Nội dung

Journal of Computer Science and Cybernetics, V.30, N.4 (2014), 313–334 DOI: 10.15625/1813-9663/30/4/5762 REVIEW PAPER GENERAL OVERVIEW OF CONTROL PROBLEMS IN WIND POWER PLANTS NGUYEN PHUNG QUANG Institute for Control Engineering and Automation, Hanoi University of Science and Technology; quang.nguyenphung@hust.edu.vn Abstract. Wind power plants can be realized with different generator types using different control principles. The choice of the generator regardless of control method, potentially destabilizes the grid, and can even lead to grid collapse. For independent grid (e.g. on islands) this risk is especially great. The report aimed at giving the reader a general overview of the control methods, and the developers a better understanding of each generator type to get the right choice for their wind power project. Keywords. Wind power plant, DFIG, PMG, front-end converter, generator-side converter, grid voltage oriented control, linear control, exact linearization, flatness-based control Abbreviations DFIG DPC DTC ESS FC GC GVOC Doubly-fed Induction Generator Direct Power Control Direct Torque Control Energy Storage System Frontend Converter Generator-side Converter Grid Voltage Oriented Control 1. IG LLDG MPPT PMG SCADA WPP WT Induction Generator Low-Load Diesel Generator Maximum Power Point Tracking Permanentmagnet Excited Generator Supervisory Control and Data Acquisition Wind Power Plant Wind Turbine INTRODUCTION Currently the exploitation of wind energy receives increasing attention from the society in Vietnam. Many projects have been carried out, in parallel with both (more or less) successful and not yet successful results. The weaknesses that make exploitation of such systems more difficult are caused by insufficient understanding of the operating principles, especially the principle of control. Even the projects with (more or less) success also contain potential long-term risks to the national grid. On the one hand the paper presents an overview of the control methods in WPP system, on the other hand it points out the mistakes susceptible in WPP projects in Vietnam. We know, energy can be extracted from the wind (Figure 1, [1]) by the following formula: 1 3 P = ρw A vw C (λ, β) , 2 (1) where P : power; ρw : density of air; A: swept areas of blades; vw : wind speed; λ: ratio of the rotational speed of the turbine to wind speed; β : angle of rotor blades c 2014 Vietnam Academy of Science & Technology in common, that is, the coefficient can always reflected byspeed a class of of power curves, which C
M, C  Then, despite the fluctuation of the wind. the berotational the turbine would have to change are identical have the more form as in Figuredue 2. These characteristic are kept constantly and in theprinciple controland becomes difficult to large inertia ofcurves the rotor blades [1]. confidential by manufacturers and stored in a look-up table to control turbines. Characteristic curves in Figure 2 show: Each wind speed curve has a point with maximum capacity to exploit P. Therefore, if the consumer (the grid) is able to accept unlimited P, the control system is responsible for changing the turbine rotational speed (the working point) to reach and to maintain maximum power point. However, if the turbine is only permitted to generate a capacity of P = const, 314 NGUYEN PHUNG despite the fluctuation of the wind. Then, the rotational speed of theQUANG turbine would have to change constantly and the control becomes more difficult due to large inertia of the rotor blades [1]. Figure 1: Exploiting thepower power from Figure 2: Characteristic curves for power extraction from wind Figure 1: Exploiting the from wind wind Figure 2: Characteristic curves for power extraction from wind Figure 2: Characteristic curves for power extraction from turbines 1: Exploiting the power Figure turbines wind from wind turbines 2 CONTROL HIERARCHY HIERARCHY In formula (1), C (λ, β) is2 theCONTROL coefficient reflecting the characteristics (the ability to exploit energy) of wind turbines. This coefficient is also the secret of the manufacturer, making up the 2.1 Operating modes of wind turbines difference between the turbines of different manufacturers. However, all types of turbines always have 2.1 WeOperating modes of wind turbines can distinguish two modes and therefrom control modes wind power one thing in common, that ofis,operation, the coefficient C (λ,theβ)twocan always beofreflected by a class of power generation systems. curves, which are identical in principle and have the form as in Figure 2. These characteristic curves Weare can distinguish ofgrid operation, and therefrom thetable two tocontrol of wind power 2.1.1 Operating modetwo with the national kept confidential bymodes manufacturers and stored in a look-up controlmodes turbines. . generation systems Characteristic curves in Figure 2 show: Each wind speed curve has a point with maximum capacity This operating mode is characterized as follows: to exploit P . Therefore, thenational consumer 2.1.1 Operating mode withifthe grid(the grid) is able to accept unlimited P , the control system is responsible for changing the turbine rotational speed (the working point) to reach and to maintain maximum powermode point.isHowever, if the as turbine is only permitted to generate a capacity of P = const, This operating characterized follows: despite the fluctuation of the wind. Then, the rotational speed of the turbine would have to change constantly and the control becomes more difficult due to large inertia of the rotor blades [1]. 2. 2.1. CONTROL HIERARCHY Operating modes of wind turbines We can distinguish two modes of operation, and therefrom the two control modes of wind power generation systems. 2.1.1. Operating mode with the national grid This operating mode is characterized as follows: • The national grid can be seen as hard grid with extremely large P , with stable voltage and frequency. GENERAL OVERVIEW OF CONTROL PROBLEMS IN WIND POWER PLANTS P315 • The active power is controlled following the curve with optimal power (Figure 2), to extract maximum power from the wind. • The power factor cos ϕ is often fixed by value nearly 1. That means the WPP will neither Q generate nor consume a reactive power Q. 2.1.2 Independent operating mode without the national grid 2.1.2. Independent operating mode without the national grid Specific examples for this operating mode are WPPs on islands with following characteristics: wind based hybrid power systems P • Local grids are built by a group of diesel generators with small active power P . These are the so called wind based hybrid power systems. P • Local grids are soft grid whose voltage and frequency are unstable. • The load is divided between the group of diesel generators and the WPP. The WPP may generate only a fixed active power P = const (Figure 2) specified by the rate of distribution. • The power factor cos ϕ of WTs should be set flexibly in the appropriate value to ensure safe 2.2 Control hierarchy of a WPP and efficient exploitation of the diesel generators. Regardless of the used type of generator, the control system of a WPP is always structured by a 3level hierarchy as in Figure 3. 2.2. Control hierarchy of a WPP Regardless of the used type of generator, the control system of a WPP is always structured by a 3-level hierarchy as in Figure 3. 2.2.1. Control level I This control level has the task of a SCADA system serving the goal of WPP integration with the grid (national, local). Dependent on the operation mode this level decides the set points for P and Q. For large-scale systems (wind park ), the level plays the role of the supervisory control equipped with the ability to communicate between members of wind park and the dispatching center. With the characteristics of a SCADA system, on this level we can specify our principles of energy management. 2.2.2. Figure 3: Control hierarchy of wind power plants Figure 3: Control hierarchy of wind power plants Control level II This level realizes the task of turbine control with a feedback closed loop for the turbine rotor speed ω . Based on the measured wind speed vwind and on the pre-selected operating mode, the system uses a look-up table to find the set points for the rotor speed ω which can be controlled by varying the blade pitch angle β . There are two things to note: • In operating mode with extraction of maximum wind power the system uses a MPPT algorithm to reach the rotor speed ω on the top of the wind characteristics (Figure 2) dependently on the measured wind speed vwind . MPPT algorithm is always a secret of turbine manufacturers, and users do not have the opportunity to intervene at this stage. 316 NGUYEN PHUNG QUANG • Rotor and rotor system weigh many tons, resulting in a huge moment of inertia which limits the dynamic control of the blade pitch angle β in both operating modes P = const or P = max (Figure 2). 2.2.3. Control level III This control level contains the real-time algorithms of the generator control structure to control the flows of active power P (electric torque mG ) and reactive power Q (power factor cos ϕ), fulfilling the demands of the level I. To control P and Q, the system uses a back-to-back converter with two parts GC and FC. The implemented control methods depend on: • the type of the generator, and • the operating mode (connected to the national or local grid). It can be confirmed that the level III is responsible for the control system of WPP (characterized by rapid dynamics and small inertia, small sampling periods and small modulation periods), which is connected with grids (characterized by slow dynamics and large inertia), is really a challenge for investors. The incomplete understanding of this level is the potential risks mentioned from the beginning of the paper. 3. 3.1. CONTROL PROBLEMS OF THE LEVEL III Overview about control of generators Figure 4 gives an overview of the control problems for generator types IG, DFIG or PMG ( [2–4]) used in WPP. It can be seen: • In the case DFIG : Because the back-to-back converter is located on the side of rotor circuit (not between the stator and grid like the cases IG and PMG), the power electronic converter must only be sized with nearly 1/3 power of the generator. The cost of systems using DFIGs is always lower than the cost of systems with PMGs. • In the cases IM, PMG : Because the back-to-back converter is located between the stator and grid, the system cost is higher than the cost of DFIG systems, but easier to control. We can divide the generator control problems into 2 groups: FC control and GC control with a lot of issues that need to be addressed, but not possible to be introduced in the limited framework of this paper. Depending on the type of generator DFIG or IG/PMG, the group of GC control can also be split into different solutions. 3.1.1. FC control The control problems of this group are basically the same in all three cases IG, DFIG and PMG. It can be summarized as follows ( [3, 7]): • The main method is the GVOC. Some works have tested the method DPC inspired by the DTC of electric three-phase AC drives. • The control must ensure the decoupling between P and Q, as well as the flexible setting of cos ϕ. It only needs a linear control structure [7]. P Q • The control must satisfy the regulations of the grid harmonics. In some cases the FC control GENERAL OVERVIEW OF CONTROL PROBLEMS IN WIND POWER PLANTS 317 can be extended by an active filter function. Figure 4: Overview of the control problems for generator types DFIG, IG and PMG Figure 4: Overview of the control problems for generator types DFIG, IG and PMG 3.1.2 GC control in the case DFIG Because the stator of DFIG is directly connected to the grid, therefore this is the case with most • regarding The control must satisfy the regulations of the grid harmonics. In some cases the FC control challenge to the generator control. can be extended by an active filter function. • The main method is the GVOC. 3.1.2. GC control in the case DFIG P Q mG Because the stator of DFIG is directly connected to the grid, therefore this is the case with most challenge regarding to the generator control. • The main method is the GVOC. • The control must ensure the decoupling between P and Q (decoupling between mG and cos ϕ), as well as the flexible setting of cos ϕ. • The control structure can be either linear or nonlinear. • Crowbar control. 3.1.3. GC control in the cases IG, PMG In practice, the generator type IG is no longer used. Currently we can not find on the market this generator type used by turbine manufacturers, but only PMG. For PMG, there are 2 possible solutions for GC as follows: • GC is a simple non-controlled rectifier : In this case following characteristics are to note. + The amount of the input energy on the primary side (wind energy) is decided only by the turbine control system (control of rotor speed ω ). The input energy must be totally transferred to the grid. 318 NGUYEN PHUNG QUANG + A DC-DC boost converter must be used on the DC link to increase the magnitude of the DC voltage to the level of the FC input. • GC is a controlled rectifier : + In combination with the turbine control, the GC can effectively control the energy flow from the primary side. The often used principle is the pole flux oriented control. + Decoupling control between the electric torque mG and the pole flux ψ p . + The control structure can be either linear or nonlinear. 3.1.4. Related control problems for both groups FC and GC Beside separate control problems only for GC or FC, there are a lot of control task related to the complete system WPP: • Fulfilling the grid code (more in section 4): While symmetrical or nonsymmetrical voltage dips, the WTs should be able to stay on grid and to handle without disconnection. • The standard output voltage of WTs is 690V AC. The wind turbines must be equipped with 690V/22kV (or 690V/110kV) transformer to synchronize with the grid. This leads to problems to be solved: – The neutral-point voltage of the primary side varies. A neutral-point voltage control must be implemented to ensure that the DC current is zero. – Common-mode voltage stress : The switching action of the rectifier and inverter normally generates common-mode voltages which are essentially zero-sequence voltages superimposed with switching noise. This voltage is very harmful for the winding insulation, and causes ignition through the parasitic capacitance Cp which reduces the life of bearings. • Equipments supporting to stabilize the grid voltage in case of independent operating mode. The equipments can be either energy storage systems or low-load diesel generators. – The ESSs with the ability to charge or discharge energy very quickly can help to stabilize the grid voltage while wind fluctuation. – The main task of LLDGs is the generation of reactive power Q, and therefore the generation of the grid for WTs with DFIG. This equipment is not able to compensate the wind fluctuation. 3.2. Main difference between the DFIG and PMG control To illustrate the difference between 2 generator types, we should begin with DFIG in Figure 5a. Because the stator of DFIG is directly connected with the grid, the turbine rotor speed ω as well as the rotational speed of the DFIG is bounded in the range ±33% compared with the synchronous speed. A brief statement of the main points of DFIG control can be given as follows [7]: • Range of oversynchronous speed : Figure 5a illustrates clearly that in this range the energy, flowing through the rotor winding, is generated by the wind. That means, the reactive power Q is generated by the wind. grid frequency, and this fact allows the power extraction from wind in a relatively wide range. • WPP using PMGs is the only type of system, which can operate independently without grid 319 (national GENERAL and local).OVERVIEW OF CONTROL PROBLEMS IN WIND POWER PLANTS a) b) Figure 5: Regarding to the rotational speed: a) the control of DFIG is dependent on range of oversynchronous or Figure 5: Regarding to the rotational speed: a) the control of DFIG is dependent on range of oversynchronous or subsynchronous speed, and b) the control of PMG is nearly unlimited • Synchronous speed : At the point of synchronous speed, corresponding to the stator frequency 50Hz, the rotor frequency is zero. A DC current flows in the rotor circuit. Special attention for this point in control concept is necessary, to avoid damages of the rotor. • Range of subsynchronous speed : In this range Q is supplied by the grid. This is the main disadvantage, which limits the use of DFIGs for WPPs on islands. In contrast to DFIG, Figure 5b shows that: • PMGs are excited by permanent magnets and do not consume reactive power Q. • The stator of PMGs is not directly connected with the grid. The rotor speed is not dependent on grid frequency, and this fact allows the power extraction from wind in a relatively wide range. • WPP using PMGs is the only type of system, which can operate independently without grid (national and local). 4. CONTROL DURING GRID FAULTS Previously, in order to protect itself when grid faults occur, the control system may separate WTs out off the grid. In recent years, the exploitation of wind power has reached the scale of the plant (wind parks ); WPPs separation out off the grid potentially causes local oscillations. These local oscillations can spread out and lead to the risk of grid collapse. ride through 320 grid cod NGUYEN PHUNG QUANG To prevent this negative scenario, many countries have made regulations which strictly prohibit increases steadily 90% The (the WTs allowed low level). In the whole process of grid fau the separation out off the grid in some cases ofback gridtofaults. must have the ability to “ride 3000ms (150 grid cycles), the WPP is not allowed to separate itself out off through ” during grid faults ( [18–22]), and must be able to generate reactive power Q for supporting the grid. The W be able the so to control its output voltage exactly follows the grid voltage. During t the grid stability as well as must for avoiding spread of that voltage oscillations. control process the generation of active power P is not necessary. Fulfilling The term “grid code”the grid code is the required condition for grid connection of WPPs. At begin this iss has created new challenges for control design. In recent times this issue has been investigated also thetoHUST intensively. The mentioned ability “ride very through ” 4.1. during grid faults is standardized by the term “grid code ” illustrated in Figure 6, in which the definition of the group E.On Netz (Germany) is clearly explained. Here in words: The grid voltage amplitude suddenly drops from 100% to 15% of the nominal level. The level 15% maintains approximately 500ms (25 grid cycles), then the grid voltage recovery increases steadily back to 90% (the allowed low level). In the whole process of grid faults 3000ms (150 grid cycles), the WPP is not allowed to separate itself out off the grid. The WPP must be ableFigure so 6: The ability “ride through” is defined by the term “grid code” ([21]) Figure 6: The ability “ride through” is defined by the term to control that its output voltage ex“grid code” ([21]) 4.2 voltage. WPP During control with grid tracking actly follows the grid this control process the generation of active power P is not necessary. Fulfilling the grid code is the required condition for grid connection of WPPs. At begin this issue has created new challenges for control design. In recent times this issue has been investigated also at the HUST very intensively. 4.2. WPP control with grid tracking In section 3, all control problems in WPPs are listed. The main challenge for manufacturers is to find a solution for both problems to design the control structure and to fulfill the grid code (section 4.1). This is particularly difficult for the system using DFIG, and it can be confirmed: Not any commercial DFIG system on the market can meet this requirement. To visualize the level of difficult or easy to meet the requirement ”grid code” between the generator types DFIG and PMG, we only have to take a closer look for Figure 7. • Because the stator of PMG is not directly connected to the grid, it will be relatively easy the FC (a DC-AC converter) so to control that its output voltage exactly follows the grid voltage during grid faults, as required by grid code. • Because the stator of DFIG is directly connected to the grid, the control efforts from the rotor side only have indirect effects. In addition, when the grid voltage is suddenly decreased, stator of PMG is not directly connected to the grid (a DC-AC converter) so to control that its output voltage exactly follows the grid voltage during grid faults, as required by grid code. • Because the stator of DFIG is directly connected to the grid, the control efforts from the rotor side only have indirect effects. In addition, when the grid voltage is suddenly decreased, the DFIG operation will change into the nonlinear operating mode. These are the two main causes OVERVIEW OF CONTROL PROBLEMS IN WIND POWER PLANTS ofGENERAL difficulty in case DFIG control to meet the grid code. 321 Figure 7: DFIG so to control that WPP fulfills the grid code is much more difficult than PMG control Figure 7: DFIG so to control that WPP fulfills the grid code is much more difficult than PMG control 5 CONTROL STRUCTURES FOR DFIG The preceding sections have highlighted the difficult problems of DFIG control. This section presents the DFIG operation will change into the nonlinear operating mode. These are the two main causes of difficulty in case DFIG control to meet the grid code. 5. CONTROL STRUCTURES FOR DFIG The preceding sections have highlighted the difficult problems of DFIG control. This section presents the investigation results of recent years to overcome this. The most implemented principle is the grid voltage oriented control thereby the d-axis (the real axis) is the axis of the grid voltage vector (Figure 8). Starting from the following machine equations (2):  us = Rs is + ur = Rr ir + dψs dt dψr dt + jωs ψs + jωr ψr (2) The state space model of DFIG in the grid voltage oriented reference frame (3) will be obtained as follows [7]: dx = A x + Bs us + Br ur dt (3) 322 NGUYEN PHUNG QUANG where:  − σ1  1 Tr   −ω r A =  1  T s 0  1−σ − σLm  0 Bs =   1 With: state vector xT + 1−σ Ts  ωr − σ1  1 Tr + 1−σ Ts 0 1 Ts  1−σ σTs 1−σ σ ω − T1s −ωs − 1−σ σ ω  1−σ σTs   ;   ωs − T1s  1 0 σLr 0 1−σ  1   0 − σL σLr  m  ; B  = r   0  0 0 Lm 1 0 0 0 Lm h i / / = ird , irq , ψsd , ψsq ; stator voltage vector uTs = [usd , usq ] as input   vector on stator side; rotor voltage vector uTr = [urd , urq ] as input vector on rotor side. The used symbols in system matrix A, rotor-side input matrix Br and stator-side input matrix Bs mean: Tr , Ts : time constants of rotor and stator circuit; Lm : mutual inductance; Lr : rotor-side inductance; ω : mechanical rotor angle speed; ω r , ω s : angle speed of rotor and stator circuit; σ : total leakage factor. Outgoing from the model (3) the following physical relations (4) can be easily derived, and then illustrated in Figure 8: sin ϕ = |ψs |/Lm − irq 3 Lm ; mG = − zp ψsq ird |is | 2 Ls (4) The main conclusion following the equation (4) is that the current component ird plays the role of torque control or active power control and the current irq is the reactive power forming component. This conclusion means that the most important control loop in the structure is the inner loop. The variety of the inner current loop extends from linear to nonlinear controller whose successful designs will be presented in the next sections. The outer loop normally contains two PIcontroller for active power P as well as reactive power Q or power factor cos ϕ. Figure 9 shows the control hardware of WPPs. 5.1. Linear control Since the two rotor current components ird , irq play the role of P and Q control variables an inner control loop to impress the rotor current vector is needed. The discrete model of the rotor current can be derived by iterative integration of the equation (3): ir (k + 1) = Φ11 ir (k) + Φ12 ψs0 (k) + Hs1 us (k) + Hr1 ur (k) (5) or in component form: ( / ird (k + 1) = Φ11 ird (k) + Φ12 irq (k) + Φ14 ψsq (k) + h11s usd (k) + h11r urd (k) / irq (k + 1) = −Φ12 ird (k) + Φ11 irq (k) + Φ13 ψsq (k) + h11r urq (k) (6) In equations (5) and (6) the stator flux and the stator voltage might be regarded as disturbances to be compensated by a feed-forward control on the one side. On the other side, these values are nearly constant and therefore can be compensated exactly and fast enough by the implicit integral part of the controller, so that their feed-forward compensation may be omitted.