Wind Farm Power Research Articles

Data‐driven stochastic model predictive control for regulating wind farm power generation with controlled battery storage

AbstractThe uncertainty and fluctuation of the volatile wind power cause more reserve and frequency regulation capacity, so it incurs additional operational cost for the power grid. Therefore, a wind farm (WF) coordinated controller is essential to reduce the power fluctuation and trace the scheduled power generation with minimal wind curtailment and battery degradation cost. This paper proposes a data‐driven stochastic model predictive control (SMPC) method for WFs to realize these objectives. Firstly, to address the non‐linear dynamic model of wind turbines, the dynamic mode decomposition method accompanying Koopman operator is used to learn a lifted linear dynamic model from measurements. Secondary, this paper proposes the relationship between the curtailed wind power of wind turbines and their pitch angle, wind speed, and rotation speed, as well as the corresponding estimation formula. In addition, the prediction errors of wind speeds are characterized using the Gaussian mixture model (GMM), which may significantly affect the controller's performance. Finally, a tractable data‐driven SMPC model is developed and is verified on a DIgSILENT/PowerFactory based simulation platform. Compared with conventional model predictive control algorithms, it can significantly reduce the overall operational cost of WFs.

Wake and power prediction of horizontal-axis wind farm under yaw-controlled conditions with machine learning

The main objective of this study is to employ the Extreme Gradient Boosting (XGBoost) machine learning algorithm to predict the power, wake, and turbulent characteristics of horizontal-axis wind farms under yaw-controlled conditions. For this purpose, a series of high-fidelity numerical simulations using LES method are performed over tandem NREL-5 MW wind turbines to generate the input data for training and testing in machine learning analysis. It is observed that XGBoost is more accurate for wake prediction of the yaw-controlled wind farms compared to ANN, which was used in previous studies. The results illustrate that XGBoost can predict the power with a mean deviation of 0.94 % for different yaw angles, while ANN can estimate the power generation with a mean deviation of 2.15 % for various tested yaw angles. At far wake regions (X > 2000 m) of the second wind turbine, the deviations reach below 1 %. Moreover, XGBoost requires a much shorter training time, 87.5 % faster than ANN. The power production of both wind turbines can be predicted more accurately with XGBoost compared to ANN. The wake prediction time of XGBoost is just 0.105 sec, while this time is 4.480 for the ANN model. In conclusion, XGBoost provides a significant reduction in error and training time compared to ANN and deep learning algorithms over yaw-misaligned wind farms.

Capacity evaluation of hydropower for accommodating wind‐photovoltaic power generation in the dry season

AbstractConsidering that the benefits of hydropower stations from regulating and consuming new energy are restricted during the dry season due to the inadequate regulation capacity, this paper focuses on the accommodation capacity evaluation of hydropower for wind‐ photovoltaic (PV, for short) power generation in different supply‐grid‐load conditions. First, the factors that influence hydropower stations' accommodation capacity were initially examined in this study, in terms of power supply, power grid, and load. Then, two new indices, namely the regulatory factor of installed capacity and the regulatory factor of electric quantity, were put forward to evaluate the capacity of hydropower for accommodating wind‐PV. To reduce the peak regulation burden on the power grid and encourage the consumption of clean energy, a two‐objective optimization model with maximum source‐load matching and clean energy consumption was devised. And a two‐tier binary algorithm and progressive optimization coupled strategy was proposed to solve the model. Finally, the cascade hydropower stations in the north Panxi section of the Yalong River and the surrounding wind farm and PV power stations were taken as examples to examine the sensitivity of the factors. The results show that with each 5% increase in the regulatory electric quantity, the regulatable installed wind‐PV capacity and consumable wind‐PV electric quantity were increased by 7.57% and 8.21%, respectively, rank only second to transmission capacity. Considering the large investment required for transmission channel expansion, this result reveals that the most effective and economical way to promote wind‐PV consumption is to increase the regulated electric quantity of hydropower stations in the river basin.

Experimental study on deterioration of bedrock strength and P-wave velocity by pressurized seawater

Bedrock has to be excavated before the construction of rock-socketed pile foundations for offshore wind power turbines, leading to the exposure of fresh and dry rocks to seawater. The pile-rock interface is partially separated under the long-term action of various cyclic loads during the operation stage of offshore wind turbines, and the infusion of seawater will further weaken the bedrock. To comprehensively investigate the deterioration of bedrocks by pressurized seawater, tests on point load strength and slake durability were carried out to study the long-term deterioration of mechanical properties of granite, sandstone, and tuff collected from an offshore wind power farm. The elastic P-wave velocity was measured to analyze the variation trend of P-wave velocity, the mechanism of which was revealed by means of Gassmann’s equation. The results of mechanical experiments indicate that granite is almost impervious to pressurized seawater, while the mechanical properties of sandstone and tuff are deteriorated by water, especially by the pressurized seawater. After 60 days of immersion in seawater at 0.5 MPa pressure, decreasing amplitude of point load strength and slake durability index for the sandstone specimens in seawater reaches 53.2% and 18.5%, respectively, and that for the tuff specimens is 14.6% and 1.52%, respectively. The elastic P-wave velocity of granite shows an upward trend with increasing immersion time in different environments, but there exists a different tendency between the P-wave velocity of sandstone and tuff specimens in water or seawater with normal pressure and that in seawater with 0.5 MPa pressure. The variation of P-wave velocity of different types of rocks is studied based on Gassmann’s equation that takes into account the change of saturation degree, porosity, and compressibility of the matrix material.

Ensemble prediction aided multi-objective co-design optimizations of grid-connected integrated renewables for green hydrogen production

The use of renewables including wind and photovoltaic (PV) powers to produce green hydrogen will be of great importance for promoting cleaner productions. However, there is a need for the co-design optimizations of integrated renewables-based hydrogen productions considering interactions between renewables and power grid. This work investigates this type of green hydrogen production which is achieved in a distributed manner through the power grid. An overall modelling framework is provided to describe characteristics of the main subsystems including wind farms, power grid, PV array, pumped hydro storages and the electrolyzers considering wake steering conditions of wind farms. Then, a co-design optimization methodology using ensemble learning aided wind farm power predictions is proposed. The proposed methodology is tested by using design experiments based on the IEEE power grid. The results indicate that wind farm power prediction errors can be reduced from around 13%–4% by using the ensemble learning compared with a gaussian method. It is then possible that parameters of the renewables and water electrolyzer can be optimized simultaneously such that the grid-connected hydrogen productions can be kept stable. Also, the voltage oscillations (grid cost) can be well kept within appropriate stable range of [0, 0.15] p. u. under different wind conditions.

Optimal selection of time windows for preventive maintenance of offshore wind farms subject to wake losses

AbstractThe maintenance of wind farms is one of the major factors affecting their profitability. During preventive maintenance, the shutdown of wind turbines causes downtime energy losses. The selection of when and which turbines to maintain can significantly impact the overall downtime energy loss. This paper leverages a wind farm power generation model to calculate downtime energy losses during preventive maintenance for an offshore wind farm. Wake effects are considered to accurately evaluate power output under specific wind conditions. In addition to wind speed and direction, the influence of wake effects is an important factor in selecting time windows for maintenance. To minimize the overall downtime energy loss of an offshore wind farm caused by preventive maintenance, a mixed‐integer nonlinear optimization problem is formulated and solved by the genetic algorithm, which can select the optimal maintenance time windows of each turbine. Weather conditions are imposed as constraints to ensure the safety of maintenance personnel and transportation. Using the climatic data of Cape Cod, Massachusetts, the schedule of preventive maintenance is optimized for a simulated utility‐scale offshore wind farm. The optimized schedule not only reduces the annual downtime energy loss by selecting the maintenance dates when wind speed is low but also decreases the overall influence of wake effects within the farm. The portion of downtime energy loss reduced due to consideration of wake effects each year is up to approximately 0.2% of the annual wind farm energy generation across the case studies—with other stated opportunities for further profitability improvements.

Elman short-term wind power prediction based on the optimized seagull algorithm

Accurate prediction of wind farm power output can relieve the pressure of grid frequency regulation and peak regulation and improve grid stability. With the goal of improving power prediction accuracy and reducing overall prediction error, this paper proposes an Elman short-term wind power prediction model on the basis of an optimized seagull algorithm. Firstly, the Elman network is used as the base prediction model, and the seagull algorithm is applied to seek the best values for its weights. Secondly, the chaotic circle mapping with better initial characteristics is improved to equalize its sequence distribution for optimizing the population initialization. Then, to address the lack of local search capability, an optimized iterative approach using the sine cosine operator is used to achieve a balance between local exploitation ability and global search capability. Finally, after simulation and analysis of the actual data set, it is verified that the model has a better prediction effect.

Blockage and speedup in the proximity of an onshore wind farm: A scanning wind LiDAR experiment

To maximize the profitability of wind power plants, wind farms are often characterized by high wind turbine density leading to operations with reduced turbine spacing. As a consequence, the overall wind farm power capture is hindered by complex flow features associated with flow modifications induced by the various wind turbine rotors. In addition to the generation of wakes, the velocity of the incoming wind field can reduce due to the increased pressure in the proximity of a single turbine rotor (named induction); a similar effect occurs at the wind-farm level (global blockage), which can have a noticeable impact on power production. On the other hand, intra-wind-farm regions featuring increased velocity compared to the freestream (speedups) have also been observed, which can be a source for a potential power boost. To quantify these rotor-induced effects on the incoming wind velocity field, three profiling LiDARs and one scanning wind LiDAR were deployed both before and after the construction of an onshore wind turbine array. The different wind conditions are classified according to the ambient turbulence intensity and streamwise/spanwise spacing among wind turbines. The analysis of the mean velocity field reveals enhanced induction and speedup under stably stratified atmospheric conditions. Furthermore, a reduced horizontal area between adjacent turbines has a small impact on the induction zone but increases significantly the speedup between adjacent rotors.

A power storage system planning model for the Wolfe Island wind farm

As one of the alternatives to conventional energy sources, wind power is a fast-growing renewable technology. The intermittent characteristic of wind speed leads to unstable wind power production; therefore, a storage system is typically needed at wind power farms to stabilize the power output. This project aims to develop a power storage system planning model to optimize the power transfer between wind turbines and storage devices on an hourly basis to stabilize power output and maximize annual net earnings. The developed model utilizes the dynamic programming technique, and it is applied to a case study of the Wolfe Island Wind Farm in Ontario, Canada, for the year of 2020. Two scenarios with different power output tolerance considerations (i.e., tolerance ranges based on 6- and 8-hour subperiods) are analyzed to compare the fluctuations of power output and the total net earnings. The optimized number of storage devices is also discussed. The developed model can provide technical support for the design and operation of power storage systems at wind power farms.

Wind farm power maximization based on analytical sensitivity model considering wake effect

To reduce the wake effect on wind energy capture, it is necessary to optimize the operating parameters of each wind turbine (WT) in the wind farm (WF) to improve the overall wind power. Considering the complex coupling characteristics of wind energy captured by upstream and downstream WTs caused by the wake effect, existing optimization methods for the WF usually adopt intelligent algorithms or numerical sensitivity-based search algorithms. However, most of the existing studies overlook the exploration of the physical mechanism underlying the wake influence on WF power, and the optimization speed still needs to be improved. To this end, in this paper, the analytical sensitivity model of WF power is formulated with the adjustable parameter - the pitch angle which is a key factor for the wake effect. On this basis, the improvement mechanism of how the WF power can be improved by changing the wake effect through the pitch angle is analyzed using the Blade Element Momentum theory, which can be used to guide and accelerate the optimization process. Furthermore, the WF power maximization method based on the analytical sensitivity model is proposed, and its effectiveness and advancement are verified based on SimWindFarm simulation under different operation scenarios.

Energy Price Decoupling and the Split Market Issue

Load scheduling and dispatch by merit order on electricity generation networks has used a wholesale market electricity system operator model focused on system marginal pricing, in which the spot price of electricity at any point in time is equal to the system marginal cost given by the higher value of the price, which rations demand to capacity or the operating cost of the most expensive plant on the system, which is usually a fossil fuel price. This idea has come under challenge because renewable technologies such as wind power farms or solar power farms are treated as having close to zero operating costs. The challenges, under the general heading of energy price decoupling, include suggestions for operating split markets possibly overseen by a regulator, and the prediction that marginal cost pricing should be abandoned. This review evaluates these in terms of their economic impact, relating them to the policy debates on electricity market reform.

A wind tunnel study on cyclic yaw control: Power performance and wake characteristics

In this study, wind-tunnel experiments are used to investigate cyclic yaw control as a wake mitigation strategy aiming to improve wind farm power production. The control strategy is applied to a single wind turbine and a wind farm comprising of a column of three wind turbines in full wake state. The power performance of the wind farm is optimized by adjusting the control parameters, which include the yaw angle amplitude and yaw Strouhal number, and the wake flows are investigated using particle-image velocimetry measurements. The yaw angle of the controlled turbine is varied dynamically in a sinusoidal manner. A power gain of more than 10 % compared to an uncontrolled wind farm is achieved when the yaw angle amplitude and yaw Strouhal number are in the range of 10°–35° and 0.09-0.22, respectively, with a peak power gain of 15.2 %. The power consumed to generate the control is found to be small and predictable. Results from wake measurements reveal that by deflecting the wake cyclically, this control strategy can lead to relatively flattened mean wake velocity deficit profiles, faster wake recovery and very different wake turbulence intensity distributions, compared to the uncontrolled case. The results show that cyclic yaw control has a great potential in improving the power production of wind farms. A scaling of the control parameters based on the kinematic similarity between the wind-tunnel scale and utility-scale wind farms is also proposed.

Modelling the impact of trapped lee waves on offshore wind farm power output

Abstract. Mesoscale meteorological phenomena, including atmospheric gravity waves (AGWs) and including trapped lee waves (TLWs), can result from flow over topography or coastal transition in the presence of stable atmospheric stratification, particularly with strong capping inversions. Satellite images show that topographically forced TLWs frequently occur around near-coastal offshore wind farms. Yet current understanding of how they interact with individual turbines and whole farm energy output is limited. This parametric study investigates the potential impact of TLWs on a UK near-coastal offshore wind farm, Westermost Rough (WMR), resulting from westerly–southwesterly flow over topography in the southeast of England. Computational fluid dynamics (CFD) modelling (using Ansys CFX) of TLW situations based on real atmospheric conditions at WMR was used to better understand turbine level and whole wind farm performance in this parametric study based on real inflow conditions. These simulations indicated that TLWs have the potential to significantly alter the wind speeds experienced by and the resultant power output of individual turbines and the whole wind farm. The location of the wind farm in the TLW wave cycle was an important factor in determining the magnitude of TLW impacts, given the expected wavelength of the TLW. Where the TLW trough was coincident with the wind farm, the turbine wind speeds and power outputs were more substantially reduced compared with when the TLW peak was coincident with the location of the wind farm. These reductions were mediated by turbine wind speeds and wake losses being superimposed on the TLW. However, the same initial flow conditions interacting with topography under different atmospheric stability settings produce differing near-wind-farm flow. Factors influencing the flow within the wind farm under the different stability conditions include differing, hill and coastal transition recovery, wind farm blockage effects, and wake recovery. Determining how much of the differences in wind speed and power output in the wind farm resulted from the TLW is an area for future development.

Optimization of floating wind farm power collection system using a novel two-layer hybrid method

Deep-sea floating wind farms have gained significant attention in recent years, however, their high construction and operation cost remains a notable challenge. In the floating wind farm, floating wind turbines (WTs) have a certain range of movement, and dynamic cables are used, thus the collection system needs a careful design that relies heavily on iterative optimization. In this study, a life cycle cost model for the collection system of floating wind farms is firstly established considering the floating characteristics. Due to the complexity of the optimization model, the two-layer optimization framework is presented. The upper layer uses Binary Particle Swarm Optimization (BPSO) to determine the optimal connection between the WTs and offshore substations (OSs), while the lower layer optimizes the cable topology between WTs via Improved Monte Carlo Tree Search (IMCTS). Simultaneously, the cable selection scheme corresponding to the obtained topology is optimized by the lower level. Simulations are performed to validate the proposed method. The results demonstrate that the proposed algorithm achieves a minimum optimization cost of 45.277 million euros, and attains significant reductions of 2.31, 0.486 and 1.667 million euros, respectively, compared to Genetic Algorithm (GA), BPSO-Sweep-IMCTS and BPSO-AA-GA. Further, this study evaluates the differences in the optimized structures of the collection system between floating and fixed wind farms, and analyzes the effects of the OS position and environmental factors. The analysis results show that locating OSs at the central position and upstream side reduces the cost of the collection system, and the wind condition has a considerable influence on the design of the collection system.

Effectiveness of wake control optimization for multiple in-line wind turbines by combinatorial machine learning wake model

In the wind turbine wake control research, the analytical wake model does not consider the influence of direct control parameters on the wind turbine wake flow field, and its prediction accuracy is by no means satisfying. The numerical simulation method can accurately predict the wake effect, but its low computational efficiency renders it unable to be used in the iterative control optimization study. In this paper, a new wake model coupled with the direct control parameters (tip speed ratio and pitch angle) of a single wind turbine is built based on the machine learning algorithm for which the widely applied Artificial Neural Network (ANN) is applied for instance. By combining the ANN wake model with different wake superposition models, the best combinatorial model for quantitative evaluation of multiple wake interferences is achieved, which not only has great accuracy but also ensures high calculation efficiency. On top of the established combinatorial wake model, three in-line wind turbines are investigated for assessing the effectiveness of wake control optimization with respect to the traditional greedy control. The results show that by optimizing the control parameters (tip speed ratio and pitch angle) of wind turbines, the total power of the three wind turbine units is increased by 1.15~3.48% depending on the incoming wind speeds and distances apart between wind turbines. The power variation of each wind turbine shows that the increase in total power is achieved by the decreased power of the first wind turbine and the increased power of the second and third wind turbines. The research verifies the effectiveness of adopting optimized wind turbine wake control on the total wind farm power improvement numerically.

Data‐driven modelling of turbine wake interactions and flow resistance in large wind farms

AbstractTurbine wake and local blockage effects are known to alter wind farm power production in two different ways: (1) by changing the wind speed locally in front of each turbine and (2) by changing the overall flow resistance in the farm and thus the so‐called farm blockage effect. To better predict these effects with low computational costs, we develop data‐driven emulators of the ‘local’ or ‘internal’ turbine thrust coefficient as a function of turbine layout. We train the model using a multi‐fidelity Gaussian process (GP) regression with a combination of low (engineering wake model) and high‐fidelity (large eddy simulations) simulations of farms with different layouts and wind directions. A large set of low‐fidelity data speeds up the learning process and the high‐fidelity data ensures a high accuracy. The trained multi‐fidelity GP model is shown to give more accurate predictions of compared to a standard (single‐fidelity) GP regression applied only to a limited set of high‐fidelity data. We also use the multi‐fidelity GP model of with the two‐scale momentum theory (Nishino & Dunstan 2020, J. Fluid Mech. 894, A2) to demonstrate that the model can be used to give fast and accurate predictions of large wind farm performance under various mesoscale atmospheric conditions. This new approach could be beneficial for improving annual energy production (AEP) calculations and farm optimization in the future.

Optimal control of wind farm power output with delay compensated nested-loop extreme seeking control

In this paper, we propose to enhance the nested-loop extremum seeking control (NLESC)-based wind farm control strategy with the predictor-based delay compensation in order to improve its convergence characteristics under fluctuating wind. Earlier work has shown the effectiveness of NLESC for region-2 wind farm operation, i.e., maximizing the total power output of cascaded wind turbine array, while its convergence speed is highly limited by the delay of power output for downstream turbines due to wake propagation along the wind direction. By utilizing the delay compensated ESC proposed by Oliveira and Krstic, the delay compensated NLESC (DCNLESC) wind farm control is proposed, allowing the dither frequencies to be of similar magnitude as that in the single-turbine ESC. This can significantly improve the convergence speed of optimum tracking for real-time wind farm control. The wake propagation delay is estimated from turbine power outputs using cross correlation and proper filtering. Using the SimWindFarm platform, the proposed DCNLESC strategy is simulated with both a single-column three-turbine array and a 2 × 3 turbine array, under different wind speeds. The results show that the convergence speed toward the calibrated optimum is significantly improved over the NLESC operation. The convergence time for the upstream turbines' torque gain is reduced by 55%–14% in terms of integral time-weighted absolute error, while the impact on turbine fatigue loads is as low as no more than 3.5% increase on turbine tower and shaft.

A D-stacking dual-fusion, spatio-temporal graph deep neural network based on a multi-integrated overlay for short-term wind-farm cluster power multi-step prediction

The uncertainty of wind energy due to its non-stationary and random nature poses a major challenge to engineers responsible for power system scheduling. In the present research, a spatio-temporal graph deep neural network is used to learn the spatial characteristics and explore the strong correlation characteristics of wind-farm clusters. The Bagging, long short-term memory (LSTM), and random forest (RF) are superimposed based on the stacking integration algorithm. By analyzing the decomposed wind electron sequence characteristics, two Stacking models are used to predict the wind power of target wind-farms respectively and then a dual stacking model fusion strategy is formed. Finally, a multi-step prediction model with the spatio-temporal characteristics of D-stacking multi-integration fusion is designed for rolling multi-step prediction of wind-farm cluster power to obtain high-precision target wind-farm power. Through the actual wind power generation measured in northwest China to conduct case studies and comparative tests, it concludes that：(1) Spatio-temporal method in this paper can effectively extract the deep spatial features of wind farm clusters. (2) Dual fusion strategy improves Stacking effectively. (3) The proposed model can obtain accurate wind power prediction results, which is superior to 14 comparative algorithms proposed by other researchers.

Wind farm control and power curve optimization using induction-based wake model

This paper proposes a control strategy to achieve minimum wake-induced power losses in a wind farm. At first, the axial-induction-based wake model is developed to consider the aerodynamic wake interactions among wind turbines. To optimize the generated power of the whole wind farm, the axial induction factor of each wind turbine is calculated by the genetic algorithm. As a supervisory controller, each wind turbine’s optimal axial induction factor calculated by the genetic algorithm is implemented as a setpoint of each wind turbine’s internal controller. In the internal control loop, a comprehensive controller is designed to track the commanded axial induction factor. In the partial load region, the commanded axial induction factor was attained by tuning the generator torque. In the transient and full load regions, the blade pitch angle is tuned to keep the generator speed and torque at the rated values. The performance of the proposed control strategy is investigated through case studies, including three different wind speeds and a time-varying wind speed case in a 3 × 3 wind-farm layout. The simulation results show the satisfactory performance of the proposed approach.

Study on Mitigation of Wake Interference by Combined Control of Yaw Misalignment and Pitch

Yaw misalignment can make a wake steer, which is an effective method to increase the power of wind farms but it also increases the fatigue load of the turbines. In this paper, the combination of yaw offset and pitch control (CYMP) is studied to analyse the potential mitigation of wake, focusing on the wind velocity and turbulence of the wake distribution, power increment, and fatigue load reduction. The simulation case study shows that the method of CYMP can reduce the fatigue load by 10.29% and increase the total power by 1.7% compared with only wake steering in FAST.Fram. The Collaborative MPC (CoMPC) method based on CYMP is proposed to the real-time wake control in this research, which can increase power by more than 2% and reduce thrust by more than 4% than greedy control under 10 m/s turbulent wind.