Wind Time Series Research Articles

Optimal Long Short Term Memory Networks for long-term forecasting of real wind characteristics

The capability of wind power to meet the energy demand inspired researchers to develop techniques for harnessing this clean and renewable energy. As a primary step, accurate forecasting of wind characteristics by modelling the stochastic nature of wind is done. Although, statistical methods provide good results in forecasting, they are inferior to Deep Learning based tools while handling extreme nonlinearities in wind characteristics. The authors in this work implemented Long Short Term Memory (LSTM) networks, a deep learning based tool, for modelling wind time series data due to its efficiency in handling long term dependencies. However, several hyper-parameters like activation function and design are chosen heuristically, making the modelling process tedious and inefficient. In this study, novel multi objective optimization formulation, driven by NSGA II, is proposed to design the LSTM networks with respect to the conflicting objectives of accuracy and parsimony. The resultant optimal LSTM models are used for long term forecasting (2 years) of wind characteristics data, with an accuracy of 97%, obtained from a real wind farm in France. To demonstrate the importance of this forecasting, a study of wind power calculations on a real wind farm is conducted. For a given layout, the effect of wind frequency scenarios, generated from the time series data of wind, on the annual power calculations, is determined. The existing and forecasted values of wind speed and direction over longer periods of time resulted in realistic values of expected power from wind farm. This study demonstrates the importance of forecasting while evaluating the power which can impact research in fields such as wind farm layout optimization and control under uncertainty.

Temporal and Spatial Variations of Monsoonal Upwelling along the South West and East Coasts of India

The implicit/explicit impact of global warming or the anthropogenic climate change on the coastal upwelling processes affects the marine ecosystem significantly and continuous monitoring of the coastal upwelling along the coastal regions is essential. The Ekman transport due to alongshore winds and Ekman pumping due to wind stress curl play important roles on the coastal upwelling, especially along the southern part of Indian peninsula. Upwelling Indexes based on wind and SST are computed in the study to demarcate the upwelling features along the coasts. Since upwelling leaves its trail on surface features, long time series of satellite-derived wind, Sea Surface Temperature (SST), Sea level anomaly (SLA) and Surface Chlorophyll-a concentration (Chl-a) are analyzed to understand the upwelling along the coastal regions of south India. The present study reveals a decrease in the intensity of upwelling along the western coast (Chl-a ∼ 0.2 mg/m3 dec−1, Upwelling Index based on wind () and Vertical Velocity of Ekman Pumping (VVEP) is approximately 0.0012 m2/s dec−1 and 0.120 cm/day dec−1 respectively) and a slight increase all along the eastern coast of southern India during summer monsoon (Chl-a along 11.3°N and 13.9°N is 1 mg/m3 dec−1, and VVEP along 10.5°N is 0.0023 m2/s dec−1 and 0.2 cm/day dec−1 respectively).

New SVM kernel soft computing models for wind speed prediction in renewable energy applications

This paper proposes a hybrid multi-step wind speed prediction model based on combination of singular spectrum analysis (SSA), variational mode decomposition (VMD) and support vector machine (SVM) and was applied for sustainable renewable energy application. In the proposed SSA–VMD–SVM model, the SSA was applied to eliminate the noise and to approximate the signal with trend information; VMD was applied to decompose and to extract the features of input time series wind speed data into a number of sub-layers; and the SVM model with various kernel functions was adopted to predict the wind speed from each of the sub-layers, and the parameters of SVM were fine-tuned by differential evolutionary algorithm. To investigate the effectiveness of the proposed model, various prediction models are considered for comparative study, and it is demonstrated that the proposed model outperforms with better prediction accuracy.

COMPARATIVE ANALYSIS OF LSTM, RF AND SVM ARCHITECTURES FOR PREDICTING WIND NATURE FOR SMART CITY PLANNING

Abstract. Meteorological data and its effect have been the attention of the researchers of the smart city planning for thorough utilization and management of resources, that help in effective government management, convenient public services and sustainable industrial development. Renewable sources of energy like wind, solar, are being integrated into city planning to improve environmental quality. Wind energy is utilized through wind turbines and requires foreknowledge of wind parameters like speed and direction. The aim of this paper is to predict dominant wind speed and direction for time-series wind dataset, that can be incorporated into city planning for selecting suitable sites for wind turbines. This paper proposes three one-dimensional (1D) algorithms using Long Short Term Memory (LSTM), Random Forest (RF) and Support Vector Machine (SVM) for dominant wind speed and direction prediction. The proposed 1D LSTM (1DLSTM), RF (1DRF) and SVM (1DSVM) take successive time values in terms of wind speed and direction as input and predict the future dominant speed and direction, separately. The proposed algorithms are trained and tested using historical wind dataset of Stuttgart and Netherlands respectively. Prediction using 1DLSTM results in total accuracies reaching up to 93.9% and 94.7%, up to 92.8% and 93.8% using 1DSVM and up to 88.7% and 89.3% using 1DRF for speed and direction, respectively. Thus, prediction of wind nature using the proposed algorithms, will give city planners advanced knowledge of wind conditions.

Wind Intensity Is Key to Phytoplankton Spring Bloom Under Climate Change

The onset of the spring bloom (OSB) occurs when phytoplankton growth exceeds losses and is promoted by a transition from deep convection to a shallow mixing layer concurrent with increasing light intensities in nutrient-enriched waters. We have combined remotely sensed chlorophyll-a data and high-resolution sea-surface winds to quantify and understand high-latitude spring-bloom dynamics and the effect of varying winds. Increasing winds strengthen turbulent mixing and may eventually cause the mixing depth to extend beyond the depth at which light is favorable for net growth and delay the OSB. We find that wind intensity accounts for up to 60 % of the interannual variation in the OSB as revealed by remotely-sensed chlorophyll-a values at the key spawning ground (62-63 oN) of one of the worlds’ largest herring stocks. The OSB is, on average, one month later and with about half the variability farther north at the main spawning ground (67-68 oN) of one of the world’s largest cod stocks. Since the atmospheric reanalysis considered here extends wind time series much further back in time (1958) than remote sensing (1998), the former may act as a good proxy for investigating OSB trends on the time scales of multi-decadal variability and climate change. We find a weak but non-significant signal of delay in the OSB across these extended time periods. More importantly, our results clearly show that predictions of future productivity and ecosystem dynamics under global warming based on earth system models require accurate representation of winds.

On the driving forces of historical changes in the fatalities of tropical cyclone disasters in China from 1951 to 2014

Tropical cyclone (TC) disasters have frequently caused casualties in the coastal areas of China. According to the statistics of dead and missing people due to TCs from 1951 to 2014, the number of fatalities has been significantly decreasing over time. However, deadly TC events have still caused great losses of life in recent years, which are characterized as significant abrupt fluctuations superimposed along the downward trend of the long-term fatality time series. The numbers of fatalities caused by TC disasters are influenced by variables such as the intensity of TC hazards, the population exposed to TCs and the vulnerability of people to TC hazards. It is thus of great significance to analyze their temporal characteristics and understand the forces driving these changes. First, the time series of the TC wind, precipitation, spatial distribution of population, fatality and disaster risk reduction (DRR) measures of China from 1951 to 2014 are reconstructed. Second, the improved power dissipation index, total precipitation, integrated intensity and index of exposed population are calculated, and the population vulnerability indices, including mean and relative fatality rates, are derived. Third, the change trend of each index is detected using the Mann–Kendall test. Finally, the main driving factors of the long-term change trend and fluctuations of the TC fatalities are analyzed by a negative binomial regression model and standard deviation statistics. It is found that the decrease in vulnerability based on the improvement in structural and non-structural measures is the main driving force of the decreases in fatalities over the past six decades. Although the total population and exposure have increased dramatically in the coastal areas of China, their contributions to the increase in the fatality risk were counteracted by the decrease in vulnerability. Abrupt and catastrophic disasters were mostly caused by TCs with hazards of high intensity that surpassed the capacity of structural measures; the lack of forecasting or early warning, as well as improper emergency response actions, may also have triggered the great loss of lives. To reduce the fatalities of future TCs, especially those that may exceed the capacity of structural measures, the enhancement of non-structural measures and the adaptation of resilience strategies should be priorities for future people-centered disaster management.

Pattern identification for wind power forecasting via complex network and recurrence plot time series analysis

Renewable energy sources, where wind energy is an important part, are increasingly participating in developing economies and environmental benefits. Wind power is strongly dependent on wind velocity and thus identifying patterns in wind speed data is an important issue for forecasting the generated power from a wind turbine and it has significant importance for the renewable energy market operations. In this work we approach the problem of identification of the underlying dynamic characteristics and patterns of wind behavior using two approaches of non-linear time series analysis tools: Recurrence Plots (RPs) and Complex Network analysis. The proposed methodology is applied on wind time series collected by cup anemometers located on a wind turbine installed in Greece. We show that the proposed approach provides useful information which can characterize distinct two time intervals of the data, one ranging from 2 to 4.5 days and another from 5 to 8.5 days. Also analysis can identify and detect dynamical transitions in the system's behavior and also reveals information about the changes in state inside the whole time series. The results will be useful in wind markets, for the prediction of the produced wind energy and also will be helpful for wind farm site selection.

“Dust in the Wind...”, Deep Learning Application to Wind Energy Time Series Forecasting

To balance electricity production and demand, it is required to use different prediction techniques extensively. Renewable energy, due to its intermittency, increases the complexity and uncertainty of forecasting, and the resulting accuracy impacts all the different players acting around the electricity systems around the world like generators, distributors, retailers, or consumers. Wind forecasting can be done under two major approaches, using meteorological numerical prediction models or based on pure time series input. Deep learning is appearing as a new method that can be used for wind energy prediction. This work develops several deep learning architectures and shows their performance when applied to wind time series. The models have been tested with the most extensive wind dataset available, the National Renewable Laboratory Wind Toolkit, a dataset with 126,692 wind points in North America. The architectures designed are based on different approaches, Multi-Layer Perceptron Networks (MLP), Convolutional Networks (CNN), and Recurrent Networks (RNN). These deep learning architectures have been tested to obtain predictions in a 12-h ahead horizon, and the accuracy is measured with the coefficient of determination, the R² method. The application of the models to wind sites evenly distributed in the North America geography allows us to infer several conclusions on the relationships between methods, terrain, and forecasting complexity. The results show differences between the models and confirm the superior capabilities on the use of deep learning techniques for wind speed forecasting from wind time series data.

Reversibility of Turbulent and Non-Collisional Plasmas: Solar Wind

AbstractWe have studied turbulent plasma as a complex system applying the method known as Horizontal Visibility Graph (HVG) to obtain the Kullback-Leibler Divergence (KLD) as a first approach to characterize the reversibility of the time series of the magnetic fluctuations. For this, we have developed the method on Particle In Cell (PIC) simulations for a magnetized plasma and on solar wind magnetic time series, considering slow and fast wind. Our numerical results show that low irreversibility values are verified for magnetic field time series associated with Maxwellian distributions. In addition, considering the solar wind plasma, our preliminary results seem to indicate that greater irreversibility degrees are reached by the magnetic field associated with slow solar wind.

A regime-switching recurrent neural network model applied to wind time series

This paper proposes a regime-switching recurrent network model (RS-RNN) for non-stationary time series. The RS-RNN model emits a mixture density with dynamic nonlinear regimes that fit flexibly data distributions with non-Gaussian shapes. The key novelties are: development of an original representation of the means of the component distributions by dynamic nonlinear recurrent networks, and derivation of a corresponding expectation maximization (EM) training algorithm for finding the model parameters. The results show that the RS-RNN applied to a real-world wind speed time series achieves standardized residuals similar to popular previous models, but it is more accurate distribution forecasting than other linear switching (MS-AR) and nonlinear neural network (MLP and RNN) models.

Deep Learning is blowing in the wind. Deep models applied to wind prediction at turbine level

Wind Energy generation depends on the existence of wind, a meteorological phenomena intermittent by nature, with the consequence of generating uncertainty on the availability of wind energy in the future. The grid stability processes require continuous forecasting of wind energy generated. Forecasting wind energy can be performed either by using weather forecast data or by projecting (or regressing) the past time-series data observations into the future. This last method is the statistical or time series approach. Wind Time Series show non-linearity and non-stationarity properties, and these two properties increase the complexity of the forecasting task using statistical methodologies. In this paper we explore the use of deep learning techniques, which can represent non-linearity, to the wind speed prediction using the largest public wind dataset available, the Wind Toolkit from the National Renewable Laboratory of the US. Several deep network architectures like Multi Layer Perceptrons, Convolutional Networks or Recurrent Networks have been tested on the 126,692 wind-sites and with the results obtained valuable comparisons and conclusions have been obtained. The distribution of the wind sites across the North American Geography has allowed to include in the analysis relationships between terrain, wind forecast complexity and deep methods. With the developed testing workbench and with the availability of the Barcelona Supercomputing Center new architectures are being developed. This work concludes with the feasibility of deep learning architectures for the wind and energy forecasting.

Wavelet variance scale-dependence as a dynamics discriminating tool in high-frequency urban wind speed time series

High frequency wind time series measured at different heights from the ground (from 1.5 to 25.5 meters) in an urban area were investigated by using the variance of the coefficients of their wavelet transform. Two ranges of scales were identified, sensitive to two different dynamical behavior of the wind speed: the lower anemometers show higher wavelet variance at smaller scales, while the higher ones are characterized by higher wavelet variance at larger scales. Due to the relationship between wavelet scale and frequency, the results suggest the existence of two frequency ranges, where the wind speed variability change according to the position of the anemometer from the ground.This study contributes to better understanding of the high frequency wind speed in urban areas and to a better knowledge of the underlying mechanism governing the wind fluctuations at different heights from the ground in particular in urban area.

Past, present and future environmental footprint of the Danish wind turbine fleet with LCA_WIND_DK, an online interactive platform

Renewable energy systems are promoted and developed notably due to their low environmental footprint. Fleet-wide robust environmental assessments are needed to drive the sustainable transition of energy systems worldwide. This study introduces a tailored comprehensive impact assessment methodology for fleets of renewable energy systems based on Life Cycle Analysis and its application to Danish wind turbines fleet through an online platform LCA_WIND_DK (viewer.webservice-energy.org/lca-wind-dk/). This platform enables to visualize environmental performances of wind turbines in Denmark and their temporal evolution. The fleet is known in detail from 1980 to 2016 and projected from 2017 to 2030 based on national objectives for onshore/offshore capacity and pre-approved offshore projects. Each turbine's future electricity production is estimated from its power curve and geo-localized wind time-series. More than 10,000 cradle-to-grave life cycle inventories are generated, considering the spatio-temporal context and technological characteristics. The comprehensive analysis of the Danish fleet over fifty years reveals long-term trends for several impact categories. Improvements in all categories follow similar trends as in climate change, which decreases from 40 to 13 g CO2-eq/kWh between 1980 and 2030. Improvements stem from combined economies of scale and higher load factors linked to increasingly large and powerful turbines. The interactive mapping tool LCA_WIND_DK may provide statistics to support renewable energy oriented policy scenarios and unique spatio-temporal environmental information to project developers. This novel approach designed for large territories, here applied to the Danish wind turbine fleet, is generic and can be applied to other renewable energy systems and/or to other territories.

강원 중부 산지 및 동해안 지상 바람 시계열 스펙트럼 분석

한반도 전체적으로 기상청의 지상 관측 지점은 거의 600 소에 이른다. 하지만, 여기서 관측된 시계열을 이용한 스펙트럼 연구는 매우 드물다. 본 연구는 강원도 산지 및 동해안 지역의 약 65 km × 8 35 km 영역 내 평균 거리 5- 내지 10-km 간격으로 위치한 기상청의 13곳 지상 관측소에서 5년간 수집한 1분 간격 풍속 시계열 자료를 스펙트럼 분석한 결과이다. 풍력 스펙트럼은 십 분 정도 보다 고주파수 영역 즉 십여 분보다 작은 시간 규모에서 명확한 에일리어싱(aliasing)을 보였다. 이는 수집된 시계열의 관측 간격 1분의 니퀴스트 (Nyquist) 주파수보다 고주파수 영역에서 심각한 난류 에너지에 기인한다고 사료 된다. 관측이 복잡한 지형과 해안에 인근 한 지역 내에 이루어졌음을 고려하면, 이러한 강한 난류의 특성은 당연한 결과라 할 수 있다. 이 분석 결과의 실용적 조언은, 특히 산지 및 해안 지역상에서 관측된 1분 간격 시계열의 경우, 이를 10분 평균 또는 이보다 긴 시간에 대한 평균으로 사용하는 것이 바람직하다는 것이다. 이에 더하여 풍속 스펙트럼은 기온과 수증기압 스펙트럼이 1년 주기에서 가장 강한 피크를 보인 것과는 상이하게, 1일 주기에서 가장 강한 피크를 보였다. 즉, 이 지역 바람 장의 경우, 종관장, 계절 변화 또는 연변화와 연관된 바람 장보다는, 해륙풍 및 산곡풍 등 일변화와 연관된 바람 장에 주목할 필요가 있다는 것을 의미한다. 이외에도 본 연구는 여름철보다 겨울철에 스펙트럼 에너지가 크다는 것을 확인하였다. 하루를 기준으로 여름철은 이 이하의 시간 규모 영역에서, 겨울철이 이보다 큰 시간 규모 영역에서 스펙트럼 에너지가 상대적으로 많음을 확인하였다. 이 결과와 바람 장미 분석 결과의 연결에 대하여 본문에 자세히 설명하였다.

Transitions in the wintertime near-surface temperature inversion at Dome C, Antarctica.

In this work we study the dynamics of the surface‐based temperature inversion over the Antarctic Plateau during the polar winter. Using 6 years of observations from the French–Italian Antarctic station Concordia at Dome C, we investigate sudden regime transitions in the strength of the near‐surface temperature inversion. Here we define “near‐surface” as being within the domain of the 45‐m measuring tower. In particular, we consider the strongly nonlinear relation between the 10‐m inversion strength (T 10m – T s) and the 10‐m wind speed.To this end, all individual events for which the 10‐m inversion strength increases or decreases continuously by more than 15 K in time are considered. Composite time series and vertical profiles of wind and temperature reveal specific characteristics of the transition from weak to very strong inversions and vice versa. In contrast to midlatitudes, the largest variations in temperature are not found at the surface but at a height of 10 m.A similar analysis was performed on results from an atmospheric single‐column model (SCM). Overall, the SCM results reproduce the observed characteristics of the transitions in the near‐surface inversion remarkably well. Using model output, the underlying mechanisms of the regime transitions are identified. The nonlinear relation between inversion strength and wind speed at a given level is explained by variations in the geostrophic wind speed, changes in the depth of the turbulent layer and the vertical divergence of turbulent fluxes. Moreover, the transitions between different boundary layer regimes cannot be explained without considering the contribution of subsidence heating.

21st Century Climate Change Impacts on Key Properties of a Large-Scale Renewable-Based Electricity System

Efforts to reduce climate change, but also falling prices and significant technology developments currently drive an increased weather-dependent electricity production from renewable electricity sources. In light of the changing climate, it is highly relevant to investigate the extent of weather changes that directly impacts the best system design decisions for these weather-dependent technologies. Here, we use three IPCC representative CO2 concentrations pathways for the period 2006--2100 with six high-resolution climate experiments for the European domain. Climate elements are used to calculate bias adjusted 3-hourly time series of wind and solar generation, and temperature corrected demand time series for 30 European countries using state-of-the-art methodology. Weather-driven electricity system analysis methodology is then applied to compare four key metrics of highly renewable electricity systems. We find that climate change does not have a discernible impact on the key metrics of the combined electricity system dynamics, and conclude that the effect on important system design parameters can likely be ignored.

Linearity versus non-linearity in high frequency multilevel wind time series measured in urban areas

In this paper, high frequency wind time series measured at different heights from the ground (from 5.5 to 25.5 meters) in an urban area were investigated. The spectrum of each series is characterized by a power-law behaviour at low frequency range, with a mean spectral exponent of about 1.5, which is rather consistent with the Kolmogorov spectrum of atmospheric turbulence. The detrended fluctuation analysis was applied on the magnitude and sign series of the increments of wind speed, in order to get information about the linear and nonlinear dynamics of the time series. Both the sign series and magnitude series are characterized by two timescale ranges; in particular the scaling exponent of the magnitude series in the high timescale range seems to be related with the height of the sensor. This study aims to understand better high frequency wind speed in urban areas and to disclose the underlying mechanism governing the wind fluctuations at different heights.

Characterization of internal inertia gravity wave over a low latitude: results from the RONAC-2012 campaign

High height resolution GPS sonde measurements have been utilized to understand the characteristics of internal inertia gravity waves (IGWs) over a low-latitude station Trivandrum (8.5°N, 76.9°E) during an intensive campaign “Research on OrganisatioN of Atmospheric convection (RONAC), which was initiated to understand the organization of convective systems during the Indian summer monsoon (ISM). GPS sondes had been launched at every 3 hourly intervals for a period of 10 days before the onset of ISM from 23 May to 1 June 2012. This unique dataset is utilized to understand the characteristics of IGW over this latitude where the effect of the Coriolis force is minimum. Observations show that the onset of tropical easterly jet (TEJ) was established well before the onset of ISM. Internal IGW activity with ~ 78 h periodicity associated with TEJ was observed in the upper troposphere and lower stratosphere (UTLS) regions. We have reconstructed the time series of winds to understand the propagation characteristics and employed Stokes parameter method to extract the other wave parameters. Vertical and horizontal wavelengths of ~ 2.3 km and ~ 2620 km were observed, respectively. The significance of the present study lies in the characterization of internal IGW over Trivandrum for the first time, which will be an important addition to the present understanding of the IGW and hence to the gravity wave community.

Assessment of Wind Energy Prospect by Weibull Distribution for Prospective Wind Sites in Australia

In this study, three potential wind sites such as Hamilton Island, Proserpine, and South Johnstone wind speed data has been analyzed using Weibull distribution. The half-hourly time series wind speed data used in this analysis were obtained from Bureau of Meteorology, Australia wind monitoring stations. With the aim assessing the wind energy potential, method of moments (MOM), empirical method (EM), and power density method (PDM) were employed for determining the Weibull factors of the available data. The accuracy of the results was analyzed using a statistical test tool named chi-square error (χ2). The EM method was found to be the best method among the three studied methods for calculating Weibull parameters. Based on the analyzed data, a horizontal axis wind turbine specification was proposed. Finally, wind power profile of Hamilton Island site was evaluated for the proposed turbine specification.

Fatigue analysis of jacket support structure for offshore wind turbines

In the past few decades and up to now, the fossil energy has exerted tremendous impacts on human environments and gives rise to greenhouse effects while the wind power, especially in offshore region, is an attractive renewable energy resource. For offshore fixed wind turbine, stronger foundation like jacket structure has a good applicability for deeper water depth. Once water depth increases, dynamic responses of offshore wind turbine (OWT) support structures become an important issue. The primary factor will be the total height of support structure increases when wind turbine is installed at offshore locations with deeper water depth, in other words the fatigue life of each components of support structure decrease. The other one will experience more wind forces due to its large blades, apart from wave, current forces, when makes a comparison with offshore oil and gas platforms. Summing up two above reasons, fatigue analysis, in this research, is a crucial aspect for design of offshore wind turbine structures which are subjected to time series wind, wave loads and carried out by aiding of SACS software for model simulation (P-M rules and S-N curves) and Matlab code. Results show that the fatigue life of OWT is decreased accordingly by increasing the wind speed acting on the blades, especially with the simultaneous interaction between wind and wind-induced wave. Hence, this should be considered in wind turbine design.
 Keywords: offshore wind turbine; Jacket structure; fatigue analysis; P-M rules; S-N curves.
 Received 01 October 2018, Revised 19 November 2018, Accepted 31 January 2019