The rapid development of wind generation technology has boosted types of the new topology wind turbines. Among the recently invented new wind turbines, the front-end speed regulated (FSR) wind turbine has attracted a lot of attention. Unlike conventional wind turbine, the speed regulation of the FSR machines is realized by adjusting the guide vane angle of a hydraulic torque converter, which is converterless and much more grid-friendly as the electrically excited synchronous generator (EESG) is also adopted. Therefore, the drive chain control of the wind turbine owns the top priority. To ensure that the FSR wind turbine performs as a general synchronous generator, this paper firstly modeled the drive chain and then proposed to use the variable-universe fuzzy approach for the drive chain control. It helps the wind generator operate in a synchronous speed and outperform other types of wind turbines. The multipopulation genetic algorithm (MPGA) is adopted to intelligently optimize the parameters of the expansion factor of the designed variable-universe fuzzy controller (VUFC). The optimized VUFC is applied to the speed control of the drive chain of the FSR wind turbine, which effectively solves the contradiction between the low precision of the fuzzy controller and the number of rules in the fuzzy control and the control accuracy. Finally, the main shaft speed of the FSR wind turbine can reach a steady-state value around 1500 rpm. The response time of the results derived using VUFC, compared with that derived from a neural network controller, is only less than 0.5 second and there is no overshoot. The case study with the real machine parameter verifies the effectiveness of the proposal and results compared with conventional neural network controller, proving its outperformance.
The recent progress in wind energy generation has advanced the ever-increasing development of large-scale wind generation. As a complex electromechanical system that converts wind energy into electricity, wind turbines are developing in the way of high power, high efficiency, high reliability, and low cost. As the wind is changeable and unpredictable due to its randomness, current studies mainly focus on how to convert such stochastic energy into power electricity smoothly [
Compared with the CSCF type wind turbines, the wind turbines functioning in a VSCF model can behave more efficiently since they can operate in a maximum power point tracking (MPPT) mode. Since both of the CSCF units and the VSCF units connect to the grid through a back-to-back converter, power quality becomes an important issue to them, which is more urgent to the VSCF type wind turbines since they still cannot meet some of the grid connection requirements automatically, e.g., the low-voltage ride through (LVRT) capability. From such aspect, the CSCF wind turbines have certain advantages compared with the VSCF units as additional facilities like LVRT module, reactive power compensation devices, and filters are not necessarily required. In order to make use of the advantages of both the VSCF type wind turbines and the CSCF type units, thus letting the wind turbine work in a gird-friendly way like traditional thermal power, hydropower, or nuclear power with fewer harmonics and high power quality, great efforts have been made [
However, the intrinsic drawbacks of traditional VSCF wind turbines are obvious since converters are necessarily required by either the PMSG or the DFIG for integrating into the power grid. For DFIG, its converter power is generally one-third of its ratings, which makes the cost of the machine reduce to a certain extent. But it also makes the generator more sensitive to the fluctuation of the power grid, i.e.; once the voltage of grid-side fluctuates, the DFIG is prone to trip and disconnect from the grid. In terms of PMSG, a full-scale power converter is adopted, which makes it no longer sensitive to the voltage fluctuation of the grid-side. But the total cost of the machine, in comparison, is much higher.
For the sake of minimizing the interferences of harmonics generated by voltage source converters on the power grid to the greatest extent, current source based new converter topology and kinds of filters are proposed [
Although types of control methods are proposed, as mentioned before, drawbacks introduced by the converter in wind turbines are inevitable and costly, and the asynchronization of the stator and rotor speed asynchronous generator is also a deficiency compared with the electrically excited synchronous generator (EESG). The control of such machines is realized through, without exception, a converter, which can be seen as a “Back-End” mode. In front of that, prototypes of “Front-End” controlled wind turbines with EESG are proposed and some of them are even converterless [
Schematic of front-end speed regulated wind turbines.
As the converter is reduced, speed regulation of the EESG in Figure
The rest of this paper is organized as follows: Section
As shown in Figure
Drive chain of the FSR wind turbine with a hydraulic torque converter (red in the figure).
In the beginning, the wind wheel is driven by the wind at a speed of
For WinDrive, as it can achieve either continuous adjustment of the speed of the main shaft or regulation of the torque in the drive chain, once there appears a sudden change in wind speed, it can release part of the energy stored to increase the damping of the drive chain to smooth the torque. Therefore, the modeling of the drive chain of FSR wind turbine will consider both the speed transfer and the torque delivery.
By neglecting the inertial, the input and output relationship of the gearbox can be described as
The dynamical balance equation of the WinDrive can be formulated as
Ignoring the moment of inertia of high-speed shaft, its dynamic equation can be summarized as
Based on the relationship between rotation speed and torque, the following equation can be obtained from [
In the ring of outer gear, the torque delivered by two rings are equivalent, namely,
From the above relationships, the following equation can be obtained:
Considering that the MPGA algorithm is more stable and faster for continuous function optimization compared with the standard genetic algorithm (SGA) or particle swarm optimization (PSO) algorithm, which is consequently introduced to optimize the parameters of the extension factor of the variable-universe fuzzy controller (VUFC), in the iteration of SGA, real coding, multipopulation, and multiobjective parallel searching are used for chromosomes, and the minimum preserving algebra of the optimal individual is used as the termination judgment of the algorithm.
In this paper, the designed optimal VUFC was applied to the drive chain for the purpose of speed regulation and torque control of the FSR wind turbine. The objective function is formulated according to the performance of the speed loop. Adaptive control of drive chain is hence realized by adjusting the basic universe adaptively, which thus helps to improve its dynamic and steady performance.
In the design of VUFC, the speed control of drive chain can improve its accuracy according to the variation of expansion factor. The expansion factor, in this sense, can be defined as adjusting the domain of each language control variable according to the value of the current control index.
According to the analysis of two commonly used structures of the expansion factor in [
The SGA is a highly parallel, stochastic, and adaptive global optimization probabilistic search algorithm developed from natural selection and evolutionary mechanism of biology, which has strong robustness and global search ability as it does not depend on the gradient. However, with the wide application of SGA, its problem of premature convergence is gradually exposed. It is mainly manifested in the fact that all individuals of the population tend to the same state and stop evolving. As a result, the algorithm cannot give a satisfactory solution. Based on that, this paper proposes to use MPGA to optimize the expansion factor. The schematic of the MPGA algorithm is shown in Figure
Schematic diagram of the MPGA algorithm.
In Figure
The VUFCs accomplish the expansion of their basic domains by corresponding expansion factors. The expansion of basic domains is equivalent to the adjustment of the control rules. The adaptive law is reflected in the expansion factors
Schematic diagram of the variable universe.
The idea of using MPGA to optimize the VUFC is to let the population, based on selected performance indicators, find the optimal parameters of each expansion factor in each sampling period intelligently relying on the mechanism of global search and local search. After several iterations, the optimal solution of the problem satisfying the performance indicators is thus obtained, which is then taken as the parameter of the expansion factor of the VUFC at the next sampling time. Based on variable-universe theory, an MPGA based variable-universe dual input and single output fuzzy control system for drive chain control of the FSR wind turbine is designed, which is shown in Figure
Diagram of variable-universe adaptive fuzzy control system based on MPGA.
In Figure
Based on the structure given in Figure
The adaptive function of individual adaptive value is compiled and calculated according to the objective function, which is given as follows.
From (
Schematic diagram of the MPGA algorithm.
To realize the proposed VUFC optimized by MPGA, the dynamic model of the drive chain, together with the aerodynamics model of the wind turbine and the speed relationship between wind wheel and planetary gears, is implemented in MATLAB/Simulink. Parameters of the wind turbine and drive chain are given in Table
Parameters of the wind turbine and its drive chain used for simulation.
R (m) | | | | | | | | | |
| |||||||||
45.3 | 1.29 | 3 | 875000 | 10 | 10 | 80 | 300 | 1380 | 3.75 |
| |||||||||
Tooth number of the gear ring | Tooth number of solar wheel | Speed Ratio of WinDrive | |||||||
| |||||||||
161/160 | 42/74 | 468/1500 |
When implementing the proposed VUFC for drive chain in MATLAB/Simulink using the parameters provided in Table
Performance of the proposed VUFC.
As shown in Figure
As shown in Figure
For the purpose of comparison, the simulation results using neural network control in [
Drive chain performance comparison between the proposed VUFC and the neural network controller.
As shown in Figure
By comparing the VUFC and the neural network controller, it is clear that the VUFC is more feasible. With the VUFC introduced, the guide vane angle of the hydraulic torque converter can be well adjusted and the output torque of hydraulic torque converter is changed accordingly. The output speed in the control process can adapt to the uncertainty of the input speed, when the VUFC is adopted.
Based on the above analysis, it can be concluded that the simulation model can have a better response to the variation of wind speed, and the wind energy can be captured to the greatest extent when the FSR wind generator operates in a steady state. The speed of the main shaft or the output speed of the WinDrive can be stabilized near 1500 rpm within only 0.5 seconds, which verifies the outperformance of the VUFC.
In this study, the optimized control of drive chain of the FSR wind turbine is realized through a variable-universe fuzzy controller by modeling speed relationship between the wind wheel and the planetary gears, the aerodynamics of the wind turbine, and the dynamics of the drive chain. Compared with the traditional neural network control algorithm, the drive chain control based on multipopulation genetic algorithm improves the control accuracy and speed to a certain extent and ensures reliable operation and quick response of the FSR wind turbine. For the VUFC, a response time less than 0.5 seconds can be achieved to ensure that the main shaft can rotate around 1500 rpm, which is much shorter than a 3-second response utilized with a neural network controller.
Wind speed
Structure parameter of the first planetary gear set
Structural parameters of the second planetary gear set
Turbine speed
Constant speed
Guide vane angle
Solar wheel speed
Speed of planetary gear frame
Speed of outer gear ring
Output speed of torque converter
Wind wheel speed
Radius of wind wheel
Mechanical power
Blade angle of the wind wheel
Power coefficient
Tip speed ratio
Air density
Inertial of the wheel
Coefficient of the low-speed shaft
Gear ratio
Damping coefficient of the high-speed shaft
First or second gear set
Coefficient
Pump torque
Turbine torque
Torque converter speed ratio
Density of oil filled in torque converter
Torque coefficient of the torque converter pump
Speed of the torque converter pump
Diameter of the circular circle
Gravity acceleration
Opening coefficients of the guide vane
Optimization parameters
Control variable
Rising time
Weights of the objective function.
The data used to support the findings of this study are available from the corresponding author upon request.
The authors declare that they have no conflicts of interest.
This work is financially supported by Scientific Research Projects of Colleges and Universities in Gansu Province (2016b-032).