailed that aims to provide a calibration test of the processing applied by a SODAR. Testing has shown that the transponder can determine the Doppler shift equation used by a SODAR although further work is needed to make the system applicable to all SODARs. It is shown that anechoic measurements of single elements are useful for improving array models. Measurements of the FWHM and acoustic tilt angle can be achieved in the field using a tilt mechanism and a Sound Level Meter (SLM) on a 10m mast. The same mechanism can be used to calculate an effective tilt angle using the Bradley technique. It is proposed that these methods are integrated to calculate error slopes for the SODAR measurement with regards to a secondary location. It is shown that the slopes could be between 0 and 5% if the methods are fully realised and a Computational Fluid Dynamics (CFD) model is incorporated.

Wind turbines typically rely on feedback controllers to maximize power capture in below-rated conditions and regulate rotor speed during above-rated operation. However, measurements of the approaching wind provided by Light Detection and Ranging (lidar) can be used as part of a preview-based, or feedforward, control system in order to improve rotor speed regulation and reduce structural loads. But the effectiveness of preview-based control depends on how accurately lidar can measure the wind that will interact with the turbine.

In this thesis, lidar measurement error is determined using a statistical frequency-domain wind field model including wind evolution, or the change in turbulent wind speeds between the time they are measured and when they reach the turbine. Parameters of the National Renewable Energy Laboratory (NREL) 5-MW reference turbine model are used to determine measurement error for a hub-mounted circularly-scanning lidar scenario, based on commercially-available technology, designed to estimate rotor effective uniform and shear wind speed components. By combining the wind field model, lidar model, and turbine parameters, the optimal lidar scan radius and preview distance that yield the minimum mean square measurement error, as well as the resulting minimum achievable error, are found for a variety of wind conditions. With optimized scan scenarios, it is found that relatively low measurement error can be achieved, but the attainable measurement error largely depends on the wind conditions. In addition, the impact of the induction zone, the region upstream of the turbine where the approaching wind speeds are reduced, as well as turbine yaw error on measurement quality is analyzed.

In order to minimize the mean square measurement error, an optimal measurement prefilter is employed, which depends on statistics of the correlation between the preview measurements and the wind that interacts with the turbine. However, because the wind speeds encountered by the turbine are unknown, a Kalman filter-based wind speed estimator is developed that relies on turbine sensor outputs. Using simulated lidar measurements in conjunction with wind speed estimator outputs based on aeroelastic simulations of the NREL 5-MW turbine model, it is shown how the optimal prefilter can adapt to varying degrees of measurement quality.

Cette thèse propose des algorithmes pour estimer la vitesse et la direction du vent pour des éoliennes et des parcs éoliens.Tout d'abord, nous proposons des méthodes basées sur les données pour estimer la vitesse effective du rotor (REWS) sans nécessiter la connaissance de certains paramètres physiques de l'éolienne, qui pourraient être inconnus de l'opérateur. Nous fournissons deux méthodes basées sur les données, l'une basée sur la régression par processus gaussien et l'autre combinant la régression par processus gaussien avec un observateur grand gain.Ensuite, en nous basant sur cette estimation locale de la REWS, au niveau d'une éolinenne, nous abordons la question de l'estimation du vent en écoulement libre au niveau du parc éolien.Nous commençons par nous concentrer sur l'estimation de la vitesse du vent, pour une direction du vent connue. Pour un parc éolien de géométrie simple, nous démontrons qu'une mesure locale de la vitesse perturbée par la présence des éoliennes peut être utilisée pour estimer la vitesse du vent en écoulement libre. Nous fondons notre méthodologie d'estimation sur une modélisation simplifiée de l'effet de sillage qui consiste en des équations aux dérivées partielles hyperboliques du premier ordre en cascade, et dont la vitesse de transport est la vitesse du vent en écoulement libre. Nous proposons d'utiliser une solution analytique de ces équations, impliquant des retards de transport, pour effectuer une estimation de la mesure locale et mettre à jour l'estimation de la vitesse du vent en écoulement libre. Nous démontrons formellement la convergence de cette estimation et illustrons numériquement l'efficacité de cette méthode.Enfin, nous passons à une configuration plus générale où à la fois la vitesse et la direction du vent en écoulement libre sont inconnues. Nous proposons d'utiliser une modélisation bidimensionelle du sillage et de nous appuyer sur une méthode basée sur l'optimisation. Le problème d'identification que nous formulons se révèle être particulièrement difficile en raison de l'apparition de retards de transport, mais nous montrons comment contourner cette difficulté en considérant une valeur moyenne de l'historique de la vitesse du vent en écoulement libre. Des résultats de simulation obtenus avec le simulateur FAST.Farm illustrent l'intérêt de la méthode proposée
This thesis designs algorithms to estimate the wind speed and direction for wind turbines and wind farms.First, we propose data-based methods to estimate the Rotor Effective Wind Speed (REWS) for a single turbine without prior knowledge of certain physical parameters of the turbine that might be unknown to an operator.We provide two data-based methods, based respectively on Gaussian Process Regression (GPR) and on an combination of GPR with high-gain observers.Second, grounding on this REWS estimation at the local level of one turbine, we address the question of estimating the free-flow wind at the level of a wind farm.We start by focusing on wind speed estimation, for a given known wind direction. For a wind farm with a simple geometry, we prove that a local speed measurement disturbed by the presence of the turbines can be used to estimate the free-flow wind speed. We ground our estimation methodology on a simplified wake model, which consists of first-order hyperbolic partial differential equations, the transport speed of which is the free-flow wind speed. We propose to use an analytical solution of these equations, involving transport delays, to perform an estimate of the local measurement and to update the free-flow wind speed estimate. We formally prove the convergence of this estimate and numerically illustrate the efficiency of this method.Finally, we move to a more general setup where both the free-flow wind speed and direction are unknown. We propose to use a two-dimensional wake model and to rely on an optimization-based method. This identification problem reveals to be particularly challenging due to the appearance of transport delays, but we illustrate how to circumvent this issue by considering an average value of the free flow wind speed history. Simulation results obtained with the simulator FAST.Farm illustrate the interest of the proposed method

The Weibull distribution, an extreme value distribution, is frequently used to model survival, reliability, wind speed, and other data. One reason for this is its flexibility; it can mimic various distributions like the exponential or normal. The two-parameter Weibull has a shape (γ) and scale (β) parameter. Parameter estimation has been an ongoing search to find efficient, unbiased, and minimal variance estimators. Through data analysis and simulation studies, the following three methods of estimation will be discussed and compared: maximum likelihood estimation (MLE), method of moments estimation (MME), and median rank regression (MRR). The analysis of wind speed data from the TW Daniels Experimental Forest are used for this study to test the performance and flexibility of the Weibull distribution.

Impulsionado pela mecânica de leilões de energia, o aproveitamento energético de recursos eólicos no Brasil atravessa um momento de expansão em participação na matriz de energia elétrica nacional. Não obstante, o desempenho da geração dos parques eólicos que estão em operação foi monitorado e apresentou, em média, resultados aquém daquilo que fora confiado ao Sistema Interligado Nacional, revelando que as estimativas de geração projetadas e declaradas por alguns dos projetos vencedores dos processos licitatórios podem ter sido supervalorizadas. Tal cenário provocou a exigência de medidas mais conservadoras para participação nos leilões de energia, como a já vigente adoção do P90 no cálculo da Garantia Física e o aumento da duração da campanha de medição anemométrica, a entrar em rigor a partir de 2017. Sendo o vento uma variável estocástica, existem incertezas intrínsecas à Avaliação de Recursos Eólicos que influenciam no processo de estimação da geração por um parque eólico e que devem, desta forma, ser identificadas, quantificadas e reduzidas, na medida do possível. Nesse sentido, este trabalho estuda a influência da duração da campanha de medição anemométrica na Avaliação de Recursos Eólicos com base na aplicação do método MCP ferramenta imprescindível no processo de caracterização do regime eólico no longo prazo com vistas para aprimorar a exatidão das previsões de geração pela fonte eólica. Para tanto, foram utilizadas quatro bases de dados contendo séries temporais de velocidade e direção do vento referentes a uma região de interesse. Inicialmente, nove diferentes métodos MCP foram testados e comparados, sendo que o método Vertical Slice aplicado com auxílio do software Windographer destacou-se dos demais e mostrou-se mais aderente aos dados utilizados conforme as métricas de Erro Absoluto Médio e Raiz Quadrada do Erro Quadrático Médio. Posteriormente, as bases de dados foram configuradas para simular campanhas de medição anemométricas com durações que variavam de 2 a 6 anos, de modo a avaliar o comportamento da incerteza relativa à caracterização histórica de recursos eólicos e analisar em que medida esta incerteza impacta no cálculo da estimativa de geração de eletricidade por um conjunto de aerogeradores hipoteticamente dispostos naquele local de interesse. Foi possível verificar que, para os dados e casos analisados, à medida que se aumentou a duração da campanha de medição anemométrica, a incerteza da caracterização histórica de recursos eólicos sofreu queda significativa; determinando, por conseguinte, redução da incerteza total que permeia a geração eólica. Ademais, a quantidade de energia estimada para o parque eólico hipotético exemplificado também decresceu, permitindo melhora na acurácia da previsão de geração e beneficiando a confiabilidade da fonte eólica no sistema elétrico brasileiro.
Driven by the energy auctions system, the energetic harnessing of wind resource in Brazil is now going through a phase of expansion in participation in the national electric energy mix. Nevertheless, the performance of power generation of in-operation wind farms was monitored and the results proved to be, on average, below what was initially entrusted to the National Grid System, indicating that the energy production estimations projected by some energy auctions winners could have been overestimated. This scenario has caused the requirements for participating in the energy auctions to be more conservative, with measures such as the adoption of the P90 on the calculation of the physical guarantee and the increase of the wind measurement campaigns time span the latter to be enforced as of 2017. The wind is a stochastic resource, hence there are uncertainties intrinsic to the Wind Resource Assessment that influence a wind farms power generation estimation and that need to be properly identified, quantified and reduced, as far as possible. In this respect, the influence of a wind measurement campaigns time span on the Wind Resource Assessment based on MCP methods an important tool in the process of characterizing the long-term wind regime was studied in order to detect the potential of enhancing the accuracy of wind power generation forecasts. For this purpose, four databases containing time series of wind speed and direction belonging to a target site were used. Firstly, nine different MCP methods were tested and compared, of which the Vertical Slice method implemented on the software Windographer outperformed all the others according to the Mean Absolute Error and Root Mean Square Error metrics. Subsequently, the databases were set to simulate campaigns with time spans varying from 2 to 6 years, in such a way to evaluate the behavior of the uncertainty in the long-term wind speed and to analyze how this uncertainty impacts the calculation of the energy production estimation of an array of wind turbines hypothetically placed on that target site. From the analyzed data and cases, it was verified that, as the wind measurement campaigns time span was increased, the uncertainty in the long-term wind speed was significantly diminished, thereby reducing the overall uncertainty that pervades the wind power harnessing. Furthermore, the energy production estimation of the exemplified hypothetical wind farm also decreased, allowing an improvement on the accuracy of the energy generation prediction and benefiting the reliability of wind power in the Brazilian electric system.

In electrical power system, prediction of Renewable energy sources has become essential for designing a control strategy to manage the electricity on the grid. To help the system operators for integration of wind power system to the existing power system, wind speed and power prediction is essential.Basically neural network is aimed for short-term forecasting problems as it is capable to learn non-linear relationship between inputs and outputs by a non-statistical approach and don’t require any predefined mathematical model. This thesis investigates the effectiveness of recurrent wavelet neural network (RWNN) and artificial wavelet neural network (AWNN) dynamics for wind speed forecasting. We evaluate the RWNN and AWNN against multilayer feed-forward neural network. The RWNN and AWNN are trained using back propagation gradient descent algorithm. The experimental results show that the performance of RWNN and AWNN approaches outperforms the multilayer feed-forward neural network. All the three models use Hourly averaged time series data (2982 numbers of samples) for wind speed collected from the National Renewable Energy Laboratory (NREL).

碩士
國立高雄科技大學
電腦與通訊工程系
107
Wind speed measurements are used in many environments, such as highways, MRT elevated roads, and wind turbines. Excessive wind power may cause disasters, so detecting wind speed is one of the most important factors for safety. At present, the general wind speed warning system uses the average wind speed or the instantaneous wind speed as the decision-making basis for issuing warnings. It may cause the measurement data to be averaged, so that the average wind speed cannot reflect the current real situation, and the instantaneous wind speed may float near the warning threshold without correctly triggering the warning. Therefore, this thesis proposes a wind quality decision system, using Kalman filter as the wind parameter estimator, and applying new wind parameters to the wind system model for parameter estimation. The results show that the wind quality decision system can provide earlier decisions than the previous system model, and the probability of correctly triggering the warning increases as the instantaneous wind speed floats near the wind speed warning threshold. Therefore, the overall performance may be improved by using the proposed decision system for determining the wind quality.

碩士
國立高雄第一科技大學
電腦與通訊工程系碩士班
106
Wind speed measuring are used in many environments where winds are too large to cause disasters, so detecting wind speed is one of the most important factors for safety. Current warning system works out with an average or instant maximum wind speed as a parameter. The measurement data might be desalinated or activated by detecting maximum wind speed and lead to tragedy without making a current decision. To improve the accuracy or lower the probability of misjudgment, we use Kalman Filter to estimate wind speed and wind acceleration on its recursive estimation and correction formula. Average wind speed estimation might have credibility problem by averaging critical values. It could cause the detention and inappropriate to the warning system and make a unsuitable decision of current situation. After Kalman Filtering, wind speed and wind acceleration were used as decision parameter. We observed the untness using average or instant maximum wind speed as a decision parameter from different experiment. The wind parameter after Kalman Filtering can be suitable to different kinds of situations. The results show three different kinds of wind parameter processed by Kalman Filter and combined with sliding window signal processing giving suitable feedback to reduce the probability of misjudgment, avoiding the dangerous to drive from strong wind. By using the proposed decision assisted system, we can improve the accuracy of warning system and improve the safety factors of related applications.

碩士
南台科技大學
電機工程系
97
It is an important issue to master the characteristics of wind speed distribution exactly during estimating the wind energy. To evaluate the goodness of fit for the wind speed distribution in Taiwan，statistical inference is used to conduct four different probability density functions (pdf) of wind speed. The statistical characteristics of wind speed are analyzed from the histograms. The thesis tests the impact of variations of histogram interval number to the Root Mean Square probability error. The Monte Carlo based simulation, modified from maximum likelihood method, is proposed to conduct the best shape and scale parameters in the pdf of wind speed. The wind speed data form eight weather stations of the Central Weather Bureau during year of 2003-2008 are used to test. Results show that the Weibull distribution may be not the best pdf in the all eight testing locations in Taiwan. The Gamma pdf is more suitable to represent the characteristics of wind speed in the Chengkung & Tungchitao terrain. The most potential wind resource in Taiwan, Lanyu, Gamma or Rayleigh pdf are suitable than Weibull model. 　　The capacity & available factor are used as index to verify the most suitable wind unit in Taiwan from the 22 chosen popular wind generators. The wind speed estimated at 100 meter height in the testing locations are evaluated the capacity factor of 22 wind generators. Results show that the Vestas V90_1.8M site is the appropriate choice in the seven testing locations, expect for YungKang.

Global warming, which is one of the most serious consequence of climate change, can be expected to have different effects on the atmosphere, the ocean, icebergs, etc. Global warming has also brought secondary consequences into nature and human society directly. The most negative effect among the several effects of global warming is the rising sea level related to the large typhoons which can cause flooding on low-level land, coastal invasion, sea water flow into rivers and underground water, rising river level, and fluctuation of sea tides. It is crucial to recognize surge level and its return period more accurately to prevent loss of human life and property damage caused by typhoons. This study researches two topics. The first purpose of this study is to develop a statistical model to predict the return period of the storm surge water related to typhoon Maemi, 2003 in South Korea. To estimate the return period of the typhoon, clustered separated peaks-over-threshold simulation (CSPS) has been used and Weibull distribution is used for the peak storm surge height’s fitting. The estimated return period of typhoon Maemi’s peak total water level is 389.11 years (95% confidence interval 342.27 - 476.2 years). The second aim is related to the fragility curves with the loss data caused by typhoons. Although previous studies have developed various methods to mitigate damages from typhoons, the extent of financial loss has not been investigated enough. In this research, an insurance company provides their loss data caused by the wind speed of typhoon Maemi in 2003. The loss data is very important in evaluating the extent of the damages. In this study, the damage ratio in the loss dataset has been used as the main indicator to investigate the extent of the damages. The damage ratio is calculated by dividing the direct loss by the insured amount. In addition, this study investigates the fragility curves of properties to estimate the damage from typhoon Maemi in 2003. The damage ratios and storm induced wind speeds are used as the main factor for constructing fragility curves to predict the levels of damage of the properties. The geographical information system (GIS) has been applied to produce properties’ spatial wind speeds from the typhoon. With the damage ratios, wind speeds and GIS spatial data, this study constructs the fragility curves with four different damage levels (Level I - Level IV). The findings and results of this study can be basic new references for governments, the engineering industry, and the insurance industry to develop new polices and strategies to cope with climate change.

碩士
國立臺灣科技大學
營建工程系
92
Many typhoons attack Taiwan, in Autumn and Summer. Raids of typhoons frequently damage buildings and cause the loss of lives and fortunes. Derivation of extreme wind speeds is an important factor in Wind Resistant Design. It is very difficult to estimate extreme wind speeds accurately based on short-time wind speed data. Many statistical methods used to compute n-year return period wind speeds and n-year maximum wind speeds. This study adjusts the wind speed data of Central Weather Bureau according to the prescribed criteria. Unified wind speed data are grouped into normal-speed data and typhoon-speed data; their relevant probability distribution function are studied. Monte Carlo Simulation method, in conjunction with Latin Hypercube Sampling, are used to generate synthetic wind speeds time histories in which each sample is the respective daily maximum wind speed. Finally, we use Taipei wind speeds as example. It is found that the synthetic time series yield reasonable estimates of the year-means and year-standard of the observed series. The proposed method in used to find n-year (n=10, 50, 100, 475, 1000) return period wind speeds and probability distribution statistic of n-year maximum wind speeds.