Method and arrangement for tracking the maximum power point of a photovoltaic module

Abstract

 A method and apparatus of tracking the maximum power point of a photovoltaic module, the method comprising measuring output voltage (v_C) of the photovoltaic module, determining output voltage (e) of the inverter connected to the photovoltaic module, measuring output current (i₁) of the inverter connected to the photovoltaic module. The method further comprises defining a variable (z) relating to energy of the capacitor using the measured PV module output voltage, extracting second harmonic component (z̃) from the defined variable (z), estimating the second harmonic component (p̃) of the module output power (p) using the defined variable (z), measured output current (i₁) of the inverter and the determined output voltage (e) of the inverter, multiplying the obtained harmonic components with each other, extracting a DC component (〈p̃ z̃〉_DC) from the obtained product (p̃ z̃), and forming a control signal (P;v_ref) for controlling the inverter connected to the photovoltaic module by using the extracted DC component in a PI algorithm.

Description

FIELD OF THE INVENTION

[0001] The present invention relates to a method and an arrangement for tracking the maximum power point (MPP) of a photovoltaic cell or module. More specifically, the invention relates to a method and an arrangement with which the maximum power point can be tracked fast and accurately in single-phase photovoltaic applications without measurement of output current from the photovoltaic cell or module and in environments where the irradiation and temperature of the cell changes rapidly.

BACKGROUND OF THE INVENTION

[0002] It is known in the art of photovoltaic (PV) cells, modules or arrays to use a maximum power point tracker (MPPT), which tries to ensure that the maximum available power is extracted from the irradiated cells.

[0003] Most of the conventional MPPT methods deliver a reference for the PV voltage, which is later used in a limited bandwidth proportional plus integral (PI) controller to generate the amplitude of the grid-side current reference.

[0004] The voltage and current output characteristic of a photovoltaic cell or module is usually represented in the form of i-v or p-v curves, as shown in Figure 1. In Figure 1 the i-v curve is the curve that starts at i = 1 and ends at v = 1, and the p-v curve is the curve having a maximum point when voltage v = V_MPP.

[0005] The most important electrical characteristics of PV panels are presented in Figure 1. Figure 1 shows the short circuit current (I_SC), which is the maximum value of current the cell can generate, and it is produced under short-circuit condition (v= 0 V). Open circuit voltage (V_OC) corresponds to the highest value of voltage generated in open circuit condition (i = 0 A). Further, Figure 1 shows the maximum power point (MPP), which is the operating point (voltage V_MPP and current I_MPP) where the PV cell produces the maximum power (P_MPP = V_MPP.I_MPP).

[0006] The relationship between voltage and current can be expressed as the following implicit static nonlinearity

where V_T is referred to as the thermal voltage, which is calculated according to the Boltzmann constant (K = 1.38.10^-23 J/K), magnitude of electron charge (q = 1.6.10^-19 C), PV temperature (T = (K)), idealizing factor (1 <m<2) and number of cells in series (N), according to

; R_s is the series resistance, which depends on V_MPP,I_MPP,V_OC and I_SC under standard conditions and V_T .

[0007] The previous expression (1) can be further reduced if R_s << 1

[0008] It should be noted that current i = i(v) is a static explicit nonlinear function of the voltage, i.e., dynamics are not considered. The power delivered by the PV module can be computed simply as a product of current and voltage as follows

[0009] Manufacturers of photovoltaic cells and modules usually provide the open circuit voltage (V_OC), the short circuit current (I_SC) and the maximum power point (P_MPP=V_MPP.I_MPP) under standard test conditions (irradiance of 1000W/m² at 25°C cell temperature). However, these parameters, and consequently the i-v PV characteristic, are affected by temperature and solar irradiation as shown in Figure 2.

[0010] Figure 2 shows that the open circuit voltage V_OC varies with both irradiance and temperature. In fact, V_OC has a negative temperature coefficient and depends logarithmically on the irradiance [1]. Figure 2 also shows that, although the short circuit current I_SC changes proportionally to the irradiance, it is relatively insensitive to temperature variation [1].

[0011] In any case, the MPP is varying in function of such environmental conditions, and thus it is important to have a strategy to guarantee the operation of the PV module on the MPP at all times. These strategies are referred to in the PV systems literature as maximum power point tracking (MPPT) algorithms [2], [3] and [4].

[0012] The PV panel must be forced to operate in the MPP as shown in the p-v plot of Figure 1. This guarantees that the power extracted from the PV module is the maximum power available. This objective can be recast as a regulation objective that can be fulfilled if one of the following is satisfied as t→ ∞:

where i_MPP,ν_MPP and P_MPP are the current, voltage and power in the MPP. From now on, (·)* represents the reference for (·).

[0013] In the case of single-stage inverters, as those shown in Figures 3 and 4, this regulation objective is achieved by modulating the amplitude of the grid-side current i₀. The reference for this current, referred to as P, is generated by the MPPT algorithm. In Figures 3 and 4 a basic topology for the photovoltaic system is presented. A solar panel, string or module 31 produces a DC voltage v. A capacitor C is connected in parallel with the panel, and the voltage from the parallel connection is fed to an inverter 32, which is presented in Figures 3 and 4 as a voltage source inverter (VSI). The output of the inverter is filtered with a filter 33 and fed further to the grid 34.

[0014] In the known MPPT algorithms, the generation of P is performed indirectly by means of an intermediate PI controller 35 as shown in Figure 3. In this case, the MPPT 36 generates the voltage reference ν_Cref for the PV voltage ν_C, which is then compared to the measured ν_C, and the difference is used by the PI controller 35 to generate the amplitude P. The control block 40 of Figure 3 also includes a synchronization block 37, which reconstructs the frequency and phase of the grid voltage for producing a desired inverter output current, and a grid control block 38, which produces a voltage reference for the modulator 39. The PI controller of Figure 3 must be tuned to have a relatively small bandwidth to alleviate the effect of the 2^nd harmonic fluctuation. As a result, a poor dynamic response is obtained as the response speed is considerably reduced.

[0015] The current reference is computed from the obtained amplitude information P as

where ν_s,1 is the fundamental component of the grid voltage, and v_S,RMS its RMS value. Usually ν_s,1 is obtained by means of an external PLL or any other synchronization process. A current control loop is then designed to guarantee that the grid-side current i₀ follows such a reference i^*₀ defined above in an accurate and fast manner.

[0016] In the case of dual-stage converters, as shown in Figure 5, the MPPT generates the voltage reference ν_ref for the input capacitor C_PV (PV voltage), which is then used to compute the duty ratio u in block 54 for the DC-DC converter 51. In most dual-stage topologies, the output voltage of the DC-DC converter is controlled by the inverter, while the input voltage v is controlled by the DC-DC converter. That is, the DC-DC converter 51 is responsible for guaranteeing the operation in the MPP. The amplitude P for the current reference is generated by a PI controller 52 that guarantees the regulation of the capacitor voltage ν_C towards a given constant reference ν_Cref. The voltage reference ν_Cref is a design parameter defined externally. The amplitude P is fed to a grid control block 55, which operates to produce a voltage reference e for producing a desired current i₁.

[0017] The most common MPPT algorithms are the constant voltage (CV), the perturbation and observation (P&O) and the incremental conductance (IncCond), and modifications to them. In [2], [3] and [4] a survey on the different MPPT schemes is presented as well as a comparative study. In fact, both P&O and IncCond are based on a perturb and observe approach. The idea behind this approach consists in perturbing the PV voltage by adding or subtracting a small step and then observing the resulting changes in power. A decision based on these changes is then made to decrease or increase the PV voltage in the next sampling time.

[0018] From these algorithms, a reference for the PV voltage is obtained, which is later used in a PI system to generate the final control signal, usually the amplitude of the reference for the grid-side current. Both methods, P&O and IncCond, usually oscillate close to the MPP as they are based on a perturb and observe process. On the other hand, the CV has no oscillations but it rarely reaches the MPP.

[0019] An interesting issue in MPPT schemes that has attracted the attention of many researchers is the performance under rapidly changing atmospheric conditions [5], [6], [7], [8]. It has been observed that P&O suffers from big excursions in the wrong direction after rapidly changing irradiation conditions, that is, P&O fails to track the MPP effectively, while IncCond may still show good accuracy and efficiency in these conditions.

[0020] In [9] the authors present MPPT methods based on the idea of reconstructing the variation of power with respect to voltage on the PV (dp/dv) or the variation of power with respect to the duty ratio of the DC-DC converter attached to the PV (dp/dD). The authors deal with the problem of attaching a battery charger after the DC-DC converter, which restricts the output voltage to be constant. Thus, the maximization of output power turns out to be equivalent to maximizing the output current of the DC-DC converter. Hence, the measurement of the PV voltage becomes unnecessary. The interesting point here is that the authors not longer maximize the PV power but the power after the DC-DC converter, which is referred as the actual usable power.

[0021] Common drawback in the known MPPT schemes is that current measurement from the PV panel is required.

BRIEF DESCRIPTION OF THE INVENTION

[0022] An object of the present invention is to provide a system and a method so as to solve the above problems. The object of the invention is achieved by a system and a method, which are characterized by what is stated in the independent claims. Preferred embodiments of the invention are disclosed in the dependent claims.

[0023] The invention is based on estimating the derivative of power from the cell as a function of voltage of the cell. Further the power from the cell is not calculated based on the direct measurement of current from the cell, instead, the power is reconstructed from signals measured from the AC side of the inverter of the system. Especially harmonic components of the PV voltage and estimated power are used in the MPP tracker.

[0024] In contrast to most conventional maximum power point trackers (MPPT) schemes, the present method does not follow the perturbation and observation approach. Instead, the idea behind the proposed method is to use the information on the gradient of power with respect to the PV voltage to establish the amplitude of the grid-side current, which, in its turn, guarantees convergence of state trajectories towards the MPP. The present scheme is thus referred to as a current sensorless direct gradient maximum power point tracker "DG-MPPT-iless".

[0025] Most of the conventional MPPT methods deliver a reference for the PV voltage, which is later used in a limited bandwidth proportional plus integral (PI) controller to generate the amplitude of the grid-side current reference. In contrast, the present method delivers the amplitude directly, thus naturally guaranteeing a considerably faster response.

[0026] The proposed DG-MPPT-iless for the single-stage topology can directly deliver the power reference P, which is used as the modulation amplitude to build the grid side current reference. In addition, it can be re-structured so as to deliver a PV voltage reference as in most conventional methods. This reference is then used in an additional PI controller to generate P.

[0027] The method does not require measurement of the PV current normally used to generate the PV power signal, thus reducing the number of sensors. Instead, estimators have been designed to recover information provided by the PV power. The design of the estimators is based on the structure of the system mathematical model, and uses information available on the AC side of the inverter.

[0028] The proposed DG-MPPT-iless for the single-stage topology is not based on the perturbation and observation concept, therefore very small ripple is expected in the generation of the modulation amplitude P. This has the additional advantage of producing a cleaner grid side current.

[0029] The method utilizes phase information of needed signals for tracking the MPP of a PV module and hence the absolute amplitude of the signals is not needed. This is very useful for the simplification of the method in practical implementation. In addition, variations e.g. in capacitance of the input capacitor interfacing PV module or losses of the inverter stage do not affect the performance of the method significantly.

BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In the following, the invention will be described in greater detail by means of preferred embodiments with reference to the attached drawings, in which

Figure 1 shows i-v and p-v curves of a PV panel,

Figure 2 shows the influence of (a) solar irradiance and (b) cell's temperature on the PV i-v characteristic curve,

Figure 3 shows a connection of the PV panel to the grid by means of a single-stage inverter, and controller using an indirect MPPT,

Figure 4 shows a connection of the PV panel to the grid by means of a single-stage inverter,

Figure 5 shows a connection of the PV panel to the grid by means of a dual-stage inverter,

Figure 6 shows a direct regulation controller DG-MPPT for the single-stage inverter,

Figure 7 shows DG-MPPT for a single-stage inverter,

Figure 8 shows a single-phase single-stage inverter grid-connected through an LCL filter,

Figure 9 shows an alternative implementation of the DG-MPPT for single-phase single-stage inverter,

Figure 10 shows an estimator of the 2^nd harmonic component of the PV power p̃ (2H-POW),

Figure 11 shows a simplification to the estimator of the 2^nd harmonic component of the PV power p and variable z,

Figure 12 shows an estimator of the DC component of the PV power (DC-POW),

Figure 13 shows a notch filter used to cancel out the 2^nd and the 4^th harmonic components,

Figure 14 shows a simplification to the structure of Figure 13,

Figure 15 shows an embodiment of the present invention,

Figure 16 shows another embodiment of the invention,

Figure 17 shows another embodiment of the invention,

Figure 18 shows a simplified structure of an embodiment,

Figure 19 shows another simplified structure of an embodiment,

Figure 20 shows biased sinusoidal irradiation profile for the simulations carried out, and

Figures 21, 22, 23 and 24 show simulation results.

DETAILED DESCRIPTION OF THE INVENTION

[0031] An approach to develop the present invention is to first develop an MPPT controller with PV current measurement using the information of the derivative of power p as a function of ν_C i.e.

The idea of using the derivative

has a very intuitive significance since it represents the slope of the p-v characteristic curve. It should be noted that the derivative is exactly zero at ν_C = ν_MPP , it is positive for ν_C < ν_MPP , and negative for ν_C > ν_MPP .

[0032] As seen from Figure 4, which shows the development towards the present invention, both current from the photovoltaic panel and capacitor voltage are measured. These measured values are fed to a control block 101, and more specifically to a maximum power point tracker MPPT 102. Figure 6 shows the basic structure of the MPPT having the measured current and capacitor voltage as inputs.

[0033] From the block diagram of Figure 6 it can be seen that the measured current and voltage are multiplied, and the obtained product, representing power, is fed to a derivative block 200 together with the measured capacitor voltage. The block 200 calculates the above-mentioned derivative and further divides the derivative by the measured capacitor voltage. The division by ν_C makes a normalization and also helps to balance the slope whose absolute value is much lower in ν_C < ν_MPP than in ν_C > ν_MPP . However, ν_C is positive in the whole operating region, and thus this division does not affect the sign of the slope. The functioning of the method is based on the information of the sign of such a slope.

[0034] The obtained derivative is then used in the PI plus feedforward controller as shown in the block diagram of Figure 6, where k_p and k_i are the proportional and integral gains in blocks 201 and 202, respectively. As further seen in Figure 6, the results from the blocks 201 and 202 are subtracted from the calculated power signal p to produce a signal P, which is the outcome of the MPPT and can be used as a reference value for the amplitude of the output current, as above explained.

[0035] The estimation of the derivative

is based on the information provided by the 2^nd harmonic component of the unavoidable fluctuation present in both power and voltage, p̃ and ṽ_C, respectively. These harmonic components are then processed as shown in the more detailed diagram of the DG-MPPT shown in Figure 7.

[0036] The above MPPT method based on the concept of direct-gradient is referred as DG-MPPT method. This method itself provides fast response in case of rapid changes in irradiation and temperature and it is usable in connection with a single-phase grid connection. The idea behind the DG-MPPT consists in the reconstruction of the rate of change of PV power with respect to PV voltage (dp/dν), which is later integrated to directly generate the power reference. For the reconstruction of dp/dν the harmonic ripple present in those two signals are extracted and correlated. In the single-stage case, PV signals are naturally perturbed by a second harmonic caused by the fluctuating delivered power. That is, DG-MPPT method uses this second harmonic ripple already contained in the PV signals rather than introducing additional perturbation as in most conventional MPPT methods. DG-MPPT requires the measurement of both PV voltage and PV current, where the latter is needed for the computation of the PV power. In DG-MPPT the 2^nd harmonic components of both PV power p and capacitor voltage ν_C are estimated, which were referred as p̃ and ṽ_C, respectively, as shown in Figure 7.

[0037] The above DG-MPPT with current measurement is explained only for better understanding of the present invention, in which the measurement of PV current is not needed.

[0038] Let us consider the single-stage case of a single-phase inverter connected to the grid by means of an LCL filter as shown in Figure 8. The mathematical model of the capacitor on the DC side of the VSI of Figure 8 can be obtained as a power balance in the following form

where ν_C is the voltage in the capacitor connected in parallel to the PV, and thus, ν_C =ν; i is the current of the PV; e is the injected voltage, which is generated by the VSI; i_inv is the current on the inverter DC side; i₁ is the current on the inverter AC side, which is equal to the current on the inductor L₁ .

[0039] And for the LCL filter on the AC side of the inverter the model is

where i₀ is the current of inductor L₀ , also referred as current on the grid side; ν_C0 is the voltage in the capacitor C₀; and ν_s is the grid voltage.

[0040] It is assumed that the VSI has no losses, and thus, the power at its input equals the power delivered at its output. Moreover, in what follows it is assumed that e is a known signal coming out of the grid controller. A grid controller is a device which synchronizes the operation of the inverter with the grid voltage and produces switching sequences for the switching devises in the VSI.

[0041] Model (6)-(7) and the developments that follow hold for different topologies of inverters. This is possible as only the modulation signal e has been considered in the model. Out of this signal, and depending on the topology selected, the switching sequences can be generated using a suitable modulation algorithm.

[0042] In the following, the present invention, in which PV current is not measured, is described in more detail. Instead of estimating the PV current based on a model of the system, the idea is to indirectly reconstruct the 2^nd harmonic component of the PV power p using the information of the power in the inverter AC side, that is, the power given by p_inv = ei₁ . Notice that p_inv can be obtained as the product of two known signals.

[0043] To facilitate the design, consider the following transformation

[0044] Out of which the model gets the form

where we have used the fact that p = ν_Ci, which represents the power delivered by the PV. Notice that with this transformation, the MPP, originally at [p_MPP, v_MPP], has been mapped to the point [p_MPP, z_MPP]. Based on transformation (8) a new (power to z variable) pz-characteristic curve can be obtained preserving the same convexity of the original (power to voltage) pv-characteristic curve. Moreover, the MPP is also reached whenever the power reaches p_MPP . Roughly speaking, it is equivalent to use variable z to search the MPP in this new pz-characteristic curve, as using the voltage variable ν_C in the pv-characteristic curve.

[0045] Based on the model description (9), the DG-MPPT can also be realized using the new variable z as shown in Figure 9. This alternative implementation of the DG-MPPT is the base for the current sensorless method as it is described next. All along the paper the notation 〈x〉_k, is used as the operator that extracts the k-th harmonic component of variable x . For instance,〈x〉₂ is the operator that extracts the 2^nd harmonic component of x . All along the paper we use indistinctly 〈x〉₀ or 〈x〉_DC to represent the operator that extracts the DC component of signal x .

[0046] As above stated, the measurement of the PV current can be substituted by estimations using available information. To better understand the idea behind this estimation process let us extract the 2^nd harmonic component on both sides of (9), i.e., after application of operator 〈·〉₂ we obtain

[0047] Solving for 〈p〉₂ yields

where linearity of the operator 〈·〉₂ has been assumed.

[0048] Roughly speaking, the 2^nd harmonic component of the PV power can also be obtained by adding the 2^nd harmonic components of both the power in AC inverter side p_inv and the power handled by the capacitor C, with the advantage that the current on the DC side is no longer necessary. These harmonic components can be estimated by using, for instance, the band-pass filters (BPF) 2H-QSG shown in Figure 10, which are quadrature signal generators (QSG) tuned at the 2^nd harmonic of the fundamental.

[0049] The estimate of 〈p〉₂ can now be used in the place of p̃ , which appeared in the DG-MPPT scheme of Figure 9 based on current measurement. Figure 10 shows the block diagram of the proposed estimator, referred as 2H-POW, used to reconstruct p̃ . This estimator uses mainly two 2H-QSG to reconstruct the 2^nd harmonic components 〈ż〉₂ and 〈ei₁〉₂. They are referred as 2H-DQSG-1 and 2H-QSG-1, respectively.

[0050] In Figure 10 block 2H-DQSG-1 receives the calculated variable z as input together with the fundamental frequency of the grid voltage ω₀ which can be obtained with an external phase locked loop (PLL), for example. Block 2H-QSG-1 receives the same frequency and the product ei₁. The frequency response of 2H-DQSG-1 and 2H-QSG-1 consists of a very narrow resonant peak tuned at twice the fundamental frequency. These very selective estimators can thus be used as effective estimators of the second harmonic component with a relatively fast response.

[0051] The output of the integrator on the top of 2H-DQSG-1 in Figure 10 is the estimate of the 2^nd harmonic of the input signal. Therefore, the input to such integrator must be the time derivative of this 2^nd harmonic component estimate, i.e.,

In other words, the time derivative of the estimate of the 2^nd harmonic component is available from the 2H-DQSG-1. Here 〈ż〉₂ is reconstructed by using such available signal

According to (10), the gain λ₁ must be the same in both 2H-DQSG-1 and 2H-QSG-1.

[0052] For the estimation of 〈ż〉₂ it is assumed that both the operators 〈·〉₂ and

commute. This assumption is valid if the fundamental frequency ω₀ varies relatively slow. This is equivalent to say that the QSG is almost linear.

[0053] The 2^nd harmonic component of the variable z, i.e., 〈z〉₂ can still be estimated as in the DG-MPPT of Figure 9, that is, by using yet another QSG tuned at the 2^nd harmonic to get z̃ . This QSG is similar band-pass filter as the 2H-QSG-1 of Figure 10 and is referred as 2H-QSG-2 in Figures 10 and 11.

[0054] In the case that the fundamental frequency ω₀ is a known constant, then the estimators of the 2^nd harmonic component of power p and variable z represented by p̃ and z̃ , respectively, can be reduced to simple BPFs as shown in Figures 10 and 11.

[0055] In the DG-MPPT with current measurement shown in Figure 9, the PV power p was also used as a feedforward term, however, this signal is not available anymore. According to an embodiment the DC component of p is reconstructed by means of another estimator, and then this estimate is used as feedforward term.

[0056] Consider the above model (9), and assume that the DC component of p , referred as p̅ = 〈p〉₀, is an unknown constant. Based on the structure of model (9), the following estimator can be structured, which is referred as DC-POW

where λ₀ > 0 and γ₀ > 0 are two design parameters; ẑ and

represent the estimates of z and p̅, respectively. According to an embodiment the estimate

is used then as the feedforward term for the PI in the DG-MPPT-iless. This feedforward term improves the dynamical performance of the DG-MPPT-iless.

[0057] As the signal

contains 2^nd and 4^th harmonic components, then, according to an embodiment, such components are filtered out before using the signal as the feedforward signal. This filtering is performed, for example, by means of a notch filter of the form 2&4-NOTCH-1 as shown in Figure 13, thus avoiding reinjection of harmonic distortion to the construction of P. A block diagram of this estimator is presented in Figure 12 and it is based on the equation (12).

[0058] The idea behind the above referred notch filter consists in designing an estimator for the harmonic distortion y_h, which in this case is mainly composed by the 2^nd and 4^th harmonic components. This estimated disturbance y_h is then subtracted from the overall polluted signal

as shown in Figure 13, thus yielding the DC component mainly. Notice that the estimator is composed of two second-order harmonic oscillators (SOHO) tuned at the 2^nd (2H-SOHO) and 4^th harmonics (4H-SOHO), where γ₁ and γ₂ are two positive design parameters that simultaneously fix the gain and the quality factor of the notch filters.

[0059] In the case of a well known and constant fundamental frequency ω₀, the structure of Figure 13 can be reduced to a structure shown in Figure 14.

[0060] Summarizing, the block diagram of the modified DG-MPPT-iless for the single-stage PV inverter is shown in Figure 15. The method comprises the estimator 2H-POW for p̃ , the estimator DC-POW for the DC component of power

, and uses variable z in the place of the capacitor voltage v_C .

[0061] It can be shown that the DG-MPPT-iless scheme is robust with respect to uncertainties in the capacitance C. For this, notice that the DC component of the product of both disturbances p̃ and z̃, which is used as an estimation of

in the DG-MPPT-iless, can be computed as follows

where the term 〈C〈ż〉₂〈z〉₂〉₀ vanishes as it is the product of two signals having a phase shift difference of 90 degrees. This produces mainly higher order harmonics, which are filtered out, with no DC component. In other words, the DG-MPPT-iless method is robust with respect to uncertainties in the capacitor C as the term associated to C vanishes in the steady state during the extraction of the DC component of the product p̃ z̃ . Based on this idea, the term associated to C could even be eliminated from the DG-MPPT-iless method. However, it has been observed that this term may prevent higher transients, and thus, it may have a positive effect on the dynamic response.

[0062] On the other hand, it has been observed that uncertainties in the capacitance C may produce higher distortion in the estimate

, mainly composed of 2^nd and 4^th harmonics. Therefore, an extra notch filter of the form 2&4-NOTCH-3 to

is included, as shown in Figure 15, before using it as a feedforward term.

[0063] A low pass filter can also be included to the design together with the proportional term k_p of the PI controller, as shown in Figure 15. This simple modification may alleviate the ripple in the modulation amplitude P. For instance, a LPF of the form

with a time constant τ , might be enough.

[0064] Moreover, the DG-MPPT-iless of Figure 15 as well as the versions of the DG-MPPT shown in Figures 7 and 9 using current measurement can be re-structured as voltage reference generators, just as conventional MPPTs. For this purpose, consider the following definition

where β is a design parameter.

[0065] Out of this, the input to the PI controller, originally

would be

[0066] A scheme of this alternative description of the DG-MPPT-iless is shown in Figure 16.

[0067] In other words, the DG-MPPT-iless scheme computes a time varying increment β

which is added to the actual capacitor voltage (or PV voltage) v_C to form an intermediate variable referred as the reference voltage v_ref . The objective of the PI consists now in guaranteeing that the capacitor voltage v_C follows such a reference v_ref . Notice that the increment depends directly on the rate of change

which has the same sign of

Therefore, it is expected that the capacitor voltage will reach the MPP following the direction of the gradient.

[0068] A low pass filter may also be included to filter out additional ripple from signal β

and to keep a smooth variation of such increment. For instance a first order filter of the form

with τ₂ the time constant, would be enough.

[0069] Moreover, to guarantee a good performance in a wider range of power, the gain β can be made a function of the DC component the estimated power

.

[0070] An advantage of the embodiment, where the DG-MPPT-iless delivers an intermediate reference voltage, is that the DG-MPPT-iless can be combined with other conventional MPPT schemes delivering also a voltage reference. Notice that in all these cases, the capacitor voltage is forced to reach the reference by means of a PI controller as well.

[0071] Another option to restructure the DG-MPPT-iless scheme consists in computing the voltage reference as the integral of

as follows.

where β is a design parameter. In this case the integral term will integrate the variable

to generate the voltage reference v_ref , and the integration process will stop exactly at the point where

= 0, which happens exactly at the MPP. A scheme of this representation is shown in Figure 17.

[0072] Some of the terms and signals in the DG-MPPT-iless can be eliminated to reduce the complexity of the scheme without compromising the overall performance. For instance, the division by 〈z̃²〉₀ does not affect the sign of

, as it only gives an appropriate scaling, which makes the result slightly more linear. Its computation, however, is quite involved, and thus, it can be eliminated. The feedforward term allows a faster response during big transients and allows a better tuning of the PI gains. However, during operation in the MPP, and for relatively slow changes in irradiation, this term does not show a considerable effect, and thus it can be eliminated as well.

[0073] As above mentioned, the term associated to the capacitor power 〈C〈ż〉₂〈z〉₂〉₀ can be eliminated as it vanishes in the steady state. It may be, however, necessary to retune the parameters of the PI scheme to allow a slower response. As a result of all these simplifications, the DG-MPPT-iless can be considerably reduced as observed in the diagram of Figure 18.

[0074] It can be seen from Figure 18 that the input variables to the system are output voltage of the inverter e, which can be obtained from the grid controller, output current of the inverter i₁ and capacitor voltage ν_C. Power signal p and parameter z are calculated from these variables. The second harmonic components are extracted from p and z, and these components are multiplied with each other. Further a DC component is extracted from the obtained product, and this DC component represents derivative of power with respect to variable z.

[0075] Figure 19 shows yet another modification to allow uniform growth of the reference voltage v_ref in the admissible operation region. For this purpose, a function is included to extract the sign of

which is later integrated. Notice that with this modification, additional slight oscillations are expected once the MPP is reached, as it is usual in high gain controllers.

[0076] For the simulation test, the single-phase single-stage PV inverter of Figure 8 is considered, which is grid-connected through an LCL filter. This system has been designed using the following parameters: L₁=2 mH, L₀=833 µ H, C₀=10 µ F, C=2200 µ F. This system includes a string of PV modules producing a power of about P_MPP=2680 W in the MPP at 1000 W/m² of irradiance and a temperature of 25°C. This corresponds to a voltage in the MPP of about v_MPP=362 V, and a current of i_MPP=7.36 A. The grid voltage is a sinusoidal signal with an amplitude of 230 V_RMS, and a fundamental frequency of ω₀ =100π r/s (ƒ₀ = 50 Hz).

[0077] The system is controlled by the DG-MPPT-iless of Figure 16, plus a suitable grid controller. The proportional and integral gains of the DG-MPPT-iless have been tuned to k_p = 40 and k_i = 75 . The parameters for the filters 2H-QSG-1, 2H-DQSG-1 and 2H-QSG-2 are tuned to λ₁ = λ₂ = 200 , which corresponds to a time response of T_λ1 = T_λ2 =11 ms. The parameters for the 2&4H-NOTCH-1 are tuned to γ_1-1 =γ_2-1 =400, which corresponds to Tγ_1-1 =T_γ2-1 =5.5 ms. In the controller expressions it is assumed that the fundamental frequency is a known constant of ω₀ =100π r/s. Therefore, we use the implementation of the reduced DG-MPPT-iless. A first order LPF of the form

has been included to filter the signal β

with a constant time of τ = 0.01 s. The parameters for the estimator of

have been fixed to λ₀ =1000 and γ₀ = 20. A notch filter of the form 2&4-NOTCH-3 has been applied to

prior to use it as a feedforward term, with estimation gains γ_1-3 =γ_2-3 = 25. Moreover, no division by 〈z̃²〉₀ has been considered. In its place a gain β in function of the estimated power

is considered, this gain is about 0.0001 for the maximum power, and about 0.003 for the minimum power.

[0078] To test the response of the proposed scheme to irradiation changes, a profile for the irradiation has been proposed in such a way that the irradiance changes between 400 and 1000 W/m² following the shape of a biased sinusoidal function at a frequency of 1/π Hz, i.e., G = 700 + 300sin(2t +φ) W/m², as shown in Figure 20. The cell temperature has been fixed to 25°C.

[0079] Figure 21 shows the MPP tracking capability of the proposed DG-MPPT-iless when the irradiation is changed from 400 to 1000 W/m² following the biased sinusoidal shape. On the left side is the result considering the correct value of C in the controller, i.e., 2200 µF , while on the right is the result considering a 20% mismatch, i.e., 1760 µF. The set of available MPPs at the different radiations is represented by the gray line in the middle of the actual response in black. Therefore, the PV generated power is very close to the maximum available power in average.

[0080] Figure 22 shows the responses of the (top) PV current and (bottom) PV voltage. On the left side the responses considering the correct C in the controller expressions, while on the right a mismatch of 20% has been used. Notice that both responses are very similar.

[0081] Figure 23 shows the responses of the grid side current i₀ and the scaled grid voltage ν_s /10 . Notice that they are both sinusoidal signals in phase, thus, operation with a power factor (PF) close to one is guaranteed.

[0082] Figure 24 shows the response of the modulation amplitude P used to compute the grid side current reference

Notice that this signal has an almost imperceptible ripple, and thus, no further deformation is expected in the grid side current i₀, thus guaranteeing a low total harmonic distortion (THD).

[0083] In the above, the photovoltaic system is mainly described as having a photovoltaic module. The term "module" should be interpreted broadly as a photovoltaic module consisting of any number of cells, modules, strings or arrays.

[0084] It will be obvious to a person skilled in the art that, as the technology advances, the inventive concept can be implemented in various ways. The invention and its embodiments are not limited to the examples described above but may vary within the scope of the claims.

REFERENCES

[0085] 

[1] W. Zhou, H. Yang, Z. Fang, "A novel model for photovoltaic array performance prediction," Applied Energy, Vol. 84(12), pp. 1187-1198, Dec. 2007.

[2] M.C. Cavalcanti, K.C. Oliveira, G.M. Azevedo, D. Moreira and F.A. Neves, "Maximum power point tracker techniques for photovoltaic systems," in Proc. Power Electronics and Intelligent Control for Energy Conservation PELINCEC'05, 2005, Warsaw, pp. 1-8.

[3] T. Esram, P.L. Chapman, "Comparison of photovoltaic array maximum power point tracking techniques," IEEE Transactions on Energy Conversion, Vol. 22(2), pp. 439-449, June 2007.

[4] E. Koutroulis, K. Kalaitzakis, N.C. Voulgaris, "Development of a microcontroller-based, photovoltaic maximum power point tracking control system," IEEE Trans. on Power Electronics, Vol. 16(1), pp. 46-54, Jan. 2001.

[5] A. Pandey, N. Dasgupta and A.K. Mukerjee, "High-Performance algorithms for drift avoidance and fast tracking in solar MPPT System," IEEE Transactions on Energy Conversion, Vol. 23(2), pp. 681-689, June 2008.

[6] D. Sera, R. Teodorescu, J. Hantschel, and M. Knoll, "Optimized maximum power point tracker for fast-changing environmental conditions," IEEE Transactions on Industrial Electronics, Vol. 55(7), pp. 2629- 2637, July 2008.

[7] M. Miyatake, T. Inada, , I. Hiratsuka, H. Zhao, H. Otsuka and M. Nakano, "Control characteristics of a Fibonacci-search-based maximum power point tracker when a photovoltaic array is partially shaded," in Proc. 4th International Power Electronics and Motion Control Conference IPEMC 2004, 14-16 Aug. 2004, Vol. 2, pp. 816 - 821.

[8] T.Y. Kim, H.G. Ahn, S.K. Park and Y.K. Lee, "A novel maximum power point tracking control for photovoltaic power system under rapidly changing solar radiation," in Proc. IEEE International Symposium on Industrial Electronics ISIE 2001, 12-16 June, 2001, Vol. 2, pp. 1011-1014.

[9] C.R. Sullivan and M.J. Powers, "A high-efficiency maximum power point tracker for photovoltaic arrays in a solar-powered race vehicle," in Proc. 24th Annual IEEE Power Electronics Specialists Conference, PESC '93, 20-24 June 1993, pp. 574 - 580

Claims

1. A method of tracking the maximum power point of a photovoltaic module, the method comprising
measuring output voltage (ν_C) of the photovoltaic module,
determining output voltage (e) of the inverter connected to the photovoltaic module,
measuring output current (i₁) of the inverter connected to the photovoltaic module, characterized by
defining a variable (z) relating to energy of the capacitor using the measured PV module output voltage,
extracting second harmonic component (z̃) from the defined variable (z),
estimating the second harmonic component (p̃) of the module output power (p) using the defined variable (z), measured output current (i₁) of the inverter and the determined output voltage (e) of the inverter,
multiplying the obtained harmonic components with each other,
extracting a DC component (〈p̃ z̃〉_DC) from the obtained product (p̃ z̃), and
forming a control signal (P; ν_ref) for controlling the inverter connected to the photovoltaic module by using the extracted DC component in a PI algorithm.

2. A method according to claim 1, characterized in that the forming of the control signal (P) comprises
estimating the DC component

of the power of the module from the output voltage of the photovoltaic module, output voltage of the inverter and output current of the inverter,
using the estimated DC component

of the power (p) of the module as a feedforward signal and subtracting the output of the PI algorithm from the feedforward signal, and using the control signal (P) as a current amplitude reference.

3. A method according to claim 1 or 2, characterized in that the forming of the control signal (v_ref) comprises
multiplying the extracted DC component (〈p̃ z̃〉_DC) with a constant (β) and
adding the output voltage (v_c) of the photovoltaic module to the obtained product for producing control signal (v_ref).

4. Method according to claim 1 or 2, characterized in that the forming of the control signal (v_ref) comprises
taking the sign of the extracted DC component (〈p̃ z̃〉_DC) multiplying the sign with a constant (β), and
adding the output voltage (v_C) of the photovoltaic module to the obtained product for producing control signal (v_ref).

5. Method according to claim 1 or 2, characterized in that the forming of the control signal (v_ref) comprises
integrating the extracted DC component (〈p̃ z̃〉_DC) , and
multiplying this integral with a constant (β) for producing control signal (v_ref).

6. Method according to any one of the claims 3 to 5, characterized in that the method comprises
using the control signal (v_ref) as a voltage reference for the inverter for controlling the output voltage of the photovoltaic module.

7. Method according to any one of the claims 3 to 5, characterized in that the method comprises
using the control signal (v_ref) as input to a PI algorithm for obtaining control signal (P) for controlling the current amplitude of the photovoltaic module.

8. A method according to claim 1 or 2, characterized in that the forming of the control signal (P) comprises
calculating the square of the second harmonic component of the variable relating to energy of the capacitor,
extracting a DC component from the calculated square,
dividing the DC component of the product of second harmonic components of variable (z) and power of the module with the DC component of the square of the second harmonic of variable (z), and
feeding the result of the division to the PI algorithm for obtaining the control signal (P).

9. Method according to any one of the claims 1 to 8, characterized in that the variable (z) relating to energy of the capacitor is defined as square of the capacitor voltage divided by two

10. An arrangement for tracking the maximum power point of a photovoltaic module, the arrangement comprising
means for measuring output voltage (v_C) of the photovoltaic module,
means for determining output voltage (e) of the inverter connected to the photovoltaic module,
means for measuring output current (i₁) of the inverter connected to the photovoltaic module, characterized by
means for defining a variable (z) relating to energy of the capacitor using the measured PV module output voltage,
means for extracting second harmonic component (z̃) from the defined variable (z),
means for estimating the second harmonic component (p̃) of the module output power (p) using the defined variable (z), measured output current (i₁) of the inverter and the determined output voltage (e) of the inverter,
means for multiplying the obtained harmonic components with each other,
means for extracting a DC component (〈p̃ z̃〉_DC) from the obtained product (p̃ z̃), and
means for forming a control signal (P;v_ref) for controlling the inverter connected to the photovoltaic module by using the extracted DC component in a PI algorithm.

11. A photovoltaic inverter comprising the arrangement of claim 10.

Drawing