Grid-connected inverter model prediction control current spectrum dispersion optimization method

Abstract

The invention relates to the technical field of model predictive control, and discloses a grid-connected inverter model predictive control current spectrum dispersion optimization method, computer equipment and a computer readable storage medium in order to solve the problem of current spectrum dispersion caused by unfixed switching frequency in three-phase grid-connected inverter model predictive control. The method comprises the following steps: determining a current constraint cost function based on a mathematical model of a three-phase grid-connected inverter and a switching state of each phase bridge arm of the inverter at the current moment; establishing a period constraint cost function according to the expected average switching period and the time interval of two times of continuous upward and downward commutation of the switching state of each phase bridge arm of the inverter; combining the current constraint cost function and the period constraint cost function to obtain a new cost function; and determining the switching state applied to the inverter at the next moment. By adopting the method, the problem of current spectrum dispersion caused by unfixed switching frequency in three-phase grid-connected inverter model prediction control is solved, and the current quality is improved.

Description

Technical Field

[0001] The present application relates to the technical field of model predictive control, and more specifically, to a method for optimizing the current spectrum dispersion of model predictive control of a grid-connected inverter, a computer device, and a computer-readable storage medium. Background Art

[0002] Model Predictive Control (MPC for short) is to establish a dynamic model of a system and use this model at each control moment to predict the behavior of the system at future moments. Based on these predictions, an optimal control sequence can be generated, and then the system state is adjusted by executing the first control action, and then recalculated and executed at the next moment. This process is repeated to enable the system to optimize a specific performance index within a certain period in the future.

[0003] Model predictive control has the control characteristics of multi-objective, multi-variable and multi-constraint, as well as an intuitive and simple design concept, and is widely used in motor drive systems.

[0004] In the current model predictive methods applied to three-phase grid-connected inverters, there is still a problem of current spectrum dispersion caused by the non-fixed frequency of the switching signals of the inverter. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides a method for optimizing the current spectrum dispersion of model predictive control of a grid-connected inverter, a computer device, and a computer-readable storage medium, which will overcome the problem of current spectrum dispersion caused by the non-fixed switching frequency in the model predictive control of three-phase grid-connected inverters, improve the current quality, and at the same time will not affect other control objectives such as response time and calculation burden.

[0006] To achieve the above object, according to the first aspect of the present invention, there is provided a method for optimizing the current spectrum dispersion of model predictive control of a grid-connected inverter, the method comprising: Determine a current constraint cost function based on the mathematical model of the three-phase grid-connected inverter and the switching states of each phase leg of the inverter at the current moment; Establish a period constraint cost function according to the expected average switching period and the time intervals of two consecutive upward and downward commutations of the switching states of each phase leg of the inverter; divide the period constraint cost function by the sampling period to establish a simplified period constraint cost function, and obtain an integer matrix for upward commutation and an integer matrix for downward commutation; Respectively use the switching signals corresponding to multiple candidate voltage vectors as the predicted values of the switching states of each phase leg of the inverter at the next moment of the current moment; Calculate the integer matrices for upward commutation and the integer matrices for downward commutation corresponding to multiple candidate voltage vectors based on the predicted values of each switch state, the switch state at the current moment, the integer matrix for upward commutation at the current moment, and the integer matrix for downward commutation at the current moment. Perform a weighted sum of the current constraint cost function and the simplified period constraint cost function to obtain a new cost function for the mathematical model. Substitute the integer matrices for upward commutation and the integer matrices for downward commutation corresponding to multiple candidate voltage vectors into the new cost function for calculation, and determine the switch states of each phase leg of the inverter to be applied at the next moment.

[0007] According to the second aspect of the present invention, there is also provided a computer device, which includes a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of any one of the above methods.

[0008] According to the third aspect of the present invention, there is also provided a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above methods are implemented.

[0009] Generally speaking, compared with the prior art by the above technical solutions conceived by the present invention, the following beneficial effects can be achieved: (1) An optimized method for model predictive control of grid-connected inverter current spectrum dispersion provided by the present invention adds a period constraint cost function to the traditional current constraint-based cost function, that is, a period constraint cost function is established through the time interval between two consecutive identical commutations of the switch state of each phase leg of the inverter, and the period constraint cost function is divided by the sampling period to establish a simplified period constraint cost function; a weighted sum of the current constraint cost function and the simplified period constraint cost function is performed to obtain a new cost function, and the new cost function is used to determine the switch states of each phase leg of the inverter to be applied at the next moment; when the switching frequency of the inverter is not fixed, it can also accurately control the switch states of each phase leg of the inverter to be applied at the next moment, achieving the purpose of avoiding current spectrum dispersion and improving current quality.

[0010] (2) An optimized method for model predictive control of grid-connected inverter current spectrum dispersion provided by the present invention uses an artificial neural network to adjust the weight factor in the new cost function in real time, realizing efficient and accurate control of the grid-connected inverter under multiple working conditions and being able to adapt to dynamic working condition changes. Description of the Drawings

[0011] To more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0012] Figure 1 It is a schematic flow chart of an optimization method for current spectrum dispersion of model predictive control of a grid-connected inverter provided by an embodiment of the present application; Figure 2 It is an overall schematic diagram of a three-phase grid-connected inverter provided by an embodiment of the present application; Figure 3 It is a schematic diagram of voltage vector projection in two subspaces provided by an embodiment of the present application; Figure 4 It is the same commutation time interval provided by an embodiment of the present application T u and T d schematic diagram; Figure 5 It is a schematic flow chart of the model predictive control steps based on periodic control provided by an embodiment of the present application; Figure 6 It is a schematic diagram of a neuron provided by an embodiment of the present application; Figure 7 It is a schematic diagram of a feedforward artificial neural network provided by an embodiment of the present application; Figure 8 It is an overall block diagram of the model predictive control of a grid-connected inverter provided by an embodiment of the present application; Figure 9 It is an internal structure schematic diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0013] In order to make the objectives, technical solutions and advantages of the present invention clearer, the following further details the present invention in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0014] In the description, claims and the above-mentioned drawings of this application, the terms "first", "second", "third", etc. are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include steps or units not listed, or may optionally further include other steps or units inherent to these processes, methods, products or devices.

[0015] As Figure 1 shown, an optimization method for current spectrum dispersion of grid-connected inverter model predictive control is provided. This method can be executed by a terminal or by a server that communicates with the terminal through a network. Among them, the terminal can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, etc. The server can be an independent server or can be implemented by a server cluster composed of multiple servers. Taking the application of this method to a terminal as an example, the following steps are included: Step 101: Determine the current constraint cost function based on the mathematical model of the three-phase grid-connected inverter and the switching states of each phase bridge arm of the inverter at the current moment.

[0016] Step 102: Establish a period constraint cost function according to the expected average switching period and the time intervals of two consecutive upward and downward commutations of the switching states of each phase bridge arm of the inverter; divide the period constraint cost function by the sampling period to establish a simplified period constraint cost function, and obtain an integer matrix for upward commutation and an integer matrix for downward commutation.

[0017] Step 103: Respectively use the switching signals corresponding to multiple candidate voltage vectors as the predicted values of the switching states of each phase bridge arm of the inverter at the next moment of the current moment.

[0018] Step 104: Calculate the integer matrix for upward commutation and the integer matrix for downward commutation corresponding to multiple candidate voltage vectors according to the predicted value of each switching state, the switching state at the current moment, the integer matrix for upward commutation at the current moment, and the integer matrix for downward commutation at the current moment.

[0019] Step 105: Perform weighted summation on the current constraint cost function and the simplified period constraint cost function to obtain a new cost function of the mathematical model.

[0020] Step 106: Substitute the integer matrix for upward commutation and the integer matrix for downward commutation corresponding to multiple candidate voltage vectors into the new cost function for calculation, and determine the switching states of each phase bridge arm of the inverter to be applied at the next moment.

[0021] In the above grid-connected inverter model predictive control current spectrum dispersion optimization method, a cost function based on period constraint is added to the traditional cost function based on current constraint, that is, a period constraint cost function is established through the time interval between two consecutive identical commutations of the switching states of each phase leg of the inverter (i.e., the time interval between two consecutive upward commutations and the time interval between two consecutive downward commutations), and the period constraint cost function is divided by the sampling period to establish a simplified period constraint cost function; the cost function based on current constraint and the simplified period constraint cost function are weighted and summed to obtain a new cost function, and the new cost function is used to determine the switching states of each phase leg of the inverter applied at the next moment; since the new cost function includes the simplified period constraint cost function, and the simplified period constraint cost function is related to the sampling period, when the switching frequency of the inverter is not fixed, it can also accurately control the switching states of each phase leg of the inverter applied at the next moment, achieving the purpose of avoiding current spectrum dispersion and improving current quality.

[0022] In one embodiment, in step 101 above, based on the mathematical model of the three-phase grid-connected inverter and the switching states of each phase leg of the inverter at the current moment, a current constraint cost function is determined, including calculating the output voltage equation of the inverter on the α-β axis at the current moment based on the mathematical model of the three-phase grid-connected inverter and the switching states of each phase leg of the inverter at the current moment; according to the output voltage equation and the first-order forward Euler formula, the current prediction equation of the inverter on the α-β axis at the next moment after discretization is obtained; according to the current prediction equation, the current prediction equation after one-step delay compensation at the next moment is calculated; based on the α-β axis current reference value and the current prediction equation after one-step delay compensation, the current constraint cost function is determined.

[0023] In one embodiment, in the above step, based on the mathematical model of the three-phase grid-connected inverter and the switching states of each phase leg of the inverter at the current moment, the output voltage equation of the inverter on the α-β axis at the current moment is calculated, including calculating the phase voltage of each phase output by the three-phase grid-connected inverter according to the DC bus voltage and the switching states of each phase leg of the three-phase grid-connected inverter, and taking the phase voltage equations of the three phases as the mathematical model of the three-phase grid-connected inverter; through coordinate transformation, the phase voltages of each phase are mapped to the α-β subspace to obtain the output voltage equation of the inverter on the α-β axis at the current moment.

[0024] In one embodiment, in step 106, the upward commutation integer matrix and the downward commutation integer matrix corresponding to multiple candidate voltage vectors are respectively substituted into the new cost function for calculation to determine the switching states of each phase leg of the inverter applied at the next moment, including calculating the current vectors of the inverter on the α-β axis after one-step delay compensation corresponding to multiple candidate voltage vectors according to the predicted value of each switching state and the current prediction equation after one-step delay compensation; substituting the upward commutation integer matrix, the downward commutation integer matrix, and the current vectors of the inverter on the α-β axis after one-step delay compensation corresponding to multiple candidate voltage vectors into the new cost function for calculation to determine the switching states of each phase leg of the inverter applied at the next moment.

[0025] In one embodiment, the grid-connected inverter model predictive control current spectrum dispersion optimization method includes the following steps: Step 1: Establish a mathematical model of a three-phase grid-connected inverter.

[0026] (1) Figure 2 shows the working principle diagram of a three-phase grid-connected inverter. As Figure 2 shown, the three-phase grid-connected inverter is respectively connected to the DC bus voltage V dc , and converts the direct current provided by the DC bus voltage into alternating current. It can be seen from Figure 2 that u a , u b , u c respectively represent the phase voltages of each phase output by the three-phase grid-connected inverter; i a , i b , i c respectively represent the phase currents of each phase output by the three-phase grid-connected inverter; e a , e b , e c respectively represent the three-phase grid voltages; L , R respectively represent the inductance and resistance on the output side of the three-phase grid-connected inverter.

[0027] First, take the equations of the phase voltages of each phase output by the three-phase grid-connected inverter as the mathematical model of the three-phase grid-connected inverter. The phase voltages of each phase output by the three-phase grid-connected inverter can be expressed in the form of the DC bus voltage and the switching states as follows:

[0028] Wherein, S a , S b , S c respectively represent the switching states of each phase bridge arm. A value of 1 indicates that the upper bridge arm is conducting and the lower bridge arm is off, and a value of 0 indicates that the upper bridge arm is off and the lower bridge arm is conducting. There are a total of 2 3 = 8 switching states; V dc represents the DC bus voltage. Figure 3 Illustrates the 8 voltage vectors corresponding to the 8 switching states: u 0 (000), u 1 (100), u 2 (110), u 3 (010), u 4 (011), u 5 (001), u 6 (101), u 7 (000).

[0029] (2) Using the Clarke transformation (a coordinate transformation method), map the phase voltages of each phase output by the three-phase grid-connected inverter to the α-β subspace to obtain the output voltage equation of the grid-connected inverter on the α-β axis:

[0030] Wherein, u α , u β , i α , i β are the output voltage and current components of the grid-connected inverter on the α-β axis; e α , e β are the grid voltages on the α-β axis; R is the resistance on the output side of the inverter; L is the inductance on the output side of the inverter.

[0031] Step 2. Based on periodic control, implement a model predictive control current spectrum dispersion optimization method.

[0032] (1) According to the mathematical model and transformation matrix of the three-phase grid-connected inverter in Step 1, the switching states of each phase leg of the inverter at the current moment (i.e., k moment) can be used to calculate the output voltage equation of the inverter on the α-β axis at the current moment:

[0033] According to the output voltage equation of the inverter on the α-β axis at the current moment and the first-order forward Euler formula, the predicted current equation of the inverter on the k axis at the next moment (i.e., α-β +1 moment) after discretization can be obtained:

[0034] In the formula, T s represents the sampling period.

[0035] In order to eliminate the influence of delay, according to the predicted current equation of the inverter on the α-β axis at the next moment, the predicted current equation after one-step delay compensation at the next moment (i.e., k +2 moment) is calculated:

[0036] In the formula, and are the predicted currents on the α-β axis at the k +2 moment, and the superscript p represents the predicted value.

[0037] (2) Determine the cost function of model predictive control The traditional cost function based on current constraint (also known as the current constraint cost function) is:

[0038] In the formula, represents the α-β axis current reference value, = T .

[0039] The cost function of the mathematical model of the three-phase grid-connected inverter adopts a periodic control method, adding the same commutation time interval T u and T d ​Take the time interval between two consecutive identical commutations of the switching state of one phase bridge arm of a three-phase grid-connected inverter as an example. T u and T d The meaning of Figure 4 As shown. In this embodiment, by correcting the time variable between two consecutive identical commutations, the purpose of changing the average output can be achieved. The reference value is the expected average switching period, and the purpose of tracking the two time variables is to achieve a modulation-like behavior. The cost function based on the period constraint (also called the period constraint cost function) is:

[0040] In the formula, = [ T ref T ref T ref ] T , = [ T ua T ub T uc ] T , = [ T da T db T dc ] T , T ref is the expected average switching period, T ua , T ub , T uc are the time intervals between two consecutive upward switching of the switch states of each phase bridge arm, T da , T db , T dc They respectively represent the time interval between two consecutive downward switching of the switching state of each phase bridge arm.

[0041] In model predictive control, the time interval between two consecutive identical commutations of the switching state of each phase bridge arm is T u and T d Must be the sampling period Ts an integer multiple of, for example, the sampling period is the time interval between the two dashed lines shown as Figure 4 . Therefore, in order to reduce the computational burden, the periodic constraint cost function is transformed according to the sampling period to obtain:

[0042] where = / T s , = / T s , is an integer matrix for upward commutation, is an integer matrix for downward commutation, and each element in is an integer; = / T s The elements in are not necessarily integers.

[0043] (3) Solve and

[0044] is solved in two steps, and the workflow is as Figure 5 shown.

[0045] The first step is to calculate the integer matrix for upward commutation and the integer matrix for downward commutation at the current moment according to the switching states of each phase leg of the inverter at the previous moment and the current moment, that is, to calculate k at the -1 moment and k at the moment to obtain k at the moment and . If no commutation occurs, and the values of each element in are incremented by 1 and reset to 1 when the corresponding upward or downward commutation occurs. The specific calculation is as follows:

[0046] where S x = S a S b S c T represents the switching state of each phase leg of the inverter, represents​S x Invert.

[0047] The second step is to obtain multiple voltage vectors. The eight voltage vectors shown in Figure 3 ( u 0 , u 1 , u 2 , u 3 , u 4 , u 5 , u 6 , u 7 ) are used as candidate voltage vectors, and the switching signals corresponding to the candidate voltage vectors (000, 100, 110, 010, 011, 001, 101, 000) are respectively used as the predicted values of the switching states of each phase leg of the inverter at the next moment of the current moment. , and the superscript p represents the predicted value.

[0048] According to the predicted value of each switching state, the switching state of each phase leg of the inverter at the current moment S x ( k ), the integer matrix for upward commutation at the current moment and the integer matrix for downward commutation at the current moment , calculate the integer matrix for upward commutation at the next moment corresponding to the eight candidate voltage vectors , the integer matrix for downward commutation , and the superscript p represents the predicted value. The specific calculation is as follows:

[0049] In the formula, i represents the serial number of the candidate voltage vector, i = 1, …, 8.

[0050] According to the predicted value of each switching state and the current prediction equation of the current after one-step delay compensation at the next moment, calculate the current vector α-β of the inverter on the axis after one-step delay compensation corresponding to the eight voltage vectors, i = 1, …, 8.

[0051] Perform weighted summation on the current constraint cost function and the simplified periodic constraint cost function to obtain a new cost function.

[0052] The integer matrix corresponding to the upward commutation of the eight candidate voltage vectors , the integer matrix corresponding to the downward commutation , and the current vector of the inverter on the α-β axis after one-step delay compensation at the next moment are all substituted into the new cost function to find the minimum value among them, and the switching signal of the candidate voltage vector corresponding to the minimum value is used as the k switching state of each phase bridge arm applied to the inverter at S x ( k + 1). The specific calculation is as follows:

[0053] In the formula, is the weight factor, and the adjustment process of the weight factor will be introduced in Step 3.

[0054] Step 3: Real-time adjustment method of the weight factor based on ANN.

[0055] By designing and offline training an Artificial Neural Network (ANN for short), the method in the present invention realizes the automatic and real-time adjustment of the weight factor of the new cost function, enabling the system to maintain a stable switching frequency and good control performance at different operating points. The trained ANN selects the optimal weight factor T ref according to the input desired average switching period and current reference value such that the actual average switching period T ave is as close as possible to the set value T ref . Among them, the actual average switching period T ave = ( T ua + T ub + T uc + T da + T db + T dc ) / 6.

[0056] (1) Design of ANN In the actual system, the actual average switching period T ave and current reference value , weight factor The following relationship exists:

[0057] In the formula, CP Represents probability distribution, which is generated by the coupling relationship between the controller and the controlled object. It is highly nonlinear and has uncertainty. The weight factor is directly solved based on the analytical model. It's not easy.

[0058] Therefore, ANN is used to directly learn the inverse mapping:

[0059] When you need to adjust to a new target switching cycle or current amplitude, just and T ref Input the network, and ANN can output the appropriate .

[0060] Figure 6 is a schematic diagram of a neuron. Figure 6 As shown in the figure, after a neuron receives a set of inputs, these inputs will be weighted and biased respectively, and finally a value is obtained, which is called the net input. At this time, the net input needs to pass through a nonlinear function f(·) to obtain the activity value a of the neuron. The nonlinear function is called the activation function. The activation function uses the rectified linear unit (ReLU).

[0061]

[0062] The advantage of the ReLU activation function is that it has a small amount of computation and only needs to determine whether the input is a positive number. This low computational complexity makes it very suitable for hardware implementation.

[0063] The present invention uses a feedforward artificial neural network (feedforward ANN), which has a relatively simple network structure, strong flexibility and general approximation ability, and is suitable for implementation in a resource-constrained hardware environment. Figure 7 As shown in the figure. Each layer of neurons can receive signals from the previous layer of neurons and generate signals to output to the next layer. The 0th layer is called the input layer, the last layer is called the output layer, and the other intermediate layers are called hidden layers. There is no feedback in the entire network, and the signal propagates unidirectionally from the input layer to the output layer.

[0064] (2) Data sampling and training To construct the training dataset, the controller-motor system combination is iteratively sampled using the partially trained ANN, and data from the controller-motor system combination is obtained for offline training. , ) to getT ave Construct the data set in a way that some areas may not be sampled (especially T ave , when the combination of T ref and is uneven). On the other hand, what really needs to be input into the ANN in the application is T ref and . Therefore, to ensure that the training data is as uniform as possible within the

[0065] working range of and in the operating region ( λ min , λ max , , ), sample and the current reference value uniformly, and use the function to obtain T ave , forming the initial data set D . The sampling points are distributed in a grid pattern, covering the entire range of operations.

[0066] (2.2) Convex hull extraction To ensure that all sample points are within the effective operating region, the convex hull of the initial data set D is then extracted to determine the effective operating range S :

[0067] where X is the controller-motor system combination ( D extracted from the initial data set , ). In the initial data set, the sample points outside the convex hull range are screened out to ensure the effectiveness of sampling.

[0068] (2.3) ANN training After obtaining the initial data set, the initial neural network is trained by minimizing the squared error to obtain a partially trained artificial neural network:

[0069] In the formula, , , are samples extracted from the initial dataset, represents the parameters (including weights and biases) of the initial neural network. is the regularization term used to prevent overfitting.

[0070]

[0071] In the formula, W ( l ) represents the weight matrix of the l th layer, r 1 and r 2 are the coefficients of L 1 and L 2 regularization (i.e., sparse regularization and weight decay), respectively. These regularization methods are used to control the absolute value of the neural network weights to prevent overfitting.

[0072] (2.4) Data Sampling - Iterative Update During the iterative phase, a partially trained artificial neural network is used to input a set of uniformly distributed targets T ref and , and output the estimated weight factor :

[0073] Among them, is the partially trained artificial neural network.

[0074] Let the controller - motor system run respectively under , and use the function to obtain a set of real T ave .

[0075] (2.5) Dataset Update Add the new sample points obtained in the previous step ( , T ave , ) to the initial dataset D to improve the training.

[0076] After each iteration, remove the samples in the new dataset that exceed the previously defined convex hull range S or are not in the desired working interval. The purpose of doing this is to ensure that the sampled data is within the effective operating area and avoid some invalid operating points that may affect the system stability and accuracy.

[0077] After adding the new sample to the dataset, continue to train the ANN to correct the inaccuracies caused by the neural network prediction error.

[0078] (2.6) Repeat multiple iterations Continuously "train → update the ANN → generate new → collect new data → retrain...", so that the dataset gradually covers the entire operating range, and at the same time the accuracy of the ANN is continuously improving. Until the network converges or reaches the upper limit of the number of iterations, finally obtain an ANN that can accurately map ( T ref , ) → (i.e., the trained artificial neural network).

[0079] Through the above iterative sampling and continuous optimization of the training process, the neural network gradually learns the optimal weight factors of the system under different operating conditions. This sampling method effectively improves the coverage of the dataset, enabling the neural network to generalize better. The samples of the initial sampling may be concentrated in certain areas, and by guiding the sampling with the partially trained artificial neural network, new samples that cover the operating area more comprehensively can be continuously obtained. At the same time, remove the data that does not belong to the initial convex hull to ensure that the newly sampled points are valid and within the operating range of the system. This helps to improve the stability and accuracy of the neural network. In addition, use L 1 and L 2 regularization to jointly constrain the values of the weights, prevent the model from overfitting on small datasets, and thus ensure the generalization ability of the model in a wider operating range.

[0080] Step Four: Apply the offline-trained ANN to the control framework, and the optimal weight factor can be determined according to the system target average switching period T ref and the current reference value . The overall system control block diagram is as shown in . Figure 8 shown.

[0081] In summary, the method provided in this embodiment can effectively solve the problem of current spectrum dispersion caused by the unfixed switching frequency of the three-phase grid-connected inverter model prediction control, use the ANN to adjust the weight factor in the new cost function in real time, and improve the current quality. It provides an efficient, accurate and stable solution for the control of the grid-connected inverter under multiple working conditions.

[0082] This application also provides a computer device, and its internal structure diagram can be as shown in Figure 9As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for optimizing the current spectrum dispersion of a grid-connected inverter model prediction control.

[0083] Those skilled in the art can understand that Figure 9 the structure shown in the figure is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0084] As Figure 9 shown in the figure, the present application also provides a computer device, which includes a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps in the above method embodiments.

[0085] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor, it implements the steps in the above method embodiments. Among them, the computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, and magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nano-systems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0086] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0087] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0088] As described above, the above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. Those skilled in the art will readily think of the implementation schemes of the present disclosure after considering the specification and practicing the present disclosure here. This application aims to cover any variations, uses or adaptations of the present disclosure, and these variations, uses or adaptations follow the general principles of the present disclosure and include the common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

[0089] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0090] Those skilled in the art can easily understand that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing the current spectrum dispersion of a grid-connected inverter model predictive control, characterized in that: include: Determining a current constraint cost function based on a mathematical model of a three-phase grid-connected inverter and a switching state of each phase bridge arm of the inverter at a current moment; Establishing a period constraint cost function according to the expected average switching period and the time interval between two consecutive upward and downward commutations of the switching state of each phase bridge arm of the inverter; establishing a simplified period constraint cost function by dividing the period constraint cost function by the sampling period, and obtaining an integer matrix for upward commutation and an integer matrix for downward commutation; Using the switching signals corresponding to the multiple voltage vectors to be selected as the predicted values ​​of the switching states of the bridge arms of each phase of the inverter at the next moment of the current moment; Calculate the integer matrices of upward commutation and the integer matrices of downward commutation corresponding to the plurality of voltage vectors to be selected according to the predicted value of each switch state, the switch state at the current moment, the integer matrix of upward commutation at the current moment, and the integer matrix of downward commutation at the current moment; Performing weighted summation on the current constraint cost function and the simplified period constraint cost function to obtain a new cost function of the mathematical model; Substitute the upward commutation integer matrix and the downward commutation integer matrix corresponding to the multiple selected voltage vectors into the new cost function for calculation to determine the switching state of each phase bridge arm applied to the inverter at the next moment.

2. The method according to claim 1, characterized in that The method of determining the current constraint cost function based on the mathematical model of the three-phase grid-connected inverter and the switch state of each phase bridge arm of the inverter at the current moment includes: Based on the mathematical model of the three-phase grid-connected inverter and the switch state of each phase bridge arm of the inverter at the current moment, the current moment is calculated. α-β The output voltage equation of the inverter on the axis; According to the output voltage equation and the first-order forward Euler formula, the next moment after discretization is obtained α-β a current prediction equation of the inverter on the axis; According to the current prediction equation, calculating the current prediction equation after one step of delay compensation at the next moment; based on α-β The shaft current reference value and the current prediction equation after one-step delay compensation are used to determine the current constraint cost function.

3. The method according to claim 2, characterized in that Substituting the upward commutation integer matrix and the downward commutation integer matrix corresponding to the multiple selected voltage vectors into the new cost function for calculation to determine the switch state of each phase bridge arm applied to the inverter at the next moment, includes: According to the predicted value of each switch state and the current prediction equation after one-step delay compensation, the current prediction equation after one-step delay compensation corresponding to the multiple candidate voltage vectors at the next moment is calculated. α-β A current vector of the inverter on an axis; The integer matrix of upward commutation, the integer matrix of downward commutation, and the integer matrix of the next moment after one-step delay compensation corresponding to the plurality of candidate voltage vectors are α-β The current vectors of the inverter on the axis are respectively substituted into the new cost function for calculation to determine the switching state of each phase bridge arm applied to the inverter at the next moment.

4. The method according to claim 1, characterized in that The step of establishing a cycle constraint cost function according to the expected average switching cycle and the time interval between two consecutive upward and downward switching of the switching state of each phase bridge arm of the inverter comprises: In the formula, = [ T ref T ref T ref ] T , = [ T ua T ub T uc ] T , = [ T da T db T dc ] T , T ref is the expected average switching period, T ua , T ub , T uc are the time intervals between two consecutive upward switching of the switching states of the bridge arms of each phase of the three-phase grid-connected inverter, T da , T db , T dc They respectively represent the time intervals between two consecutive downward commutations of the switching state of each phase bridge arm of the three-phase grid-connected inverter.

5. The method according to claim 1, characterized in that The step of dividing the period constraint cost function by the sampling period to establish a simplified period constraint cost function and obtaining an upward commutation integer matrix and a downward commutation integer matrix includes: Dividing the period constraint cost function by the sampling period, establishing a simplified period constraint cost function, and obtaining an upward commutation integer matrix and a downward commutation integer matrix; According to the switch states of the bridge arms of each phase of the inverter at the previous moment and the current moment, an integer matrix of upward commutation and an integer matrix of downward commutation at the current moment are calculated.

6. The method according to claim 1, characterized in that Substituting the upward commutation integer matrix and the downward commutation integer matrix corresponding to the multiple selected voltage vectors into the new cost function for calculation to determine the switch state of each phase bridge arm applied to the inverter at the next moment, includes: Substitute the upward commutation integer matrix and the downward commutation integer matrix corresponding to the multiple candidate voltage vectors into the new cost function for calculation, determine the minimum value, and use the switching signal of the candidate voltage vector corresponding to the minimum value as the switching state applied to each phase bridge arm of the inverter at the next moment.

7. The method according to claim 1, characterized in that The method also includes adjusting the weight factors in the new cost function in real time through an artificial neural network.

8. The method according to claim 7, characterized in that The method of adjusting the weight factors in the new cost function in real time by using an artificial neural network includes generating an initial data set; training the initial neural network by minimizing the square error based on the initial data set to obtain a partially trained artificial neural network; Updating the initial data set by using the partially trained artificial neural network, and continuing to train the artificial neural network based on the updated data set to obtain a trained artificial neural network; The trained artificial neural network is used to calculate the expected average switching cycle and α-β The shaft current reference value determines the optimal weighting factor.

9. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images