Experimental Validation of Improved Control Strategy of Grid-interactive Power Converter for Wind Power System

In this paper grid-voltage oriented control of grid-side converter for wind power system is presented, which aims to optimize the performance of utility grid during transient conditions. To realize this objective, a digitally controlled inverter system is implemented which provides independent control of active and reactive power delivered to the grid under sudden change in reactive load. Initially, the proposed grid-voltage oriented control scheme is analyzed in order to manage the power flows into the grid followed by the mathematical modelling of grid-side converter. And then, a closed-loop control structure of the grid-side converter is implemented using a conventional proportional integral (PI) controller with space vector pulse width modulation (SVPWM) current controller. In order to analyze the results of the proposed study, in this paper, the complete system is modelled and simulated in the MATLAB/Simulink environment and the numerical results are discussed. The experimental results, obtained using DSP TMS320F2812 processor, confirm the high performance of the proposed control system in terms of good sinusoidal grid current, low harmonic content and fast dynamic response.


Introduction
Wind energy is the fastest growing and most promising renewable energy source among all renewable energy sources, due to economically viable and clean power production. Globally, about 20% penetration of renewable energy in electricity generation is considered necessary in the coming decade (by 2020) [1]. However, the stochastic nature of these sources, the controllability of the wind power system is relatively complex and it should be further investigated. As a consequence, their interconnection with the utility network can lead to grid abnormality or even failure, if these systems are not well controlled [2]. In general, wind turbines are expected to support the grid and to provide grid connection requirements to have a characteristics much like conventional power plants (e.g., active power control, frequency regulation and dynamic voltage control, and low voltage ride through (LVRT)) and they are generally achieved by grid-side converters (GSCs) with appropriate control [3].
Several generic control strategies for GSC has been developed [4][5][6][7][8][9][10]. Proper power flow regulation with grid current harmonic rejection using the complementary controller incorporating isolation transformer is discussed in [5]. Decoupled control of GSC for distributed generation systems is presented in [6] using reference frame theory for decoupling d-q components which shows the good dynamic response during load variations. In [7], synchronous proportional integral (PI) current control has also been proposed which transforms the three phase grid voltages to synchronously rotating (d-q) frame for proper decoupling. As a result, grid currents become DC variables and thus no steady state-state error adjustment is required. A method for active and reactive power control has been stated in [8] and it controls the DC-link voltage by designing a Voltage Control Loop. However, the transient time has not been taken into consideration during sudden change in load conditions. For improving the dynamic response in the dc-link voltage, an adaptive droop controller can be employed in the voltage controlled loop with optimal design of LCL filter on the grid side [9]. Still for simplicity of the decoupled grid power control, estimation of grid parameters is possible by feedback linearizing [10]. In Ref. [11], authors have proposed the predictive controller which uses the system model to predict the system behavior in each sampling interval for each voltage vector, and the most appropriate vector is then chosen according to an optimization criterion to fulfill flexible active and reactive power regulation in the grid systems. Simple vector control of grid-interactive power converter is studied in [12] and the author has proposed sigma-delta modulator current controller in the inner feedback loop for sinusoidal current injection into the grid. However, the current control strategies, based on the space vector PWM (SVPWM) is widely employed for threephase voltage source inverter (VSI), because it has constant switching frequency, well-defined harmonic spectrum, optimum switching patterns, and excellent DC-link utilization [13]. Yet, the performance of the current controllers based on SVPWM for grid -connected inverter is compromised by the grid harmonics, non-linearity in the system and lack of inherent over current protection. The design of current error compensation scheme is key factor to overcome the above drawbacks.
In this paper, Implementation of grid-side converter (GSC) control for wind power generation system is analysed based on SVPWM current controller. The proposed current control scheme employs two PI regulators in decoupled current error compensation tracks. With the well-arranged sampling timing, the control delay can be minimized along with lesser computational time. Firstly the control objectives of GSC are analyzed and the mathematical model is established in this study. Based on the mathematical model and the control strategy, the software design of the control strategy is implemented in the digital domain using DSP. Experimental results, obtained using DSP TMS320F2812 processor, are in agreement with the numerical simulations and confirm the high performance of the proposed control system in terms of good sinusoidal current, low harmonic content and a fast dynamic response.

Mathematical Modelling of Grid-Side Converter (GSC)
The basic configuration of the three-phase GSC is shown in Fig. 1 which consists of IGBT switches, line inductance, dc-link capacitor, and utility grid.
Assume that the three-phase grid is balanced with negligible distribution parameters and also the IGBT switches are ideal with negligible on-state voltage drop.
From Fig. 1, the dynamic equation of the output side of the GSC can be deduced as follows [8].
The corresponding input dc current equation is given as: The line voltages and the phase currents are transformed into d − q components using synchronous reference frame by the use of the transformation matrix;  Fig. 1 Schematic diagram of three-phase grid-side converter By assuming that the three phase voltage source is balanced without the zero sequence components we can write The s-domain representation can be shown as follows: For the input dc-side of the converter Now the instantaneous power equation is given as where * represents the complex conjugate, P (t) represents the instantaneous active power and Q(t) represents the instantaneous reactive power. Hence From the above (7) and (8). it can be seen that the dcomponent and the q-component of grid voltage are highly coupled which leads to the degradation of the dynamic performance. The proposed voltage oriented control scheme decouples these terms and thus provides the ability to control each current component independently.

Grid Voltage Orientated Control of Grid-Side Converter
In grid voltage oriented control, accurate field orientation for a GSC becomes simple since the grid flux position can be derived from the measurable grid voltages using hall sensors and a non-measurable grid flux becomes a space vector that defines a rotating grid-flux oriented reference frame. As shown in Fig. 2, the grid voltage vector is aligned with qaxis. Vector control regulates the length and position of the grid current vector in the grid voltage orientated reference frame. In this reference frame the d-component of current corresponds to reactive power and q-component of current corresponds to active power [10]. The reactive and active power can therefore be controlled independently, since the current components are orthogonal. First and foremost entity in grid-voltage oriented control is to use the feedback signals from the grid in order to generate the current references which intern control the active and reactive power independently. The advantage of grid-voltage oriented control of GSC is that they detect and instantaneously compensate for voltage unbalances at PCC by injecting leading or lagging reactive power at crucial junctures to the utility grid [11]. In addition, GSC can supply the reactive power to the grid with fast dynamics. Thus it helps in regulating the system voltage and stabilizing the grid.
The dc-link power is known by: Assuming that the inverter is lossless, we can equate the input power with the output power and hence the power balance becomes Assuming balanced case and taking the d-component of grid voltage to be zero, we get Figure 3 illustrates the schematic diagram of proposed GSC control and it was said earlier, that the elementary principle of the vector control method is to control the instantaneous active and reactive grid power which can be done by controlling the grid currents, by separate controllers independent of each other. Two current controllers are employed namely dcurrent controller andq-current controller. The grid voltages and currents are first sensed and with the help of synchronous reference frame (SRF) phase locked loop (PLL), the grid phase angle is detected in order to synchronize the converter output with utility grid. The demanded amount of current and voltage are then estimated from the grid at the desired power factor and the reference currents in a synchronous frame are calculated. Consequently, the current controller attempts to reduce the current error and makes the load current follow the reference current vector. As a result, controlling of current enunciates controlling the inherent power flow between the GSC and utility grid. The converter dc-link voltage is determined by [8] The DC voltage can be regulated using PI controller by choosing the current reference where V dcref is the reference DC link voltage, K p and K i are the constant gains of the PI controller. The reference currents can be calculated as: where are the active and reactive power references.
where v dref (t)and v qref (t) are the d and q voltage references. V d and V q are the effective voltage references.
where K p and K i are the constant gains of the PI controller. The obtained dq axis references are then applied to the current controllers to achieve the desired level of output power, such that the sudden variation in the active or reactive power demanded by grid netwrok appearing as step changes does not have a significant effect on the other one and possible controls on the active and reactive output power can be done independently.
The values of the gains K p and K i are calculated according to the formulae [14]: where τ = L/R is grid time constant; L is the coupling inductance between the GSC and the grid, R is the coupling resistance V dc is the DC-link voltage

Selection of Dc-link Capacitor
The dc-link voltage, V dc , consists of an average dccomponent, V dc(avg) , as well as six times frequency voltage ripple (300 Hz). An expression for the peak-to-peak double frequency voltage ripple, V dc , of the dc-link can be derived as (1) [15]. In (1) P g is the active power injected to the grid C dc is the dc-link capacitor and ω is the fundamental angular frequency of the grid voltage.
V dc(avg) is 420 V of the modelled grid-side converter system.
In the proposed system, C dc is calculated as 1500 μF to limit V dc approximately to 5% of V dc(avg) when GSC is injecting 4 kW of active power to the grid.

Space Vector Pulse Width Modulation (SVPWM) Current Controller
In the grid connected wind power system, we need threephase sinusoidal voltage source with controlled amplitude and phase shift, in order to control active and reactive power flows into the grid. There are different control techniques, Fig. 9 Simplified structure of SRF PLL  Fig. 4), imposing the output voltage with desired amplitude and desired phase shift. Current control is done in synchronous rotating reference frame where sinusoidal AC variables, like V α and V β , become DC quantities, i.e. d-and q-components. SVPWM module gives full control to the linear PI current regulators over the output voltage d-and q-components. These d-axis and q-axis components, defined on the current controllers outputs (see Fig. 4), are transformed back to the stationary α-β reference frame and passed through to the space vector modulator. As it could be seen in Fig. 5, the eight possible states of an inverter are represented as two null-vectors (V 0 , V 7 ) and six active-state vectors forming a hexagon (V 1 -V 6 ). SVPWM approximates the rotating reference vector in each switching cycle by switching between the two nearest active-state vectors and the null-vectors. The main task of space vector modulator is to calculate the needed vector times directly from

PI Controller
where T k represents half of the vector V k on-time (dutycycle) in switching period T s .  (3) into its real and imaginary part, and after rearranging it follows: The corresponding PWM pulse generation in Sector "Introduction" is given in Fig. 6. After T 1 , T 2 and T 0 are found out; the three-phase PWM pulses are generated by one of symmetrical methods. This method makes each switching component switch once, in one carrier period, bringing all of them to a fixed switching frequency. With the appropriate placement of zero vectors, the entire voltage vector is split into ripple frequency to the double of switching frequency. Thus SVPWM current controller in the synchronous reference frame has clearly alienated current error compensation and PWM pulse generation, only if the system nonlinearity due to the control delay and the switching dead time can be well controlled along with the back-EMF disturbance. Moreover, the back-EMF disturbance of the converter which is the grid harmonics can be well compensated since the grid voltage needs to be detected in real time for system protection purpose [16]. This makes it possible to exploit the advantages of SPWM as well as to independently design the overall control structure. The proposed current control scheme employs two conventional PI regulators in decoupled current error compensation tracks. With the wellarranged sampling timing, the control delay can be minimized along with lesser computational time. Furthermore, the feed-forward grid voltage compensation can effectively eliminate the grid harmonics disturbances and high quality sinusoidal current can be injected into the grid with very low THD.

Synchronous Reference Frame (SRF) PLL for Phase Estimation
The basic structure of three-phase SRF PLL is illustrated in Fig. 7. To obtain the phase information, the three phase (V a ,V b and V c ) grid voltages are transformed into two phases (V α and V ß ) by using Clark's transformation and these two phases are transferred into direct and quadrature(dq) axis by using Park transformation. The phase angle θ is tracked by synchronously rotating voltage space vector along q or d axis by using PI controller [17].
The corresponding voltage space vector synchronous with the q-axis is depicted in Fig. 8. The transformation matrix of the voltage phase vector synchronised with q-axis is where θ * is the estimated phase angle of the PLL system. Carrying out the transformation using V qd = T qd V αβ , yields By applying matrix multiplication and trigonometric formule, we get (6), The phase angle θ is estimated with θ * which is integral of the estimated frequency ω * .The estimated frequency is the sum of the PI controller output and feed forward frequency ω ff . The gain of the PI controller is designed such that, V d follows the reference value V * d = 0 as in Fig. 9. If V d = 0, then the space vector voltage is synchronized along the qaxis .and estimated frequency ω * is locked on the system frequency ω. So that the estimated phase angle θ * is equals to the phase angle θ .

Results and Discussion
The proposed control strategy of grid-side converter is verified by the computer simulation based on MATLAB -7.6 and realized by Simulink model. The parameters for the GSC are listed in Table 1. The evaluation of the controller is made in two situations, i.e., steady-state and transient operations. In steady-state operation, the quality of the controlled current is presented while in transient state, the controller response to a sudden reactive load change is studied. Figure 10 illustrates steady state response of proposed control algorithm for GSC. From the results it is perceived that, all the three-phase voltage, inverter current, grid current, are at steady values. The corresponding waveforms are of grid voltages (230 V) and grid currents (10 A) are illustrated in Fig. 10a and b respectively. The inverter and grid current waveforms are shown in Fig. 10c at the values of 20 A and   [18] 10 A respectively. As shown in Fig. 10d active power delivered to the grid is constant and equal to its reference value (4 kW) and Fig. 10e depicts that no reactive power is injected into the grid as the grid voltage at PCC is constant (230 V ph-ph) and hence q-axis component of grid current is zero as shown in Fig. 10f. Figure 11 shows the transient response of proposed control algorithm for grid-side converter. It is shown that for a load of (4 kW, 0 kVAr) the load current is 10 A and the grid current is 20.4 A. When a high inductive load of (4 kW, 7 kVAr) is switched on at t = 0.08 sec, the load current rises to 15 A and the grid current change instantaneously at the load change and reaches to a value of 25 A to compensate the reactive power at PCC. The grid current and load current settles to steady state values within less than a cycle and are shown in Fig. 11b and c respectively. Figure 11d shows that the active power is almost constant and equal to its input reference value (4 kW). As shown in Fig. 11e reactive power drawn from the grid is 4.2 kVAr during transient. At t = 0.14 sec, the load is removed, grid current and load current changes instantaneously to their previous values. But the voltage at PCC remains constant i.e. 230 V (ph-ph). Figure  11f shows the q-component (i.e. i qref ) of the inverter current under dynamic changes of the load reactive power. In this scheme the power command for the controller was set equal to the load power (4 kW). Therefore, there is no current fed to the load from the utility, the load nearly takes its power from wind energy system. When the load increases, the amount of increment in active power is drawn from the utility grid while the reactive power is compensated by the GSC. This controller controls the active power and reactive power independently, maintains active power at its reference value while compensating for the reactive power. It can be seen that in both cases, the transient time is less than 2 ms, which means that the current controller has an excellent dynamic response. Hence, the power control could be considered to be decoupled too, and then the decoupled system has been deduced. Nevertheless, due to measurement errors and circuit parameters uncertainty, it is practically not possible to have a fully decoupled system. Moreover, the grid output current is highly sinusoidal with a THD of 1.9% as shown in Fig. 12.

Experimental Validation and Discussion
The performance of the proposed control strategy is validated with the help of a scaled laboratory prototype of DFIG based wind power system which is shown in Fig. 13 using TMS320F2812 digital signal processor and code composer studio software. An overall block diagram of implemented  Fig. 14. The corresponding experimental set-up is given in Fig. 15. The real-time code of the control algorithm is generated using the Real-Time-Workshop toolbox in MATLAB-Simulink environment. All measured and control signals are adapted for development platform based on fixed-point digital signal processor TMS320F2812, which is used for control algorithm implementation. Developed GSC is connected to the grid through coupling inductance and transformer. DC voltage is derived from output of a three-phase IGBT-rectifier, and it was constant during all experiments in order to properly design the current controller. Tests for decoupled reactive power and active power control have been done in the laboratory. During experimental verification the circuit is simulated for time t = 0 to t = 0.12 sec instead of t = 0.2 sec. Figure 16 shows the steady state experimental results which are almost similar to the steady state simulation results. But there is a transient at t = 0.08 sec and t = 0.14 sec due to switching action of the breaker as the reactive power increases between these periods.
Next the transient response of proposed control strategy for GSC is discussed and it was shown in Fig. 17 that, the required leading current and lagging current are injected into the grid as demanded by the local loads. As stated above, when the reactive load increases, the amount of increment in active power is drawn from the utility grid while the reactive power is compensated by the GSC. This controller controls the active power and reactive power independently, maintains active power at its reference value while compensating the reactive power. As a result, through the dynamic VAR control system, reactive power is injected into in the grid with fast dynamics according to the reactive load demand. Thus it helps in regulating the system voltage and stabilizing the grid [19]. Further, the harmonic spectrum of the grid current is given in Fig. 18 and it can be seen that grid current show a THD of 1.96% which is very much within the permitted distortion levels as per IEEE 519-2014 standard [20]. Thus the experimental results confirm the good performances of the proposed control system that is capable of ensuring a sinusoidal output current with a very  fast dynamic response. Table 2 gives the comparative analysis of simulation and experimental results and the results are almost converged.

Conclusions
The paper presents the grid-voltage oriented control strategy of grid-side converter for wind power generation system. The performance of the proposed control strategy in this paper is validated with the help of a scaled laboratory prototype of DFIG based wind power system using TMS320F2812 digital signal processor. From the study, the proposed current control scheme with SVPWM controller employs two conventional PI regulators in decoupled current error compensation tracks. With the well-arranged sampling timing, the control delay can be minimized along with lesser computational time. As a result, the proposed control strategy provides excellent dynamic response and high quality sinusoidal current injected into the grid with very low THD. The feasibility and effectiveness of the proposed controller were proven through theoretical analysis, simulations, and experimental results.