A deadbeat control method for LCL-type photovoltaic inverters based on composite prediction

Abstract

The present invention discloses a deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction, including: analyzing the voltage loop equation of the photovoltaic grid-connected system, discretizing it, and conducting a preliminary theoretical analysis of deadbeat control; using the repetitive control method to calculate the output current at the k+1 moment, and on this basis, using the Lagrange interpolation method to calculate the current at the k+2 moment and perform repetitive compensation on it; adopting adaptive forward linear current prediction, sampling and predicting the forward current at intervals of two cycles, and using the Levinson-Durbin algorithm for the prediction error to obtain the update equation and the minimum error; performing composite control on the repetitive prediction and the adaptive forward current prediction, and deriving its discriminant. This method shortens the system oscillation time, enhances the dynamic response ability, and finally performs composite control on the two methods, predicting the inverter output current one beat in advance and predicting the sampling reference current two beats in advance, solving the problem of time delay.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic grid-connected inverter control, and particularly relates to a deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction. More specifically, it relates to a control method for an LCL-type grid-connected inverter based on a composite improved deadbeat strategy of repetitive control and forward current linear control. Background Art

[0002] The resonant suppression control method for photovoltaic grid connection has always been one of the important problems restricting the development of solar power generation. As an important control method for resonant suppression, deadbeat control has received wide attention due to its good dynamic response, wide frequency range of resonant suppression, and fast control speed.

[0003] At present, many scholars at home and abroad have done a lot of research on deadbeat control. The article "Deadbeat Grid-Connected Control Strategy for Cascaded Multilevel Quasi-Z-Source Inverters" in the 17th issue of "Power System Protection and Control" in 2019 shows that traditional deadbeat control can achieve deadbeat control under ideal conditions. However, in actual engineering applications, due to problems of time delay and prediction accuracy, its control effect is reduced.

[0004] The article "Harmonic Suppression Method for Photovoltaic Inverters Based on Improved PI + Repetitive Control" in the 4th issue of "Power Capacitors & Reactive Power Compensation" in 2019 proposed to suppress grid-connected resonance by introducing PI + repetitive control. PI control can act quickly when resonance occurs and can change with the change of system frequency, enabling dynamic control with strong operability and improving the anti-disturbance ability of the inverter. However, its bandwidth is narrow, the dynamic response is poor, and the resonance suppression effect is not good.

[0005] The article "Deadbeat Control Strategy for Modular Multilevel Converters" in the 6th issue of "Proceedings of the CSEE" in 2017 proposed to introduce traditional deadbeat control to solve the bandwidth and dynamic response problems of PI control, which can predict the resonance frequency in advance and suppress resonance, reducing the steady-state error and greatly improving the stability of system operation. However, the current prediction accuracy of traditional deadbeat control is not high, and there is still a time delay problem in its control, which is not conducive to resonance control.

[0006] The article "Research on Improved Deadbeat Photovoltaic Grid-Connected Resonance Suppression" in the 19th issue of "Power System Protection and Control" in 2018 proposed to improve on the basis of traditional deadbeat control, solving the problem of time delay in the traditional deadbeat control system and further strengthening the resonance suppression effect. However, its predicted current is derived through formula derivation, with low prediction accuracy and large error.

[0007] In the 11th issue of "Electronic Science and Technology" in 2021, the article "Improved Current Prediction Deadbeat Control Method for Active Power Filters" proposed using Lagrange interpolation method to improve the current accuracy of repeated prediction. When the load is stable, the resonance suppression effect is better, and the harmonic content is reduced by 90%. However, when the load changes, the oscillation time of the system is longer, and the resonance effect will be reduced a lot in its prediction method.

[0008] In the 6th issue of "Chinese Journal of Scientific Instrument" in 2014, the article "Deadbeat Harmonic Current Tracking Control Based on Composite Prediction" proposed an improved deadbeat control method with composite prediction, which shortened the oscillation time of the system and could suppress resonance when the load changed or was stable. However, its prediction accuracy is on the low side and needs to be further improved.

[0009] Considering the defects existing in the above-mentioned existing technologies, aiming at the problems of long time delay and low prediction accuracy of traditional deadbeat control, how to further optimize it is an urgent problem to be solved by practitioners in the same field. Summary of the Invention

[0010] The purpose of the present invention is to provide a deadbeat control method for LCL-type photovoltaic inverters based on composite prediction, which can solve the problems of long time delay and low prediction accuracy of traditional deadbeat control.

[0011] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0012] The present invention provides a deadbeat control method for LCL-type photovoltaic inverters based on composite prediction, including the following steps:

[0013] S1. Analyze the voltage loop equation of the photovoltaic grid-connected system, discretize it, and conduct a preliminary theoretical analysis of deadbeat control;

[0014] S2. Use the method of repetitive control to calculate the output current at time k + 1. On this basis, use Lagrange interpolation method to calculate the current at time k + 2 and perform repetitive compensation on it;

[0015] S3. Adopt adaptive forward linear current prediction, sample and predict the forward current at intervals of two cycles, and use the Levinson-Durbin algorithm for the prediction error to obtain the update equation and the minimum error;

[0016] S4. Perform composite control on the repetitive prediction in step S2 and the adaptive forward current prediction in step S3, and derive its discriminant.

[0017] Further, the discretized voltage loop equation of the photovoltaic grid-connected system in step S1 is:

[0018]

[0019] Wherein, U1(k), U2(k), and U3(k) are the output voltages of the first, second, and third inverters at time k, respectively; L 11 is the first inductor of the first inverter; L 12 is the second inductor of the first inverter; L 21 is the first inductor of the second inverter; L 22 is the second inductor of the second inverter; L 31 is the first inductor of the third inverter; L 32 is the second inductor of the third inverter; i 11 is the current flowing through L 11 ; i 12 is the current flowing through L 12 ; U g is the grid-side voltage; i 11 (k + 1) is the reference current of the filter inductor on the first inverter side at the next moment, and i 12 (k + 1) is the reference current of the filter inductor on the grid side of the first inverter at the next moment; T S is the sampling period; i 11 * (k) = i 11 (k + 1), and the actual current can achieve static-error-free tracking of the reference current.

[0020] Furthermore, in the step S2, for the first inverter, the output current at time k + 2 predicted repeatedly based on the Lagrange interpolation method and its repeated compensation equation are:

[0021]

[0022] Vi = A(1 + 0.95z -N )·e(k)z 2-N ;

[0023] In the formula, i1(k) and i1 * (k) are the actual value and the reference value of the output current of the first inverter at time k; Vi is the repeated compensation amount, A is the repeated controller gain, N is the number of samplings in one cycle, and e(k) is the error between the command value and the predicted command value in the kth cycle of the repeated control.

[0024] Furthermore, the adaptive forward linear current prediction and compensation equation in the step S3 is:

[0025]

[0026]

[0027] Among them, f is the fth cycle in m cycles, is the predicted value, x(k) is the true value, and a mn is the linear prediction coefficient, and e m k is the prediction error.

[0028] Furthermore, in step S4, adaptive forward current prediction is adopted for the first inverter i1, and its judgment expression is:

[0029] e judd (k) = |i1 * (k) - i1 * (k - 256)| ≥ e max

[0030] e judd (k) is the absolute value of the difference between the current command values of two adjacent periods; e max is the maximum error;

[0031] When it is judged that the system is operating stably, its judgment expression is:

[0032]

[0033] Select the number of periods for a single stable judgment to be 8, j = 0, 1, 2... 7, representing the period; e min is the minimum error.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention provides a deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction. This method first analyzes the reasons for time delay and low prediction accuracy, and then combines the Lagrange interpolation method and repetitive prediction when the load is stable to improve the current prediction accuracy. When the load changes, adaptive forward linear current prediction is introduced to shorten the system oscillation time and enhance the dynamic response ability. Finally, the two methods are combined for control to predict the inverter output current one beat in advance and the sampled reference current two beats in advance, solving the time delay problem;

[0036] The present invention shortens the system oscillation time and enhances the dynamic response ability by applying a composite prediction method of repetitive prediction and adaptive forward linear current prediction. Finally, the two methods are combined for control to predict the inverter output current one beat in advance and the sampled reference current two beats in advance, solving the time delay problem. Description of the Drawings

[0037] Figure 1 is the structure diagram of the photovoltaic grid-connected system;

[0038] Figure 2 is the Norton equivalent circuit diagram of a single inverter;

[0039] Figure 3 It is the Norton equivalent circuit diagram of multiple inverters;

[0040] Figure 4 It is the delay diagram of traditional deadbeat control;

[0041] Figure 5 It is the composite prediction flow chart;

[0042] Figure 6a It is the deadbeat output harmonic current diagram;

[0043] Figure 6b It is the THD diagram of the traditional deadbeat control result;

[0044] Figure 7a It is a repetitive prediction harmonic current diagram when the load is stable;

[0045] Figure 7b It is another repetitive prediction harmonic current diagram when the load is stable;

[0046] Figure 8a It is the Lagrange current waveform diagram when the load changes;

[0047] Figure 8b It is the composite prediction current waveform diagram when the load changes;

[0048] Figure 8c It is the THD diagram when the load changes. Specific implementation manners

[0049] In order to make the technical means, creative features, achieved purposes and functions of the present invention easy to understand, the present invention will be further described below in conjunction with specific implementation manners.

[0050] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "inner", "outer", "front end", "rear end", "both ends", "one end", "the other end", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0051] In the description of the present invention, it should be noted that unless otherwise clearly specified and defined, terms such as "installation", "provided with", "connection", etc. should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0052] A deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction provided by the present invention includes:

[0053] S1. Analyze the voltage loop equation of the photovoltaic grid-connected system, discretize it, and conduct a preliminary theoretical analysis of deadbeat control;

[0054] S2. Use the repetitive control method to calculate the output current at the k + 1 moment. On this basis, use the Lagrange interpolation method to calculate the current at the k + 2 moment and perform repetitive compensation on it;

[0055] S3. Adopt adaptive forward linear current prediction, sample and predict the forward current at intervals of two cycles, and use the Levinson-Durbin algorithm for the prediction error to obtain the update equation and the minimum error;

[0056] S4. Perform composite control on the repetitive prediction in step S2 and the adaptive forward current prediction in step S3, and derive its discriminant.

[0057] This method first analyzes the reasons for time delay and low prediction accuracy. Then, when the load is stable, it combines the Lagrange interpolation method and repetitive prediction to improve the current prediction accuracy. When the load changes, it introduces adaptive forward linear current prediction to shorten the system oscillation time and enhance the dynamic response ability. Finally, it performs composite control on the two methods to predict the inverter output current one beat in advance and the sampled reference current two beats in advance, solving the time delay problem; and by applying the composite controller in parallel with the composite control strategy of repetitive control and adaptive forward linear current control, the harmonic content is reduced by 92%, the oscillation time is shortened by 50%, it has higher prediction accuracy and dynamic response ability, and the control effect is better.

[0058] The following will elaborate on the embodiments of the above steps in detail:

[0059] The LCL filter can filter the grid-connected current and play a role in suppressing high-frequency harmonics. Refer to Figure 1 As shown, it is the structure diagram of the photovoltaic grid-connected system, which consists of a photovoltaic front stage, an inversion link, and a filtering link. In the figure, U1, U2, and U3 are the output voltages of the first, second, and third inverters respectively, and L11 , L 12 are the first filter inductor and the second filter inductor of the first inverter, and C 11 is the filter capacitor of the first inverter; L g is the grid-side inductor, and R g is the grid-side resistor, and U pcc is the grid-connection common point. The structures of other inverters are similar to that of the first inverter. L 21 , L 22 are the first filter inductor and the second filter inductor of the second inverter, and C 21 is the filter capacitor of the second inverter; L 31 , L 32 are the first filter inductor and the second filter inductor of the third inverter, and C 31 is the filter capacitor of the third inverter;

[0060] The current of the photovoltaic panel undergoes DC-DC conversion through the front stage of the photovoltaic system and then enters the inverter stage to be converted into alternating current. After passing through the LCL filter, it is connected to the grid through U PCC and merged into the grid. From Figure 1 , the mathematical model of the LCL grid-connected inverter can be obtained as

[0061]

[0062] where, i 11 is the current flowing through L 11 , i 12 is the current flowing through L 12 ; and so on; i 21 is the current flowing through L 21 , i 22 is the current flowing through L 22 ; i 31 is the current flowing through L 31 , i 32 is the current flowing through L 32 ;

[0063] The Norton equivalent circuit diagram of a single inverter is as shown in Figure 2 . The open-loop transfer function of the grid-connected current and the output voltage of the LCL inverter:

[0064]

[0065] In the formula: G PI (s) = K P + K i / s. G PI (s) is the transfer function of the PI controller, K P is the proportional coefficient, K i is the integral coefficient, and C is the capacitor of the first inverter, L11 and L 12 are the two inductors of the first inverter; where s is an operator in the complex frequency domain.

[0066] Figure 3 is the Norton equivalent circuit diagram of multiple inverters. From Figure 3 and Equation (2), the transfer function of the grid-connected power generation system with n inverters in parallel can be derived as:

[0067]

[0068] In the formula: n is the number of inverters, L g is the grid-side inductor, and Ts is the sampling period.

[0069] According to Equation (3), the resonance frequency of the grid-connected system with n inverters is:

[0070]

[0071] It can be seen from Equation (4) that: as the number of parallel inverters increases, the resonance frequency gradually decreases, and its resonance point also increases.

[0072] Discretizing Equation (1) gives:

[0073]

[0074] In the formula: i 11 (k + 1) is the reference current of the filter inductor on the side of the first inverter at the next moment, i 12 (k + 1) is the reference current of the grid-side filter inductor of the first inverter at the next moment, T S is the sampling period.

[0075] Under ideal conditions, the actual current can achieve static error-free tracking of the reference current, that is, i 11 (k + 1) = i 11 * (k), similarly, i 12 (k + 1) = i 12 * (k), i 21 (k + 1) = i 21 * (k), i 22 (k + 1) = i 22 * (k), i 31 (k + 1) = i 31 * (k), i 32 (k + 1) = i 32 * (k) is substituted into Equation (5) to obtain:

[0076]

[0077] As can be seen from Equation (6), the deadbeat control under ideal conditions can indeed achieve the prediction and control of the inverter current in advance. However, in actual engineering applications, time is required for current sampling, calculation, and prediction signal generation. The deadbeat control under ideal conditions becomes "beat control" in actual applications, resulting in a deteriorated resonance suppression effect and a poor dynamic response. The delay schematic diagram of the traditional deadbeat control is shown in Figure 4.

[0078] During the actual sampling process, the grid voltage frequency is often lower than the sampling frequency. According to this characteristic, the change in the inverter output current can be regarded as linear growth. Ignoring the influence of the system capacitance, the predicted current of the inverter at time k + 1 is:

[0079]

[0080] In the formula, i1(k), i1 * (k) are the actual value and the reference value of the output current of the first inverter at time k, and i1(k + 1), i1 * (k + 1) are the actual value and the reference value of the output current at time k + 1. is the predicted value of the output current at time k + 1. is the predicted reference value of the output current at time k + 1, and the same applies to the subsequent formulas.

[0081] Simplifying Equation (7) gives:

[0082]

[0083] When the load is operating in a steady state, the output current fluctuates slightly and has periodicity. At this time, repeated prediction can be used to improve the prediction accuracy. The predicted reference current of the inverter output at time k + 2 is:

[0084]

[0085] There are also errors in repeated prediction. When there are errors in the sampling values at times k - 2, k - 1, and k, it will inevitably affect the current reference value at time k + 2. Therefore, Equation (9) must be repeatedly compensated, and the compensated expression is:

[0086]

[0087] In the formula:

[0088] Vi = A(1 + 0.95z -N )·e(k)z 2-N

[0089]

[0090] In formula (10), Vi is the repetitive compensation amount, A is the repetitive controller gain, N is the number of sampling times in one period, and e(k) is the error between the command value and the predicted command value in the k-th period of repetitive control.

[0091] When the load runs unstably, due to the poor dynamic performance of repetitive prediction, it cannot well predict the change of the system current. Therefore, an adaptive forward linear current prediction with better dynamic performance is adopted. The sampling values in K sampling periods are x(0), x(1),... x(k - m)..., x(k - 1),... Among them, m values are taken and let:

[0092] x(k - 1) = i1 * (k), x(k - 2) = i1 * (k - 2),......(11)

[0093] x(k - m) = i1 * (k - 2(m - 1))

[0094] m represents taking m values from the sampling values in all periods;

[0095] Then the forward current linear prediction and its error value are:

[0096]

[0097]

[0098] In formula (12): m periods are taken from k periods, and f is the f-th period among the m periods; is the predicted value, x(k) is the true value, a mn is the linear prediction coefficient, e m k is the prediction error.

[0099] The essence of composite prediction is to select different prediction methods according to different load conditions to improve the prediction accuracy. When the load runs smoothly, repetitive prediction with Lagrangian second-order interpolation is adopted; when the load oscillates, adaptive forward current prediction is adopted, and its judgment expression is:

[0100] e judd (k) = |i1 * (k) - i1 * (k - 256)| ≥ e max (13)

[0101] In formula (13): e judd (k) is the absolute value of the difference between the current command values of two adjacent periods. By setting a maximum error e for the systemmax , as long as e is detected judd (k) is greater than e max , it can be determined that the system is oscillating, and automatically switch to forward linear current prediction.

[0102] When it is determined that the system is operating stably, only the difference between two adjacent periods is judged not to meet the requirements. Therefore, multiple periods need to be selected for judgment. When too many periods are selected, the calculation amount becomes large, and the calculated error value will become small. Considering the actual situation of this article, the number of periods for single stable judgment is selected as 8, and its judgment expression is:

[0103]

[0104] j = 0, 1, 2…7, representing the period; e min is the minimum error

[0105] The comprehensive process is as Figure 5 shown.

[0106] To verify the effectiveness and superiority of the composite prediction improved deadbeat control studied in the embodiments of the present invention, a photovoltaic power generation system model containing two inverters is built on the Matlab / Simulink software platform. Taking the first one as an example for analysis, its structure diagram is as Figure 1 . The simulation results mainly include the following parts: Figure 6a , 6b is the traditional deadbeat output harmonic current diagram and its THD diagram, Figure 7a , 7b , are the two repetitive prediction harmonic current diagrams when the load is stable, Figure 8a , 8b, 8c are respectively the Lagrange current waveform diagram and the composite prediction current waveform diagram and their THD diagrams when the load changes.

[0107] From Figure 6a , 6b , 7a, 7b, 8a, 8b, 8c, it can be seen that due to its delay and control accuracy problems, the traditional deadbeat control has serious current waveform distortion and a high harmonic content, reaching 28.51%, and the resonance suppression effect is poor; when repetitive prediction is used when the load is stable, the degree of current waveform distortion is greatly reduced, and the harmonic content also drops significantly. Especially after using the Lagrange method to improve the repetitive prediction accuracy, the harmonic current waveform is smoother and the resonance suppression effect is better; when the load changes, repetitive prediction requires two periods of oscillation for the current to return to stability, while using composite prediction only requires one period, greatly improving the dynamic response ability of the system, and the harmonic content is 2.35%, meeting the national requirement that the harmonic content does not exceed 5%, and the resonance suppression effect is better.

[0108] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction, characterized in that, It includes the following steps: S1. Analyze the voltage loop equation of the photovoltaic grid-connected system, discretize it, and conduct a preliminary theoretical analysis of deadbeat control; S2. Use the repetitive control method to calculate the output current at the (k + 1)th moment. On this basis, use the Lagrange interpolation method to calculate the current at the (k + 2)th moment and perform repetitive compensation on it; S3. Adopt adaptive forward linear current prediction, sample and predict the forward current at intervals of two cycles, and use the Levinson-Durbin algorithm for the prediction error to obtain the update equation and the minimum error; S4. Perform composite control on the repetitive prediction in step S2 and the adaptive forward current prediction in step S3, and derive its discriminant; Among them, the discretization of the voltage loop equation of the photovoltaic grid-connected system in step S1 is constructed based on three inverters as: Where U1(k), U2(k), and U3(k) are the output voltages of the first, second, and third inverters at time k, respectively; L 11 is the first inductor of the first inverter; L 12 is the second inductor of the first inverter; L 21 is the first inductor of the second inverter; L 22 is the second inductor of the second inverter; L 31 is the first inductor of the third inverter; L 32 is the second inductor of the third inverter; i 11 is the current flowing through L 11 ; i 12 is the current flowing through L 12 ; U g is the grid-side voltage; i 11 (k + 1) is the reference current of the filter inductor on the first inverter side at the next moment, and i 12 (k + 1) is the reference current of the filter inductor on the grid side of the first inverter at the next moment; T S is the sampling period; i 11 * (k) = i 11 (k + 1), and the actual current can achieve static-error-free tracking of the reference current; In step S4, for the first inverter, the adaptive forward current prediction is adopted, and its judgment expression is: e judd (k) is the absolute value of the difference between the current command values of two adjacent periods; e max is the maximum error; When it is judged that the system is operating stably, its judgment expression is: Select the number of cycles for single stable judgment to be 8. Let j = 0, 1, 2... 7, representing the cycles; e min is the minimum error.

2. A deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction according to claim 1, characterized in that, In step S2, for the first inverter, the output current at the (k + 2)th moment and its repetitive compensation equation based on the Lagrange interpolation method repetitive prediction are: where, i1(k) and i1 * (k) are the actual value and reference value of the output current of the first inverter at time k; Vi is the repetitive compensation amount, A is the repetitive controller gain, N is the number of samplings in one cycle, and e(k) is the error between the command value and the predicted command value in the kth cycle of repetitive control.

3. A deadbeat control method for an LCL-type photovoltaic inverter based on composite prediction according to claim 2, characterized in that, The adaptive forward linear current prediction and compensation equation in step S3 is: Among them, f is the f-th cycle among m cycles, is the predicted value, x(k) is the true value, and a mn is the linear prediction coefficient, and e m k is the prediction error.

Images