A predictive control method based on model-free prediction and parameter-free reference
Through the prediction control method without model prediction and parameter reference, the problem of traditional model prediction control is solved, which achieves higher robustness and stability and improves control performance.
Patent Information
- Application Number
- CN202410961542.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-07-18
Smart Images

Figure CN118508773B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of energy storage converters, and in particular to a predictive control method based on model-free prediction and parameter-free reference. Background Art
[0002] In recent years, with the increasing popularity of distributed energy, the application of energy storage inverters has received more and more attention. Energy storage inverters based on LC (passive) filtering are usually used as controllable voltage sources to provide stable and reliable voltage support. In order to improve the output voltage performance of the inverter, researchers have proposed a variety of control methods. Traditional linear control is widely used in grid-connected inverters, which use a multi-loop structure to track reference points and suppress interference, but its dynamic performance is limited.
[0003] Among them, model predictive control (MPC) has become a promising method for controlling power electronic systems due to its ability to handle nonlinearities and constraints.
[0004] Compared with traditional linear controllers, MPC has better dynamic performance and greater flexibility in objective functions and constraints. The inverter voltage control method based on MPC was first reported in the literature [P. Cortes, G. Ortiz, JIYuz, J. Rodriguez, S. Vazquez and LG Franquelo, "Model Predictive Controlof an Inverter With Output LC Filter for UPS Applications," IEEE Trans. Ind.Electron., vol. 56, no. 6, pp. 1875-1883, June 2009.], in which the high-order state space system was discretized as a prediction model to seek the optimal voltage vector.
[0005] In order to further improve the output voltage performance, the literature [C. Zheng, T. , B. and F. Blaabjerg, "Constrained Modulated Model-PredictiveControl of an LC-Filtered Voltage-Source Converter," IEEE Trans. Power Electron., vol. 35, no. 2, pp. 1967-1977, Feb. 2020.] adopt a dual-objective cost function to suppress the resonance effect of the LC filter. In addition, in the literature [C. Zheng, T. , Z. Zhang, J.Rodriguez and F. Blaabjerg, "Model Predictive Control of LC-Filtered VoltageSource Inverters With Optimal Switching Sequence," IEEE Trans. PowerElectron., vol. 36, no. 3, pp. 3422-3436, March 2021.], three voltage vectors are applied per control cycle to improve voltage performance, and their duration is calculated based on the system model. However, in the case of parameter mismatch, both the calculation and prediction of the duration are affected.
[0006] A major challenge of MPC is its sensitivity to parameter mismatch, which can lead to poor performance and instability. Researchers have proposed various methods to address this issue, including the following two types of MPC: one method is to identify system parameters online; the other method guarantees stability and output performance under parameter uncertainty by incorporating uncertainty sets in the value function or constraints.
[0007] However, these methods often have the problems of high computational complexity and limited convergence speed. Therefore, it is necessary to further study the improvement of the MPC method to deal with the impact of parameter uncertainty and improve the application ability of MPC under various parameter conditions. Summary of the invention
[0008] The object of the present invention is to provide a predictive control method based on model-free prediction and parameter-free reference, which can eliminate the influence of system parameters on voltage prediction and action time calculation.
[0009] In order to solve the above technical problems, the present invention provides a predictive control method based on model-free prediction and parameter-free reference, comprising:
[0010] Construct gradient formula;
[0011] According to the mathematical relationship between different voltage vectors, a gradient update formula is constructed, and data of the bridge arm side current gradient, capacitor voltage gradient and grid-connected side current gradient corresponding to each voltage vector are updated according to the gradient update formula;
[0012] Perform current prediction and voltage prediction based on the updated gradient of each voltage vector to obtain the predicted current value on the bridge arm side, the predicted capacitor voltage value, and the predicted current value on the grid-connected side;
[0013] According to a parameter-free capacitor voltage reference value algorithm and a parameter-free bridge arm side current reference value algorithm, a capacitor voltage reference value and a bridge arm side current reference value are obtained;
[0014] According to the value function, the value function values corresponding to the eight basic voltage vectors are calculated respectively, and the voltage vector with the smallest value function value is selected as the optimal vector to be used in the next control cycle.
[0015] Optionally, before the step of constructing the gradient formula, the step further includes: converting the sampled three-phase bridge arm side current i fabc , three-phase capacitor voltage v cabc And the three-phase grid current i gabc Perform Clarke coordinate transformation to obtain the two-phase bridge arm current i in the αβ stationary coordinate system fαβ , two-phase capacitor voltage v cαβ And the two-phase grid-connected current i gαβ .
[0016] Optionally, the process of constructing the gradient formula includes:
[0017] The mathematical model of the grid-connected inverter in the αβ stationary coordinate system is discretized to obtain the prediction model. Based on the prediction model, the (k-1)th control cycle is calculated, that is, from the (k-1)th T s To the kth T s The voltage vector v acting during ix,m (k-1) (m∈{0,1,…,7}) corresponding to the bridge arm side current gradient △i fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k):
[0018] ;
[0019] Among them, v ix,m (k-1) (m∈{0,1,…,7}) is the voltage vector acting in the (k-1)th control cycle; △i fx,m (k) is v ix,m (k-1) The current gradient on the bridge arm side generated; △v cx,m(k) is v ix,m (k-1) The capacitor voltage gradient generated; △i gx,m (k) is v ix,m (k-1) The grid-connected current gradient generated.
[0020] Optionally, the gradient update formula is constructed including:
[0021] ;
[0022] ;
[0023] Where H is the gain coefficient, where v ix,n (k-1) (n∈{0,1,…,7} and n is not equal to m) is the voltage vector that is not applied in the (k-1)th control cycle; △i fx,n (k) is if v is used ix,n (k-1) The current gradient on the bridge arm side that will be generated; △v cx,n (k) is if v is used ix,n (k-1) The capacitor voltage gradient that will be generated; △i gx,n (k) is if v is used ix,n (k-1) The grid-connected current gradient that will be generated; v ix,n (k-2) is the voltage vector that is not applied in the (k-2)th control cycle; △i fx,n (k-1) is if v is used ix,n (k-2) The current gradient on the bridge arm side that will be generated; △v cx,n (k-1) is if v is used ix,n (k-2) The capacitor voltage gradient that will be generated; △i gx,n (k-1) is if v is used ix,n (k-2) The grid-connected current gradient that will be generated; v ix,m (k-2) is the voltage vector acting in the (k-2)th control cycle; △i fx,m (k-1) is if v is used ix,m (k-2) The current gradient on the bridge arm side that will be generated; △v cx,m (k-1) is if v is used ix,m (k-2) The capacitor voltage gradient that will be generated; △i gx,m (k-1) is if v is used ix,nm (k-2) the grid-side current gradient that will be generated;
[0024] When the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is 0, there is no solution for the value of coefficient H. At this time, the gradient update formula is reconstructed as:
[0025] ;
[0026] Among them, the voltage vector v that is not used in the (k-1)th control cycle ix,n The current gradient on the bridge arm side corresponding to (k-1) is △i fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) is equal to the voltage vector v used in the (k-1)th control cycle ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k);
[0027] When the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is not 0, the coefficient H is solved to be 1, and the gradient update formula is reconstructed as follows:
[0028] ;
[0029] Among them, the voltage vector v that is not used in the (k-1)th control cycle ix,n The current gradient △i on the bridge arm side corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) The voltage vector v used in the (k-1)th control cycle ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m The voltage vector v used in the (k) and (k-2) control cycles ix,m (k-2) corresponds to the bridge arm side current gradient △i fx,m (k-1), capacitor voltage gradient △v cx,m (k-1) and the grid-connected current gradient △i gx,m The voltage vector v that is not used in the (k-1) and (k-2) control cycles ix,n The bridge arm current gradient △i corresponding to (k-2) fx,n (k-1), capacitor voltage gradient △v cx,n (k-1) and the grid-connected current gradient △i gx,n (k-1) is used to represent it.
[0030] Optionally, the process of performing current prediction and voltage prediction includes:
[0031] According to the updated gradient of each voltage vector, current prediction and voltage prediction are performed, and the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k) and the grid-connected current sampling value i gx (k) and the voltage vector v used in the (k-1)th control cycle in the lookup table ix,m The current gradient and voltage gradient corresponding to (k-1) are added to obtain the bridge arm side current, capacitor voltage and grid-connected side current of the grid-connected inverter system at the (k+1)th T s Predicted value of:
[0032] ;
[0033] Among them, the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1) and the predicted value of the grid-connected current i gx,m (k+1) by the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k), grid-connected current sampling value i gx (k) and the voltage vector v ix,m The updated bridge arm current gradient △i corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k) are added together.
[0034] Optionally, the grid-connected current prediction value, capacitor voltage prediction value, and bridge arm side current prediction value at the (k+2)th Ts are calculated to compensate for the calculation delay:
[0035] ;
[0036] Among them, (k+2) T s The predicted value of the bridge arm current i fx (k+2), capacitor voltage prediction value v cx (k+2) and the predicted value of the grid-connected current i gx (k+2) is composed of the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value vcx,m (k+1), grid-connected current prediction value i gx,m (k+1) and the current gradient △i on the bridge arm side fx (k+1), capacitor voltage gradient △v cx (k+1) and the grid-connected current gradient △i gx (k+1) is added together.
[0037] Optionally, the parameter-free capacitor voltage reference value algorithm is:
[0038] ;
[0039] ;
[0040] Among them, △V α =v cα (k)-e gα (k), △V β =v cβ (k)-e gβ (k), v cαβ (k) is k T s The capacitor voltage sampling value at time e gαβ (k) is k T s The grid-connected segment voltage sampling value at time T g is the coefficient matrix of the parameter-free capacitor voltage reference algorithm.
[0041] Optionally, the parameter-free bridge arm side current reference value algorithm is:
[0042] ; ; ;
[0043] Among them, γ and η are weight coefficients, T c It is the coefficient matrix of the parameter-free bridge arm current reference value algorithm.
[0044] Optionally, based on the sampled bridge arm side current, capacitor voltage, grid-connected side current, grid-connected side voltage and a given grid-connected side current reference value, the reference value at time (k+2) is obtained according to the parameter-free capacitor voltage reference value algorithm and the parameter-free bridge arm side current reference value algorithm:
[0045] ;
[0046] Among them, i g ref (k+2), v c ref (k+2) and i f ref(k+2) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k+2); i g ref (k), v c ref (k) and i f ref (k) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k); i g ref (k-1), v c ref (k-1) and i f ref (k-1) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k-1); i g ref (k-2), v c ref (k-2) and i f ref (k-2) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k-2).
[0047] Optionally, eight sets of current prediction values and voltage prediction values and the obtained reference values are substituted into the value function equation to obtain eight value function values, and a voltage vector with the smallest value function value is selected as the optimal vector to be used in the next control cycle;
[0048] Among them, the value function is:
[0049] ;
[0050] ;
[0051] Among them, λ f , c and g They are g f , g c and g The weight coefficient of g f , g c and g They are the bridge arm current i fx , capacitor voltage v cx And the grid-side current i gx The value function of .
[0052] Compared with the prior art, the present invention has at least the following beneficial effects:
[0053] The present invention proposes a predictive control method for a grid-connected inverter based on model-free prediction and parameter-free reference. The model-free prediction is implemented based on a lookup table, a single voltage vector is output in each control cycle, and the current gradient and the voltage gradient are calculated using sampled bridge arm side current data, capacitor voltage data and grid-connected current data, and stored and updated in the lookup table to calculate the grid-connected current prediction value, capacitor voltage prediction value and bridge arm side current prediction value at the next moment; further, the method is also based on the amplitude-phase relationship between the inductor current and the capacitor voltage in the grid-connected inverter, and the system parameters in the equation are eliminated by jointly establishing the reference value amplitude-phase equation and the sampled value amplitude-phase equation through a space vector cosine algorithm, thereby proposing a parameter-free grid-connected inverter reference value calculation method, solving the problem of reference value calculation dependence on model parameters in the traditional LCL type grid-connected inverter model predictive control method, and further improving the parameter robustness of the predictive control method. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 A topological diagram of a grid-connected inverter and a basic control block diagram thereof in an embodiment of the present invention;
[0055] Figure 2 A schematic diagram of real-time updating of voltage and current gradients in an embodiment of the present invention;
[0056] Figure 3a Schematic diagram of the relationship between the current gradient and the voltage vector coordinates in the embodiment of the present invention Figure 1 ;
[0057] Figure 3b Schematic diagram of the relationship between the current gradient and the voltage vector coordinates in the embodiment of the present invention Figure 2 ;
[0058] Figure 4a The phase relationship between the reference value and the sampling value in the embodiment of the present invention is Figure 1 ;
[0059] Figure 4b The phase relationship between the reference value and the sampling value in the embodiment of the present invention is Figure 2 ;
[0060] Figure 4c Figure 3 is the relationship between the reference value and the sampling value amplitude and phase in the embodiment of the present invention;
[0061] Figure 4d Figure 4 is a diagram showing the relationship between the reference value and the sampling value amplitude and phase in an embodiment of the present invention;
[0062] Figure 5 Schematic diagram of the process of calculating the parameter-free reference value of capacitor voltage in an embodiment of the present invention;
[0063] Figure 6Schematic diagram of the calculation process of the parameter-free reference value of the bridge arm side current in an embodiment of the present invention;
[0064] Figure 7a L is the traditional model prediction method f and L g Schematic diagram of grid-connected current THD when parameters are mismatched;
[0065] Figure 7b C and L in the traditional model prediction method g Schematic diagram of grid-connected current THD when parameters are mismatched;
[0066] Figure 7c L is the traditional model prediction method f and L g Schematic diagram of grid-connected current prediction error when parameters are mismatched;
[0067] Figure 7d C and L in the traditional model prediction method g Schematic diagram of grid-connected current prediction error when parameters are mismatched;
[0068] Figure 8a L in the embodiment of the present invention f and L g Schematic diagram of grid-connected current THD when parameters are mismatched;
[0069] Figure 8b In the embodiment of the present invention, C and L g Schematic diagram of grid-connected current THD when parameters are mismatched;
[0070] Figure 8c L in the embodiment of the present invention f and L g Schematic diagram of grid-connected current prediction error when parameters are mismatched;
[0071] Figure 8d In the embodiment of the present invention, C and L g Schematic diagram of grid-connected current prediction error when parameters are mismatched;
[0072] Figure 9a It is a simulation comparison diagram of the predicted value of the bridge arm side current in the traditional model prediction method and the prediction method provided by the present invention;
[0073] Figure 9b A simulation comparison diagram of capacitor voltage prediction values in a traditional model prediction method and the prediction method provided by the present invention;
[0074] Fig.9c A simulation comparison diagram of the grid-connected side current prediction value in the traditional model prediction method and the prediction method provided by the present invention;
[0075] Fig.10aA comparison diagram of the bridge arm side current reference value simulation in the traditional model prediction method and the prediction method provided by the present invention;
[0076] Fig.10b A comparison diagram of the capacitor voltage reference value simulation in the traditional model prediction method and the prediction method provided by the present invention;
[0077] Fig.11a This is an experimental comparison diagram of the traditional model prediction method when the grid-connected current reference value changes from 10A to 20A;
[0078] Fig.11b This is an experimental comparison diagram of the prediction method provided by the present invention when the grid-connected side current reference value changes from 10A to 20A;
[0079] Fig.12a This is an experimental comparison diagram of the combination of the reference value calculation method and the prediction value calculation method in the traditional model prediction method when the grid-connected current reference value is 10A;
[0080] Figure 12b This is an experimental comparison diagram of the combination of the reference value calculation method in the traditional model prediction method and the parameter-free reference value calculation method provided by the embodiment of the present invention when the grid-connected side current reference value is 10A;
[0081] Fig.12c This is an experimental comparison diagram of the combination of the model-free prediction value calculation method and the parameter-free reference value calculation method in the prediction method provided by the present invention when the grid-connected side current reference value is 10A;
[0082] Fig.13a A comparison diagram of grid-connected current THD under parameter mismatch conditions between the traditional model prediction method and the prediction method provided by the present invention when the grid-connected current reference value is 10A;
[0083] Fig.13b The figure is a comparison diagram of grid-connected current prediction errors of the traditional model prediction method and the prediction method provided by the present invention under the condition of parameter mismatch when the grid-connected current reference value is 10A. DETAILED DESCRIPTION
[0084] A predictive control method based on model-free prediction and parameter-free reference of the present invention will be described in more detail below in conjunction with the accompanying drawings, wherein preferred embodiments of the present invention are shown, and it should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being widely known to those skilled in the art, and not as a limitation of the present invention.
[0085] The present invention is described in more detail in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the accompanying drawings are in very simplified form and are not in exact proportions, and are only used to facilitate and clearly assist in illustrating the purpose of the embodiments of the present invention.
[0086] For LCL (Inductor-Capacitor-Inductor) voltage source grid-connected inverters, if traditional model predictive control is used, although the control is relatively simple, it requires accurate model parameters. When the model parameters do not match the control parameters, it is difficult to achieve good control of the output current. Therefore, some scholars have proposed a variety of model-free predictive control methods to eliminate the dependence of the prediction value calculation process on the LCL system parameters. However, the current model-free predictive control method for LCL type grid-connected inverters cannot get rid of the dependence on the LCL system parameters in the reference value calculation process. The calculation of the current reference value on the bridge arm side of the inverter requires the capacitor C in the LCL system and the inductor L on the grid side. g Parameters, when the system parameters do not match the model parameters, the current model-free predictive control method will still be affected by the parameter mismatch, the accuracy of predictive control will be reduced, and the system stability will be further endangered; therefore, the present invention introduces a specific process of a predictive control method based on model-free prediction and parameter-free reference through the following embodiments.
[0087] Embodiment 1
[0088] The present invention provides a predictive control method based on model-free prediction and parameter-free reference. Figure 1 , including the following steps:
[0089] S1, construct the gradient formula;
[0090] S2. According to the mathematical relationship between different voltage vectors, a gradient update formula is constructed, and data of the bridge arm side current gradient, capacitor voltage gradient, and grid-connected side current gradient corresponding to each voltage vector are updated according to the gradient update formula;
[0091] S3, performing current prediction and voltage prediction according to the updated gradient of each voltage vector to obtain a bridge arm side current prediction value, a capacitor voltage prediction value, and a grid-connected side current prediction value;
[0092] S4, obtaining a capacitor voltage reference value and a bridge arm side current reference value according to a parameter-free capacitor voltage reference value algorithm and a parameter-free bridge arm side current reference value algorithm;
[0093] S5. Calculate the value function values corresponding to the eight basic voltage vectors according to the value function, select the voltage vector with the smallest value function value as the optimal vector, and apply it in the next control cycle.
[0094] Before the step of constructing the gradient formula, the step also includes: fabc , three-phase capacitor voltage v cabc And the three-phase grid current i gabc Perform Clarke coordinate transformation to obtain the two-phase bridge arm current i in the αβ stationary coordinate system fαβ , two-phase capacitor voltage v cαβ And the two-phase grid-connected current i gαβ .
[0095] Furthermore, in step S1, the process of constructing the gradient formula includes:
[0096] The mathematical model of the grid-connected inverter in the αβ stationary coordinate system is discretized to obtain the prediction model. Based on the prediction model, the (k-1)th control cycle is calculated, that is, from the (k-1)th T s To the kth T s The voltage vector v acting during ix,m (k-1) (m∈{0,1,…,7}) corresponding to the bridge arm side current gradient △i fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k):
[0097] ;
[0098] Among them, v ix,m (k-1) (m∈{0,1,…,7}) is the voltage vector acting in the (k-1)th control cycle; △i fx,m (k) is v ix,m (k-1) The current gradient on the bridge arm side generated; △v cx,m (k) is v ix,m (k-1) The capacitor voltage gradient generated; △i gx,m (k) is v ix,m (k-1) The grid-connected current gradient generated.
[0099] In step S2, constructing the gradient update formula includes:
[0100] ;
[0101] ;
[0102] Where H is the gain coefficient, where v ix,n (k-1) (n∈{0,1,…,7} and n is not equal to m) is the voltage vector that is not applied in the (k-1)th control cycle; △i fx,n (k) is if v is usedix,n (k-1) The current gradient on the bridge arm side that will be generated; △v cx,n (k) is if v is used ix,n (k-1) The capacitor voltage gradient that will be generated; △i gx,n (k) is if v is used ix,n (k-1) The grid-connected current gradient that will be generated; v ix,n (k-2) is the voltage vector that is not applied in the (k-2)th control cycle; △i fx,n (k-1) is if v is used ix,n (k-2) The current gradient on the bridge arm side that will be generated; △v cx,n (k-1) is if v is used ix,n (k-2) The capacitor voltage gradient that will be generated; △i gx,n (k-1) is if v is used ix,n (k-2) The grid-connected current gradient that will be generated; v ix,m (k-2) is the voltage vector acting in the (k-2)th control cycle; △i fx,m (k-1) is if v is used ix,m (k-2) The current gradient on the bridge arm side that will be generated; △v cx,m (k-1) is if v is used ix,m (k-2) The capacitor voltage gradient that will be generated; △i gx,m (k-1) is if v is used ix,nm (k-2) The grid-connected current gradient that will be generated.
[0103] When the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is 0, there is no solution for the value of coefficient H. At this time, the gradient update formula is reconstructed as:
[0104] ;
[0105] Among them, the voltage vector v that is not used in the (k-1)th control cycle ix,n The current gradient △i on the bridge arm side corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) is equal to the voltage vector v used in the (k-1)th control cycle ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k).
[0106] When the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is not 0, the coefficient H is solved to be 1, and the gradient update formula is reconstructed as follows:
[0107] ;
[0108] Among them, the voltage vector v that is not used in the (k-1)th control cycle ix,n The current gradient △i on the bridge arm side corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) The voltage vector v used in the (k-1)th control cycle ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m The voltage vector v used in the (k) and (k-2) control cycles ix,m (k-2) corresponds to the bridge arm side current gradient △i fx,m (k-1), capacitor voltage gradient △v cx,m (k-1) and the grid-connected current gradient △i gx,m The voltage vector v that is not used in the (k-1) and (k-2) control cycles ix,n (k-2) corresponds to the bridge arm side current gradient △i fx,n (k-1), capacitor voltage gradient △v cx,n (k-1) and the grid-connected current gradient △i gx,n (k-1) is used to represent it.
[0109] In step S3, the process of performing current prediction and voltage prediction includes:
[0110] According to the updated gradient of each voltage vector, current prediction and voltage prediction are performed, and the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k) and the grid-connected current sampling value i gx (k) and the voltage vector v used in the (k-1)th control cycle in the lookup table ix,m The current gradient and voltage gradient corresponding to (k-1) are added to obtain the bridge arm side current, capacitor voltage and grid-connected side current of the grid-connected inverter system at the (k+1)th T s Predicted value of:
[0111] ;
[0112] Among them, the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1) and the predicted value of the grid-connected current i gx,m (k+1) by the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k), grid-connected current sampling value i gx (k) and the voltage vector v ix,m The updated bridge arm current gradient △i corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k) are added together.
[0113] Furthermore, the grid-connected current prediction value, capacitor voltage prediction value, and bridge arm current prediction value at the (k+2)th Ts are calculated to compensate for the calculation delay:
[0114] ;
[0115] Among them, (k+2) T s The predicted value of the bridge arm current i fx (k+2), capacitor voltage prediction value v cx (k+2) and the predicted value of the grid-connected current i gx (k+2) is composed of the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1) and the grid-connected current prediction value i gx,m (k+1) and the current gradient △i on the bridge arm side fx (k+1), capacitor voltage gradient △v cx (k+1) and the grid-connected current gradient △i gx (k+1) is added together.
[0116] Among them, △i fx (k+1), △v cx (k+1) and △i gx (k+1) are equal to △i fx (k), △v cx (k) and △i gx (k), △i fx (k), △v cx (k) and △igx (k) is the current gradient and voltage gradient corresponding to the 8 basic voltage vectors in the updated k-th LUT. This calculation process completely eliminates the impact of parameter mismatch on the prediction model.
[0117] In step S4, the parameter-free capacitor voltage reference value algorithm is:
[0118] ;
[0119] ;
[0120] Among them, △V α =v cα (k)-e gα (k), △V β =v cβ (k)-e gβ (k), v cαβ (k) is k T s The capacitor voltage sampling value at time e gαβ (k) is k T s The grid-connected segment voltage sampling value at time T g is the coefficient matrix of the parameter-free capacitor voltage reference algorithm.
[0121] In step S4, the parameter-free bridge arm side current reference value algorithm is:
[0122] ; ; ;
[0123] Among them, γ and η are weight coefficients, T c It is the coefficient matrix of the parameter-free bridge arm current reference value algorithm.
[0124] Further, in step S4, based on the sampled bridge arm side current, capacitor voltage, grid-connected side current, grid-connected side voltage and a given grid-connected side current reference value obtained by sampling, the reference value at time (k+2) is obtained according to the parameter-free capacitor voltage reference value algorithm and the parameter-free bridge arm side current reference value algorithm:
[0125] ;
[0126] Among them, i g ref (k+2), v c ref (k+2) and i f ref (k+2) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k+2); i gref (k), v c ref (k) and i f ref (k) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k); i g ref (k-1), v c ref (k-1) and i f ref (k-1) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k-1); i g ref (k-2), v c ref (k-2) and i f ref (k-2) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k-2).
[0127] In step S5, eight sets of current prediction values and voltage prediction values and the obtained reference values are substituted into the value function equation to obtain eight value function values, and the voltage vector with the smallest value function value is selected as the optimal vector to be used in the next control cycle;
[0128] Among them, the value function is:
[0129] ;
[0130] ;
[0131] Among them, λ f , c and g They are g f , g c and g The weight coefficient of g f , g c and g They are the bridge arm current i fx , capacitor voltage v cx And the grid-side current i gx The value function of .
[0132] The embodiment of the present invention provides a model-free predictive control method for a grid-connected inverter based on model-free prediction and parameter-free reference. The relationship between the current gradient on the bridge arm side of the LCL filter, the capacitor voltage gradient, the grid-connected current gradient and the inverter output voltage vector is analyzed, and a model-free prediction method based on current gradient and voltage gradient update is further proposed; at the same time, a parameter-free reference value calculation method based on the amplitude and phase relationship between current and voltage sampling values and current and voltage reference values is provided, which eliminates the dependence of reference value calculation on system parameters in traditional model predictive control and improves the robustness of predictive control.
[0133] Embodiment 2
[0134] The embodiment of the present invention provides a model-free predictive control method for a grid-connected inverter based on no parameter reference, specifically comprising the following steps:
[0135] Step 1: Establish the mathematical model of the LCL grid-connected inverter in the αβ stationary coordinate system, and obtain its eight basic voltage vectors according to the three-phase switch state of the two-level inverter, which are u 0 (0,0,0),u 1 (1,0,0),u 2 (1,1,0),u 3 (0,1,0),u 4 (0,1,1),u 5 (0,0,1),u 6 (1,0,1) and u 7 (1,1,1).
[0136] Step 2: Use the arm-side current gradient, capacitor voltage gradient and grid-connected side current gradient to replace the prediction model of the traditional two-level grid-connected inverter.
[0137] Step 3: According to the mathematical relationship between different voltage vectors, the data of the bridge arm side current gradient, capacitor voltage gradient and grid-connected side current gradient corresponding to each voltage vector are updated.
[0138] Step 4: Add the bridge arm side current gradient, capacitor voltage gradient and grid-connected side current gradient corresponding to the 8 basic voltage vectors to the sampled bridge arm side current, capacitor voltage and grid-connected side current for current prediction and voltage prediction, so as to obtain the bridge arm side current prediction value, capacitor voltage prediction value and grid-connected side current prediction value.
[0139] Step 5: Based on the sampled bridge arm side current, capacitor voltage, grid-connected side current and grid-connected side voltage and the given grid-connected side current reference value, the capacitor voltage reference value and the bridge arm side current reference value are calculated using the proposed parameter-free reference value setting method.
[0140] Step 6: Substitute 8 sets of current prediction values, voltage prediction values and the obtained reference values into the value function equation to obtain 8 value function values, select the voltage vector with the smallest value function value as the optimal voltage vector and act on the next control cycle of the LCL type grid-connected inverter.
[0141] For details, please continue to refer to Figure 1 In step 1, a mathematical model of the LCL grid-connected inverter in the αβ stationary coordinate system is established, and its eight basic voltage vectors are obtained according to the three-phase switch states of the two-level inverter, which are u 0 (0,0,0),u 1 (1,0,0),u 2 (1,1,0),u 3 (0,1,0),u 4 (0,1,1),u 5 (0,0,1),u 6 (1,0,1) and u 7 (1,1,1).
[0142] The sampled three-phase bridge arm current i fabc、 Three-phase capacitor voltage v cabc And the three-phase grid current i gabc Perform Clarke coordinate transformation to obtain i fαβ 、v cαβ and gαβ , as shown in Formula 1:
[0143] (1)
[0144] The mathematical model of the LCL type grid-connected inverter in the αβ stationary coordinate system is shown in Formula 2:
[0145] (2)
[0146] Among them, v iαβ = [v iα , v iβ ] T , v iα and v iβ Respectively represent the voltage components output by the grid-connected inverter on the α-axis and β-axis; i fαβ = [i fα , i fβ ] T ,i fα and i fβ They represent the bridge arm current components output by the grid-connected inverter on the α-axis and β-axis respectively; v cαβ = [v cα , v cβ ] T, v cα and v cβ Respectively represent the voltage components of the capacitor voltage output on the α-axis and β-axis; e gαβ = [e gα ,e gβ ] T , e gα and e gβ Respectively represent the voltage components of the grid voltage output on the α-axis and β-axis; i gαβ = [i gα , i gβ ] T ,i gα and i gβ They represent the grid-connected side current components output by the grid-connected inverter on the α-axis and β-axis respectively; L f represents the inductance of the bridge arm side of the LCL filter, C represents the capacitance of the LCL filter, and L g Indicates the grid-side inductance of the LCL filter.
[0147] Discretizing the above mathematical model can yield its prediction model, as shown in Formula 3-4:
[0148] (3)
[0149] (4)
[0150] Among them, x∈{α, β}, T s Indicates the control period.
[0151] In step 2, the arm-side current gradient, capacitor voltage gradient and grid-connected-side current gradient are used to replace the prediction model of the traditional two-level grid-connected inverter.
[0152] Based on the above formula (3), the (k-1)th control cycle can be further calculated, that is, from the (k-1)th T s To the kth T s The voltage vector v acting during ix,m (k-1) (m∈{0,1,…,7}) corresponding to the bridge arm side current gradient △i fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k), as shown in Formula 5:
[0153] (5)
[0154] Among them, △i fx,m (k), △v cx,m (k) and △i gx,m (k) is the voltage vector v applied in the (k-1)th control cycleix,m The current and voltage gradients generated by (k-1) are stored in the voltage vector v in the kth lookup table. ix,m (k-1) corresponds to the position; i fx (k) is the kth T s The bridge arm current value, i fx (k-1) is the (k-1)th T s The current value of the bridge arm side; v cx (k) is the kth T s The capacitor voltage value, v cx (k-1) is the (k-1)th T s Capacitor voltage value; i gx (k) is the kth T s The grid-connected current value, i gx (k-1) is the (k-1)th T s The grid-connected current value.
[0155] For further information, please refer to Figure 2 as well as Figure 3a - Figure 3b In step three, according to the mathematical relationship between different voltage vectors, the data of the bridge arm side current gradient, capacitor voltage gradient and grid-connected side current gradient corresponding to each voltage vector are updated.
[0156] It should be noted that in order to ensure the accuracy of the proposed model-free prediction algorithm, all voltage vectors v that are not applied in the (k-1)th control cycle are searched in the table. ix,n (k-1) (n≠m and n∈{0,1,2,……7}) corresponding to △i fx,n (k), △v cx,n (k) and △i gx,n (k) also needs to be updated.
[0157] Combined with the prediction model in formula (3), the gradient formula (5) can be rewritten as the following formula 6:
[0158] (6)
[0159] Among them, Ψ=Φ-E, E is the unit matrix, and the selected voltage vector v ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and grid-connected current gradient △i gx,m (k) can be used with (k-1)T s The bridge arm current sampling value i fx (k-1), capacitor voltage sampling value vcx (k-1), grid-connected current sampling value i fx (k-1), inverter output voltage sampling value v ix,m (k-1) and the grid voltage sampling value e gx (k-1) represents.
[0160] The unapplied voltage vector v ix,n Substitute (k-1) into the gradient update formula (6), then △i fx,n (k), △v cx,n (k) and △i gx,n (k) can be expressed as the following formula 7:
[0161] (7)
[0162] Among them, v ix,n (k-1) is the voltage vector that is not applied, △i fx,n (k), △v cx,n (k) and △i gx,n (k) is the bridge arm side current gradient, capacitor voltage gradient and grid-connected current gradient corresponding to the voltage vector that is not applied.
[0163] Voltage vector v ix,n (k-1) and v ix,m (k-1) In the (k-1)th T s and the (k-2)th T s The corresponding current gradient and voltage gradient relationship can be expressed as the following formula 8 and formula 9:
[0164] (8) (9)
[0165] Among them, the different voltage vectors v in the (k-1)th control cycle ix,n (k-1) and v ix,m The relationship between the gradients corresponding to (k-1) is shown in the above formula 8, which can be expressed by v ix,n (k-1) and v ix,m (k-1) is used to represent the different voltage vectors v in the (k-2)th control cycle. ix,n (k-2) and v ix,m The relationship between the gradients corresponding to (k-2) is shown in the above formula 9, which can be expressed by v ix,n (k-2) and v ix,m (k-2) is used to represent it.
[0166] Furthermore, the system parameter matrix Γ can be eliminated by dividing the above formula 8 and formula 9, which can be expressed as formula 10:
[0167] (10)
[0168] Among them, v ix,n (k-1) (n∈{0,1,…,7} and n is not equal to m) is the voltage vector that is not applied in the (k-1)th control cycle; △i fx,n (k) is if v is used ix,n (k-1) The current gradient on the bridge arm side that will be generated; △v cx,n (k) is if v is used ix,n (k-1) The capacitor voltage gradient that will be generated; △i gx,n (k) is if v is used ix,n (k-1) The grid-connected current gradient that will be generated; △i fx,n (k-1) is if v is used ix,n (k-2) The current gradient on the bridge arm side that will be generated; △v cx,n (k-1) is if v is used ix,n (k-2) The capacitor voltage gradient that will be generated; △i gx,n (k-1) is if v is used ix,n (k-2) The grid-connected current gradient that will be generated; v ix,m (k-2) is the voltage vector acting in the (k-2)th control cycle; △i fx,m (k-1) is the use of v ix,m (k-2) The current gradient on the bridge arm side generated; △v cx,m (k-1) is the use of v ix,m (k-2) Capacitor voltage gradient generated; △i gx,m (k-1) is the use of v ix,nm (k-2) The grid-connected current gradient generated.
[0169] Furthermore, formula 11 is derived:
[0170] (11)
[0171] Among them, the unused voltage vector v ix,n The current gradient △i on the bridge arm side corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and grid-connected current gradient △i gx,n (k) The voltage vector v ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and grid-connected current gradient △i gx,m (k) and historical gradient.
[0172] Further, the denominator value of coefficient H is discussed, and it can be found that the value of H can be divided into two cases:
[0173] In the first case, when the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is 0, that is, when v ix,n (k-2)=v ix,m Several voltage vectors v of (k-2) ix,n (k-1) and v ix,m (k-1) selection case (respectively v in α axis iα,0 =v iα,4 、v iα,2 =v iα,6 and v iα,3 =v iα,5 and v in the β axis iβ,0 =v iβ。1 =v iβ,4 =v iβ,7 、v iβ,2 =v iβ,3 、v iβ,5 =v iβ,6 ), which will make the value of H unsolvable. At this time, the gradient update formula can be reconstructed as the following formula 12:
[0174] (12)
[0175] Among them, the voltage vector v that is not used in the (k-1)th control cycle ix,n The current gradient △i on the bridge arm side corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) is equal to the voltage vector v used in the (k-1)th control cycle ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k).
[0176] In the second case, when the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is not 0, the coefficient H can be solved to be 1. At this time, the gradient update formula can be reconstructed as the following formula 13:
[0177] (13)
[0178] Among them, the voltage vector v that is not used in the (k-1)th control cycle ix,n The current gradient △i on the bridge arm side corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) The voltage vector v used in the (k-1)th control cycle ix,m The current gradient △i on the bridge arm side corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m The voltage vector v used in the (k) and (k-2) control cycles ix,m The bridge arm current gradient △i corresponding to (k-2) fx,m (k-1), capacitor voltage gradient △v cx,m (k-1) and the grid-connected current gradient △i gx,m The voltage vector v that is not used in the (k-1) and (k-2) control cycles ix,n The bridge arm current gradient △i corresponding to (k-2) fx,n (k-1), capacitor voltage gradient △v cx,n (k-1) and the grid-connected current gradient △i gx,n (k-1) is used to represent it.
[0179] Through the above gradient update formulas (5), (12) and (13), all basic voltage vectors v in the kth lookup table are ix The corresponding current gradient and voltage gradient can be updated.
[0180] Furthermore, in step four, the bridge arm side current gradient, capacitor voltage gradient and grid-connected side current gradient corresponding to the 8 basic voltage vectors are added to the sampled bridge arm side current, capacitor voltage and grid-connected side current for current prediction and voltage prediction, thereby obtaining the bridge arm side current prediction value, capacitor voltage prediction value and grid-connected side current prediction value.
[0181] Specifically, after completing the update of all gradients in the lookup table, the current and voltage are first predicted based on the updated gradient of each voltage vector, using the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k) and the grid-connected current sampling value i gx (k) and the voltage vector v used in the (k-1)th control cycle in the lookup table ix,mThe current gradient and voltage gradient corresponding to (k-1) are added to obtain the bridge arm current, capacitor voltage and grid-connected current of the LCL grid-connected inverter system at the (k+1)th T s The predicted value is shown in the following formula 14:
[0182] (14)
[0183] Among them, the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1) and the predicted value of the grid-connected current i gx,m (k+1) by the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k), grid-connected current sampling value i gx (k) and the voltage vector v ix,m The updated bridge arm current gradient △i corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k) are added together.
[0184] In order to compensate for the computational delay in predictive control, the present invention performs a s The grid-connected current prediction value, capacitor voltage prediction value and bridge arm current prediction value are calculated as shown in the following formula 15:
[0185] (15)
[0186] Among them, (k+2) T s The predicted value of the bridge arm current i fx (k+2), capacitor voltage prediction value v cx (k+2) and the predicted value of the grid-connected current i gx (k+2) is composed of the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1), grid-connected current prediction value i gx,m (k+1) and the current gradient △i on the bridge arm side fx (k+1), capacitor voltage gradient △v cx (k+1) and the grid-connected current gradient △i gx (k+1) are added together, and their values are equal to △i fx (k), △v cx(k) and △i gx (k), △i fx (k), △v cx (k) and △i gx (k) is the current gradient and voltage gradient corresponding to the 8 basic voltage vectors in the updated k-th LUT. This calculation process completely eliminates the impact of parameter mismatch on the prediction model.
[0187] Further, in step 5, please refer to Figure 4a - Figure 4d Based on the sampled bridge arm side current, capacitor voltage, grid-connected side current, grid-connected side voltage and the given grid-connected side current reference value, the proposed parameter-free reference value setting method is used to calculate the capacitor voltage reference value and the bridge arm side current reference value.
[0188] Please refer to Figure 5 - Figure 6 Specifically, in the model predictive control process of the traditional LCL type grid-connected inverter, the calculation of the grid-connected side current reference value, the capacitor voltage reference value and the bridge arm side current reference value is shown in the following formula 16:
[0189] (16)
[0190] Among them, I g ref is the given grid-connected current reference amplitude, ω g is the system angular frequency, L g is the grid-side inductance of the LCL filter, and C is the filter capacitance of the LCL filter.
[0191] On the other hand, the capacitor voltage reference value v is established by the space vector cosine algorithm ref c (k), grid-connected voltage sampling value e g (k) and grid-connected current reference value i ref g (k) The amplitude-phase relationship in the αβ stationary coordinate system can be obtained from formula 17:
[0192] (17)
[0193] Among them, F g is the system parameter gain, △θ ref c is the grid-side voltage e g (k) and capacitor voltage reference value v ref c (k) The phase difference between them.
[0194] Furthermore, formula 18 is the phase difference expression:
[0195] (18)
[0196] Among them, cos△θ ref c is the grid voltage e in the constructed phase equation g (k) and capacitor voltage reference value v ref c (k) is the cosine of the phase difference between them.
[0197] Similarly, the capacitor voltage sampling value v can be constructed c (k), grid-connected voltage sampling value e g (k) and the grid-connected current sampling value i g (k) The amplitude-phase relationship in the αβ stationary coordinate system is shown in Formula 19:
[0198] (19)
[0199] Among them, F g is the system parameter gain. Since the physical structure of the LCL filter is fixed, F g is considered as a constant variable.
[0200] Furthermore, Δθ c is the grid side voltage sampling value e g (k) and the capacitor voltage sampling value v c The phase difference between (k) and (k) can be expressed as Equation 20:
[0201] (20)
[0202] Among them, cos△θ c is the grid voltage e in the constructed phase equation g (k) and the capacitor voltage sampling value v c (k) is the cosine of the phase difference between them.
[0203] Based on equations (17) to (20), the system parameter gain F g Eliminate and obtain the capacitor voltage reference value, thus completely eliminating the influence of parameter mismatch on the calculation of the capacitor voltage reference value. For details, see the following formulas 21-22:
[0204] (twenty one)
[0205] (twenty two)
[0206] Among them, △V α =v cα (k)-e gα (k), △Vβ =v cβ (k)-e gβ (k), T g is the coefficient matrix of the parameter-free capacitor voltage reference algorithm.
[0207] The capacitor voltage reference value v is established by space vector cosine algorithm ref c (k), grid-connected current reference value i ref g (k) and the bridge arm current reference value i ref f (k) The amplitude-phase relationship in the αβ stationary coordinate system can be obtained from formula 23:
[0208] (twenty three)
[0209] Among them, F c is the system parameter gain, △θ ref f is the grid-side current reference value i ref g (k) and the bridge arm current reference value i ref f The phase difference between (k) and (k) can be expressed as Equation 24:
[0210] (twenty four)
[0211] Among them, cos△θ ref f is the bridge arm current reference value i in the constructed phase equation ref f (k) and grid-connected current reference value i ref g (k) is the cosine of the phase difference between them.
[0212] Similarly, the capacitor voltage sampling value v can be constructed c (k), grid-connected current sampling value i g (k) and the current value i on the bridge arm side f (k) The amplitude-phase relationship in the αβ stationary coordinate system is shown in the following formula 25:
[0213] (25)
[0214] Among them, △θ f is the grid-side current sampling value i g (k) and the bridge arm current sampling value i f The phase difference between (k) and (k) can be expressed as the following formula 26:
[0215] (26)
[0216] Among them, cos△θ f is the bridge arm current sampling value i in the constructed phase equation f (k) and the grid-connected current sampling value i g (k) is the cosine of the phase difference between them.
[0217] According to equations (21) and (23), the system parameter gain F c Eliminate, and the current reference value of the bridge arm side can be obtained, thereby completely eliminating the influence of parameter mismatch on the calculation of the current reference value of the bridge arm side, as shown in the following formulas 27-29:
[0218] (27)
[0219] (28)
[0220] (29)
[0221] Among them, γ and η are weight coefficients, T c It is the coefficient matrix of the parameter-free bridge arm current reference value algorithm.
[0222] Furthermore, the reference value at time (k+2) can be obtained by Laplace recursion method, as shown in the following formula 30:
[0223] (30)
[0224] Among them, i g ref (k+2), v c ref (k+2) and i f ref (k+2) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k+2); i g ref (k), v c ref (k) and i f ref (k) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k); i g ref (k-1), v c ref (k-1) and i f ref (k-1) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k-1); i gref (k-2), v c ref (k-2) and i f ref (k-2) is the grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value at time (k-2).
[0225] Furthermore, in step 6, in order to ensure that the predicted value i in the proposed model-free prediction fx (k+2), v cx (k+2) and i gx (k+2) can accurately track the reference value i obtained from the proposed parameter-free reference value ref fx (k+1),v ref cx (k+1) and i ref gx (k+1), the present invention designs the following value function g CF , as shown in the following formula 31-32:
[0226] (31)
[0227] (32)
[0228] Among them, λ f , c and g They are g f , g c and g The weight coefficient of g f , g c and g They are the bridge arm current i fx , capacitor voltage v cx And the grid-side current i gx The value function of .
[0229] The value function values corresponding to the eight basic voltage vectors are calculated according to the value function formula (31), and the voltage vector with the smallest value function value is selected as the optimal vector to be used in the next control cycle.
[0230] In this embodiment, the present invention adopts a model-free predictive control method based on a lookup table, which can effectively improve the robustness of predictive control to model parameter changes; by increasing the number of lookup tables and analyzing the mathematical relationship between the gradients of the three control quantities, a full update formula for the current gradient and voltage gradient of the LCL grid-connected inverter is constructed, and the model-free prediction and multi-objective control of the grid-connected current, capacitor voltage and bridge arm side current of the LCL grid-connected inverter are realized, and the dependence of the predicted value calculation on parameters in the traditional model predictive control algorithm is eliminated. In addition, the scheme of the present invention analyzes the amplitude and phase relationship between the voltage-current sampling value and the voltage-current reference value through the space vector cosine algorithm, and proposes a parameter-free reference value calculation method, which realizes the parameter-free calculation of the bridge arm side current reference value, capacitor voltage reference value and grid current reference value of the LCL grid-connected inverter, and effectively eliminates the dependence of the reference value calculation on parameters in the traditional LCL grid-connected inverter control algorithm.
[0231] Furthermore, a specific simulation is provided in the embodiment of the present invention:
[0232] The effectiveness of the embodiment of the present invention is verified in Matlab / Simulink simulation, and the system parameters are shown in Table 1.
[0233]
[0234] First, in order to demonstrate the robustness of the proposed MFPC (Model Predictive Power Control) strategy under parameter mismatch, simulation verification is carried out under the condition of LCL parameter mismatch and compared with the single-vector MPC (Model Predictive Control) in the literature [P. Falkowski and A. Sikorski, “Finite Control Set Model Predictive Controlfor Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.]. Figure 7a - Figure 7d as well as Figure 8a - Figure 8d When the reference current is 10A, the current prediction error and THD (Total Harmonic Distortion) of the above two control methods are as follows.
[0235] Figure 7a - Figure 7d as well as Figure 8a - Figure 8d The three-phase voltage waveforms and their prediction error waveforms of the two methods when the inductance parameters are mismatched are shown respectively. Figure 7a - Figure 7d It can be seen that in the literature [C. Zheng, T. , Z.Zhang, J. Rodriguez and F. Blaabjerg, "Model Predictive Control of LC-Filtered Voltage Source Inverters With Optimal Switching Sequence," IEEE Trans. Power Electron., vol. 36, no. 3, pp. 3422-3436, March 2021.] In the traditional model prediction method, since its voltage prediction and vector time calculation depend on the inductor parameters, when the inductor control parameters do not match the actual system parameters, the voltage ripple and prediction error increase.
[0236] Figure 8a - Figure 8d The simulation results of the data-driven predictive control method provided in the embodiment of the present invention are shown. It can be seen that the voltage performance is not affected by the inductance parameter mismatch, thereby improving the parameter robustness of the system control.
[0237] Therefore, when the filter bridge arm side inductance L f As well as when the actual value of the capacitor C does not match its model value, the proposed MFPC has better robustness in terms of low current distortion (see THD value) and low prediction error compared to the conventional single-vector MPC without the need for model parameters.
[0238] In addition, in order to prove the effectiveness of the proposed invention in predicted value tracking and reference value calculation, the predicted value calculation method and the reference value calculation method in the traditional model predictive control algorithm are compared with the proposed invention. Figure 9a - Fig.9c and Fig.10a - Fig.10b They are the predicted value waveform and reference value waveform of the two methods respectively.
[0239] Figure 9a - Fig.9c: is a simulation comparison diagram of the predicted values of the traditional method and the control method provided by the present embodiment when the grid-connected current reference value is 10A and the actual filter inductance matches the controller inductance parameters; wherein the traditional predictive control is a single vector MPC in the literature [P. Falkowski and A. Sikorski, “Finite Control Set Model Predictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.], Figure 9a is the predicted value of the bridge arm current of the two methods, Figure 9b are the capacitor voltage prediction values of the two methods, Fig.9c are the grid-connected current prediction values of the two methods.
[0240] like Figure 9a - Fig.9c As shown in the figure, the predicted values of bridge arm current, capacitor voltage and grid-connected current generated by the proposed model-free prediction method are very close to those of the model prediction method in the literature [P. Falkowski and A.Sikorski, “Finite Control Set Model Predictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics,vol. 65, no. 4, pp. 2844-2852, April 2018.], which reflects the effectiveness and accuracy of the proposed model-free prediction calculation method.
[0241] Fig.10a - Fig.10b: is a reference value simulation comparison diagram of the traditional method and the control method provided by this embodiment when the grid-connected current reference value is 10A and the actual filter inductance matches the controller inductance parameter; wherein the traditional predictive control is a single vector MPC in the literature [P. Falkowski and A. Sikorski, “Finite Control Set Model Predictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.], Fig.10a is the bridge arm current reference value of the two methods, Fig.10b are the capacitor voltage reference values for the two methods.
[0242] like Fig.10a - Fig.10b As shown in the figure, the bridge arm current reference value and capacitor voltage reference value generated by the proposed parameter-free reference value calculation method are very close to those in the reference value calculation method in the literature [P. Falkowski and A. Sikorski, “Finite Control Set Model Predictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.], which reflects the effectiveness and accuracy of the proposed parameter-free reference value calculation method.
[0243] In addition, this embodiment also provides a specific comparative experiment, and the experimental parameters are consistent with the simulation parameters.
[0244] First, to demonstrate the effectiveness of the proposed MFPC, its steady-state performance and dynamic response were experimentally tested and compared with the single-vector MPC in the traditional literature [P. Falkowski and A. Sikorski, “Finite Control Set ModelPredictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEETransactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April2018.]. Fig.11a - Fig.11b : The current waveform, prediction error and THD (Total Harmonic Distortion) of the above two control methods when the reference current changes from 10A to 20A.
[0245] from Fig.11a - Fig.11b It can be seen that for the single vector MPC method, the THD of the output current changes from 2.36% to 1.41%. Compared with the traditional scheme, the control method proposed in the present invention has similar performance to the single vector MPC method, and the THD changes from 3.06% to 1.94%.
[0246] Fig.12a - Fig.12c When the grid-connected current reference value is 10A, the LCL filter bridge arm side inductance controller parameters, capacitor controller parameters and grid-connected side inductance controller parameters are 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.8 and 2.0 times of the actual parameters respectively, the experimental comparison diagram of different control methods; Among them, Fig.12a The reference value calculation method and the prediction value calculation method in the literature [P. Falkowskiand A. Sikorski, “Finite Control Set Model Predictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.] are combined. Figure 12bThe prediction value calculation method in the literature [P. Falkowski and A. Sikorski, “Finite ControlSet Model Predictive Control for Grid-Connected AC–DC Converters With LCLFilter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp.2844-2852, April 2018.] is combined with the parameter-free reference value calculation method in the method of the present invention. Fig.12c It is a combination of the model-free prediction value calculation method and the parameter-free reference value calculation method in the method of the present invention.
[0247] Fig.13a - Fig.13b When the grid-connected current reference value is 10A, the inductance controller parameters, capacitor controller parameters and grid-connected inductance controller parameters of the LCL filter bridge arm are 0.2, 0.6, 1.0, 1.4 and 2.0 times of the actual parameters, respectively. The three control methods are: the reference value calculation method in the literature [P. Falkowskiand A. Sikorski, “Finite Control Set Model Predictive Control for Grid-Connected AC–DC Converters With LCL Filter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.] and its prediction value calculation method combined; the reference value calculation method in the literature [P. Falkowski and A. Sikorski, “Finite Control SetModel Predictive Control for Grid-Connected AC–DC Converters With LCLFilter,” IEEE Transactions on Industrial Electronics, vol. 65, no. 4, pp. 2844-2852, April 2018.] and its prediction value calculation method combined; pp.2844-2852, April 2018.] is combined with the parameter-free reference value calculation method proposed in the embodiments of the present invention; and the model-free prediction method proposed in the embodiments of the present invention is combined with the parameter-free reference value calculation method.
[0248] also, Fig.12a - Fig.12c and Fig.13a - Fig.13b The experimental comparison when the parameters do not match is given. f , filter capacitor C and grid-connected inductor L g When the parameters are set to 0.2, 0.6, 1.0, 1.4 and 2.0 times of the actual parameter values, the grid-connected current THD of the single-vector MPC method in the literature [P. Falkowski and A. Sikorski, “Finite Control SetModel Predictive Control for Grid-Connected AC–DC Converters With LCLFilter,” IEEE Transactions on Industrial Electronics, vol. 65, no.4, pp.2844-2852, April 2018.] are 25.93%, 3.75%, 2.42%, 3.57% and 4.21% respectively; the grid-connected current THD of the combination of the traditional model prediction algorithm and the proposed parameter-free reference value calculation method are 24.2%, 3.42%, 2.98%, 3.25% and 4.05% respectively; and the grid-connected current THD of the proposed method is 3.06%. The experimental results show that the proposed invention has stronger parameter robustness than the traditional single-vector MPC method.
[0249] In order to enhance the parameter robustness of MPC and reduce the dependence of prediction value calculation and reference value calculation on LCL parameters, the present invention proposes a model-free prediction current control strategy for LCL-type grid-connected inverter based on model-free prediction and parameter-free reference value; based on the relationship between the bridge arm side current gradient, capacitor voltage gradient, grid-connected side current gradient and inverter voltage vector, a model-free prediction calculation method is proposed, which eliminates the dependence of the traditional MPC calculation prediction value on parameters; a parameter-free reference value calculation method based on the amplitude-phase relationship between the reference value and the sampled value is also proposed, and the relationship between the bridge arm side current reference value and the sampled value, the capacitor voltage reference value and the sampled value, and the grid-connected side current reference value and the sampled value is analyzed to construct a parameter-free reference value calculation equation, eliminating the dependence of the traditional MPC calculation reference value on parameters. The invention enhances the parameter robustness of the traditional single-vector MPC scheme, and achieves control performance similar to that of the traditional single-vector MPC when the model parameters are accurate. The control method adopted by the present invention can well solve the problem that the traditional model prediction control method has strong dependence on model parameters, and is particularly suitable for grid-connected power generation systems with frequent line impedance changes.
[0250] Obviously, those skilled in the art can make various changes and modifications 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 equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A predictive control method based on model-free prediction and parameter-free reference, characterized in that: include: The mathematical model of the grid-connected inverter in the αβ stationary coordinate system is established, and the sampled three-phase bridge arm side current i fabc , three-phase capacitor voltage v cabc And the three-phase grid current i gabc Perform Clarke coordinate transformation to obtain the two-phase bridge arm current i in the αβ stationary coordinate system fαβ , two-phase capacitor voltage v cαβ And the two-phase grid current i gαβ ; Constructing gradient formulas; Based on the mathematical relationship between different voltage vectors, a gradient update formula is constructed. According to the gradient update formula, the data of the bridge arm side current gradient, capacitor voltage gradient and grid side current gradient corresponding to each voltage vector are updated; Perform current and voltage prediction based on the updated gradient of each voltage vector to obtain the predicted current value on the bridge arm side, the predicted capacitor voltage value, and the predicted current value on the grid side; The capacitor voltage reference value and the bridge arm side current reference value are obtained according to a parameter-free capacitor voltage reference value algorithm and a parameter-free bridge arm side current reference value algorithm; According to the value function, the value function values corresponding to the eight basic voltage vectors are calculated respectively, and the voltage vector with the smallest value function value is selected as the optimal vector to be used in the next control cycle; The parameter-free capacitor voltage reference value algorithm is: ; ; Among them, e gαβ (k) is the kth T s The grid-connected segment voltage sampling value at time i ref gαβ (k) is the kth T s The given two-phase grid-connected current reference value, v ref cαβ (k) is the kth T s The two-phase capacitor voltage reference value at time T g is the coefficient matrix of the parameter-free capacitor voltage reference value algorithm, △V α =v cα (k)-e gα (k), △V β =v cβ (k)-e gβ (k), v cαβ (k) is the kth T s The capacitor voltage sampling value at time i gαβ (k) is the kth T s The two-phase grid-connected current sampling value at time ; The parameter-free bridge arm side current reference value algorithm is: ; ; ; Among them, i ref fαβ (k) is the kth T s The given two-phase bridge arm current reference value, v ref cαβ (k) is the kth T s The two-phase capacitor voltage reference value at time i ref gαβ (k) is the kth T s The given two-phase grid side current reference value, v cαβ (k) is the kth T s The capacitor voltage sampling value at time i gαβ (k) is the kth T s The two-phase grid-connected current sampling value at time i fαβ (k) is the kth T s The two-phase bridge arm current sampling value at time T c is the coefficient matrix of the parameter-free bridge arm side current reference value algorithm, and γ and η are weight coefficients.
2. The predictive control method based on model-free prediction and parameter-free reference according to claim 1, characterized in that: The process of constructing the gradient formula includes: The mathematical model of the grid-connected inverter in the αβ stationary coordinate system is discretized to obtain the prediction model. Based on the prediction model, the (k-1)th control cycle is calculated, that is, from the (k-1)th T s To the kth T s The voltage vector v acting during ix,m (k-1), m∈{0,1,…,7} corresponding to the bridge arm side current gradient △i fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k): ; Among them, v ix,m (k-1), m∈{0,1,…,7} is the voltage vector acting in the (k-1)th control cycle; △i fx,m (k) is v ix,m (k-1) the bridge arm side current gradient; △v cx,m (k) is v ix,m (k-1) The capacitance voltage gradient generated; △i gx,m (k) is v ix,m (k-1) The grid-connected current gradient generated.
3. The predictive control method based on model-free prediction and parameter-free reference according to claim 2, characterized in that: The gradient update formula is constructed as follows: ; ; Among them, H is the gain coefficient, where v ix,n (k-1), n∈{0,1,…,7} and n is not equal to m, is the voltage vector that is not applied in the (k-1)th control cycle; △i fx,n (k) is if v is used ix,n (k-1) will generate the current gradient on the bridge arm side; △v cx,n (k) is if v is used ix,n (k-1) The capacitor voltage gradient that will be generated; △i gx,n (k) is if v is used ix,n (k-1) The grid-connected current gradient that will be generated; v ix,n (k-2) is the voltage vector that does not act in the (k-2)th control cycle; △i fx,n (k-1) is if v is used ix,n (k-2) will generate the bridge arm side current gradient; △v cx,n (k-1) is if v is used ix,n (k-2) The resulting capacitor voltage gradient; △i gx,n (k-1) is if v is used ix,n (k-2) The grid-connected current gradient that will be generated; v ix,m (k-2) is the voltage vector acting in the (k-2)th control cycle; △i fx,m (k-1) is if v is used ix,m (k-2) will generate the bridge arm side current gradient; △v cx,m (k-1) is if v is used ix,m (k-2) The resulting capacitor voltage gradient; △i gx,m (k-1) is if v is used ix,nm (k-2) The grid-side current gradient that will be generated; When the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is 0, there is no solution for the value of coefficient H. The gradient update formula is reconstructed as follows: ; Among them, the voltage vector v that is not used in the (k-1)th control cycle is ix,n The current gradient on the bridge arm side corresponding to (k-1) is △i fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) is equal to the voltage vector v used in the (k-1)th control cycle ix,m The bridge arm current gradient △i corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m (k); When the denominator v in H ix,n (k-2)-v ix,m When the value of (k-2) is not 0, the coefficient H is solved to 1, and the gradient update formula is reconstructed as follows: ; Among them, the voltage vector v that is not used in the (k-1)th control cycle is ix,n The bridge arm current gradient △i corresponding to (k-1) fx,n (k), capacitor voltage gradient △v cx,n (k) and the grid-connected current gradient △i gx,n (k) The voltage vector v used in the (k-1)th control cycle ix,m The bridge arm current gradient △i corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and the grid-connected current gradient △i gx,m The voltage vector v used in the (k) and (k-2) control cycles ix,m (k-2) corresponds to the bridge arm side current gradient △i fx,m (k-1), capacitor voltage gradient △v cx,m (k-1) and the grid-connected current gradient △i gx,m The voltage vector v that is not used in the (k-1) and (k-2) control cycles ix,n (k-2) corresponds to the bridge arm side current gradient △i fx,n (k-1), capacitor voltage gradient △v cx,n (k-1) and the grid-connected current gradient △i gx,n (k-1) is used to represent it.
4. The predictive control method based on model-free prediction and parameter-free reference according to claim 3, characterized in that: The process of current prediction and voltage prediction includes: According to the updated gradient of each voltage vector, current prediction and voltage prediction are performed, and the kth T s The bridge arm current sampling value i fx (k), capacitor voltage sampling value v cx (k) and the grid-connected current sampling value i gx (k) and the voltage vector v used in the (k-1)th control cycle in the lookup table ix,m The current gradient and voltage gradient corresponding to (k-1) are added to obtain the bridge arm side current, capacitor voltage and grid side current of the grid-connected inverter system at the (k+1)th T s The predicted value of: ; Among them, the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1) and the predicted value of the grid-connected current i gx,m (k+1) by the kth T s The bridge arm side current sampling value i fx (k), capacitor voltage sampling value v cx (k) and the grid-connected current sampling value i gx (k) and the voltage vector v ix,m The updated bridge arm current gradient △i corresponding to (k-1) fx,m (k), capacitor voltage gradient △v cx,m (k) and grid-connected current gradient △i gx,m (k) are added together.
5. The predictive control method based on model-free prediction and parameter-free reference according to claim 4, characterized in that: For the (k+2)th T s The grid-connected current prediction value, capacitor voltage prediction value, and bridge arm current prediction value are calculated to compensate for the calculation delay: ; Among them, (k+2) T s The predicted value of the bridge arm current i fx (k+2), capacitor voltage prediction value v cx (k+2) and the predicted value of the grid-connected current i gx (k+2) is composed of the (k+1)th T s The predicted value of the bridge arm current i fx,m (k+1), capacitor voltage prediction value v cx,m (k+1) and the grid-connected current prediction value i gx,m (k+1) and the bridge arm current gradient △i fx (k+1), capacitor voltage gradient △v cx (k+1) and the grid-connected current gradient △i gx (k+1) is added together.
6. The predictive control method based on model-free prediction and parameter-free reference according to claim 5, characterized in that: Based on the sampled bridge arm side current, capacitor voltage, grid side current, grid side voltage and the given grid side current reference value, the (k+2)th T is obtained according to the parameter-free capacitor voltage reference value algorithm and the parameter-free bridge arm side current reference value algorithm. s The reference value is: ; Among them, i g ref (k+2), v c ref (k+2) and i f ref (k+2) is the (k+2)th T s The grid-connected current reference value, capacitor voltage reference value and bridge arm current reference value at the time of i g ref (k), v c ref (k) and i f ref (k) is the (k)th T s The grid-connected current reference value, capacitor voltage reference value and bridge arm current reference value at the time of i g ref (k-1), v c ref (k-1) and i f ref (k-1) is the (k-1)th T s The grid-connected current reference value, capacitor voltage reference value and bridge arm current reference value at the time of i g ref (k-2), v c ref (k-2) and i f ref (k-2) is the (k-2)th T s The grid-connected current reference value, capacitor voltage reference value and bridge arm side current reference value are shown in the figure.
7. The predictive control method based on model-free prediction and parameter-free reference according to claim 6, characterized in that: Substitute the eight sets of current prediction values, voltage prediction values, and the obtained reference values into the value function equation to obtain eight value function values. Select the voltage vector with the smallest value function value as the optimal vector and apply it to the next control cycle. Among them, the value function is: ; ; Among them, λ f ,λ c and λ g They are g f 、g c and g g The weight coefficient of g f 、g c and g g They are the bridge arm side current i fx , capacitor voltage v cx And the grid-side current i gx The value function of .
