Active front-end rectifier and predictive control method thereof

Abstract

The application discloses an active front-end rectifier and a prediction control method thereof, and the method comprises the following steps: according to a voltage model of a three-level active front-end rectifier, calculating active power and reactive power of a power grid according to instantaneous power theory; according to the voltage model, combining the active power and the reactive power, calculating a power gradient; converting the power gradient into a second-order state space equation, predicting the control input of the active front-end rectifier by using a super-helix observer, and obtaining the input voltage of the active front-end rectifier according to the predicted control input and the grid-side voltage; based on the input voltage, obtaining an optimal voltage vector by using a limited set model prediction control framework, and controlling the active front-end rectifier based on the switching state corresponding to the optimal voltage vector. The method combines power gradient modeling, super-helix observer disturbance compensation and limited set prediction control, so that the rectifier overcomes the influence of parameter mismatch, and realizes high-sinusoidal current, stable DC voltage and power tracking control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rectifier predictive control, in particular to an active front-end rectifier and a predictive control method thereof. BACKGROUND

[0002] With the development of power electronics technology, active front-end rectifiers are increasingly widely used in industrial transmission, new energy power generation and other fields. Finite set model predictive control based on instantaneous power theory has become one of the mainstream schemes due to its superior dynamic performance. However, this scheme faces the following key technical problems in actual application: the control performance is highly dependent on the accurate mathematical model of the system, and model mismatch occurs when the filter inductance and other parameters drift; the linear observation structure such as the extended state observer in related technologies cannot accurately describe the nonlinear dynamic characteristics of the rectifier; and the traditional scheme does not fully consider the voltage saturation constraint of the power switching device. These technical problems lead to phenomena such as increased current harmonics and decreased power tracking accuracy of the system under parameter changes and extreme conditions, which restricts the popularization and application of active front-end rectifiers in complex industrial environments. Therefore, it is urgent to develop a high-performance control method that can effectively overcome parameter sensitivity and saturation nonlinearity problems. SUMMARY

[0003] The embodiments of the present application provide an active front-end rectifier and a predictive control method thereof, which can realize adaptive predictive control of the active front-end rectifier without relying on accurate model parameters.

[0004] In a first aspect, the embodiments of the present application provide a predictive control method of an active front-end rectifier, which comprises:

[0005] According to the voltage model of the three-level active front-end rectifier, the active power and the reactive power of the power grid are calculated according to the instantaneous power theory;

[0006] According to the voltage model, the power gradient is calculated in combination with the active power and the reactive power;

[0007] The power gradient is converted into a second-order state space equation, the control input of the active front-end rectifier is predicted by using a super-helix observer, and the input voltage of the active front-end rectifier is obtained according to the predicted control input in combination with the grid-side voltage;

[0008] Based on the input voltage, the optimal voltage vector is obtained by using a finite set model predictive control framework, and the active front-end rectifier is controlled based on the switching state corresponding to the optimal voltage vector.

[0009] In one of the embodiments, the predictive control method further comprises using a fast gradient method to adaptively update the input gain of the super-helix observer, comprising:

[0010] When there is a mismatch in the input gain of the super-twisting observer, the second-order state space equation is expressed as:

[0011] ;

[0012] wherein, , , , , , P g represents active power, Q g represents reactive power, represents a filter resistance of the system, represents a filter inductance, , represents an axis component of a grid-side voltage , represents an axis component of a grid-side voltage , represents a grid-side angular frequency, represents an estimated value of the input gain ;

[0013] calculates a normal unit vector of a state variable X 1, performs an inner product of a transposed vector of the normal unit vector and an extended state X 2, and constructs an error equation of a fast gradient algorithm based on a result of the inner product;

[0014] estimates the input gain of the super-twisting observer based on the error equation, and updates the input gain of the super-twisting observer based on the estimated input gain.

[0015] In one embodiment, constructing the error equation of the fast gradient algorithm based on the result of the inner product includes:

[0016] part of known quantities in the result of the inner product is removed by the following equation:

[0017] ;

[0018] wherein, represents a normal unit vector of a state variable X 1, represents a grid-side angular frequency, configuring the error equation to calculate a square of a difference between zero and the inner product after removing the part of the known quantities.

[0019] In one embodiment, estimating the input gain of the super-twisting observer based on the error equation includes:

[0020] According to the error equation, a gradient of the input gain estimation value is obtained Based on the gradient of the input gain estimation value The input gain estimation value is adaptively updated using a fast gradient method in a discrete form .

[0021] In one of the embodiments, the actuator is used to clip the control input of the active front-end rectifier output by the super-spiral observer, and the input voltage of the active front-end rectifier is obtained based on the clipped control input combined with the grid-side voltage.

[0022] Among them, the execution deviation of the actuator is fitted using a neural network to obtain an execution deviation estimation value, and the control input of the active front-end rectifier output by the super-spiral observer is subtracted from the execution deviation estimation value to obtain the clipped control input.

[0023] In one of the embodiments, the neural network is a two-layer neural network, and the neural network is represented by the following formula:

[0024] ;

[0025] In the formula, V T represents the first layer weight of the neural network which is fixed and known, represents the second layer weight of the neural network, represents the known activation function, X NN =[ R*X 1] T , , P g * represents the reference value of active power, Q g * represents the reference value of reactive power.

[0026] In one of the embodiments, the predictive control method further comprises:

[0027] The update rate of the second layer weight of the neural network is set, and the second layer weight of the neural network is updated in a first-order Euler discrete manner based on the update rate, and the execution deviation estimation value is calculated based on the updated neural network.

[0028] In one of the embodiments, the update rate of the second layer weight of the neural network is configured as:

[0029] ;

[0030] In the formula, is a constant matrix, is a normal number,E This represents the tracking error of the system. b This indicates the input gain.

[0031] In one embodiment, predicting the control input of the active front-end rectifier using a superhelical observer includes:

[0032] A superspiral observer is used to establish the nonlinear sliding membrane function for the state variables in the second-order state-space equation. X 1 and expansion state X 2. Make an estimate to obtain the first observation state and the second observation state;

[0033] Discretize the first and second observation states, and tune the superspiral observer so that the second observation state approximates the extended state. X 2,

[0034] System status The control input of the superspiral observer is obtained by subtracting the estimate of the extended state from the derivative of the reference value, then subtracting the system tracking error modulated by the error coefficient, adding the robust control term of the system, and finally dividing by the input gain.

[0035] Secondly, this application also provides an active front-end rectifier configured to apply a predictive control method for an active front-end rectifier as described in the first aspect.

[0036] The aforementioned active front-end rectifier and its predictive control method calculate the active and reactive power of the power grid based on the rectifier voltage model and instantaneous power theory; calculate the power gradient according to the voltage model and power values; after converting the power gradient into a second-order state-space equation, use a superspiral observer to predict the control input and obtain the rectifier input voltage by combining it with the grid-side voltage; obtain the optimal voltage vector based on the input voltage using finite set model predictive control, and control the rectifier through the corresponding switching states. This method, through the organic combination of power gradient modeling, superspiral observer disturbance compensation, and finite set predictive control, enables the rectifier to overcome the influence of parameter mismatch and achieve high sinusoidal current, fast and stable DC voltage, and accurate power point tracking control. Attached Figure Description

[0037] Figure 1 This is a flowchart of a predictive control method for an active front-end rectifier in one embodiment;

[0038] Figure 2 This is a schematic diagram of a three-level active front-end rectifier circuit topology in one embodiment;

[0039] Figure 3 This is a flowchart illustrating the prediction of control inputs to an active front-end rectifier in one embodiment;

[0040] Figure 4 This is a waveform diagram of the three-phase current on the AC side in one embodiment;

[0041] Figure 5 The waveform of the DC-side output voltage in one embodiment is shown.

[0042] Figure 6 The waveform diagram of grid-side voltage power in one embodiment is shown.

[0043] Figure 7 This is a current harmonic spectrum in one embodiment. Detailed Implementation

[0044] The present application will be described in detail below with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present application. Any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the protection scope of the present application.

[0045] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0046] In one embodiment, such as Figure 1 As shown, a predictive control method for an active front-end rectifier is provided, which includes the following steps:

[0047] Step 101: Based on the voltage model of the three-level active front-end rectifier, calculate the active power and reactive power of the power grid according to the instantaneous power theory;

[0048] Specifically, the three-level active front-end rectifier circuit topology is as follows: Figure 2 As shown, Arm-1, Arm-2, and Arm-3 represent the three arms of the three-phase system; each arm contains four power switches, where S... a1 S a2 S a3 S a4 The four switching transistors representing phase A bridge arm, S b1 Sb2 , S b3 , S b4 represents B-phase bridge arm, S c1 , S c2 , S c3 , S c4 represents C-phase bridge arm; DC side contains two voltage-sharing capacitors C1 and C2 and load resistor R L , i dc is DC side output current; i cp may refer to the current flowing into voltage-sharing capacitor C1; i cl may refer to the current flowing into voltage-sharing capacitor C2; R g and L g are filter resistance and filter inductance, respectively, i a is AC current flowing into A-phase bridge arm, i b is AC current flowing into B-phase bridge arm, i c is AC current flowing into C-phase bridge arm. u ra , u rb , u rc represent input phase voltage of rectifier at three-phase AC side, respectively; v p may represent positive DC bus voltage (i.e. voltage across capacitor C 1); v l may represent voltage across capacitor C 2. The topology can effectively manage circulating current and ensure power quality at AC side while achieving DC bus voltage stability control by precisely controlling on-off states of 12 switching tubes.

[0049] The voltage model of three-level active front-end rectifier in stationary αβ coordinate system is as follows:

[0050] ;

[0051] wherein, e gα、 e gβ represent components of grid-side voltage in axis, respectively, u rα、 u rβ represent components of rectifier input voltage in axis, respectively, igα、 i gβ represents the component of the current on the line in the direction of the axis, R g and L g are the filter resistance and filter inductance, respectively. The voltage model can represent the grid-side voltage e gα、 e gβ the dynamic relationship between the line current i gα、 i gβ and the rectifier input voltage u rα、 u rβ To directly achieve control of power, a power calculation model based on instantaneous power theory is introduced, the core idea of which is to use the instantaneous values of voltage and current in the αβ coordinate system to perform algebraic operations, thereby avoiding complex coordinate rotation and filter links and achieving fast reactive power control.

[0052] The active power P g and the reactive power Q g of the grid side are calculated as follows:

[0053]

[0054] It should be noted that the active power P g can represent the true rate of transmission of electrical energy from the grid to the DC side; the reactive power can reflect the energy that is constantly exchanged but not consumed between the grid and the rectifier. This instantaneous value-based power calculation method has fast response speed, does not require filtering, provides accurate real-time feedback signals for subsequent predictive control algorithms, and is naturally suitable for implementation by a digital controller.

[0055] Step 102: Calculate the power gradient according to the voltage model in combination with the active power and the reactive power;

[0056] Specifically, the active power P g and the reactive power are differentiated Q g to obtain the calculation formula of the power gradient:

[0057]

[0058] where the power gradient of the active power is affected by the rectifier input voltage​​​ u rα and input voltage Components of the axis u rβ Filter resistor R g Grid-side voltage at Square of the components of the axis and grid angular frequency ω With reactive power The coupling effect has a combined influence; reactive power power gradient Then subject to the rectifier input voltage u rα and input voltage Components of the axis u rβ Filter resistor R g and grid angular frequency ω With active power The coupling effect determines this.

[0059] By analyzing the rectifier's voltage model and current power state, the changes in active and reactive power in the next second under a given control command are calculated. This calculation process comprehensively considers the rectifier's voltage regulation effect, the loss effect of circuit resistance, the fundamental driving force of the grid voltage, and the mutual influence between active and reactive power.

[0060] Step 103: Convert the power gradient into a second-order state-space equation, use a super-spiral observer to predict the control input of the active front-end rectifier, and obtain the input voltage of the active front-end rectifier based on the predicted control input and the grid-side voltage.

[0061] Specifically, the power gradient is expressed as a first-order state-space equation, with the following formula:

[0062] ;

[0063] in, a =- R g / L g , b =1.5 / L g , , , .

[0064] Further, in order to solve the problems of parameter uncertainty, unmodeled dynamics and external disturbance existing in practical applications, by introducing an extended state, the uncertainty factors difficult to accurately model are unified into the extended state X2. The state reconstruction method changes the focus of the control problem from pursuing accurate modeling to realizing accurate observation, and creates necessary conditions for subsequent feedforward compensation control based on disturbance estimation by defining the disturbance as a new state variable. Therefore, it is necessary to extend the first-order state space equation to a second-order state space equation, and the second-order state space equation is:

[0065] ;

[0066] Let V represent the rate of change of the lumped disturbance and the extended state X2, and the further second-order state space equation is:

[0067] ;

[0068] ;

[0069] .

[0070] The second-order state space equation is generally observed by a linear observer such as an extended state observer and a general proportional input-output observer. However, power converters are mathematically classified as nonlinear systems, and using a linear observer will cause problems such as inaccurate modeling and lack of stability. The super-twisting observer solves this problem by converting the discontinuous sliding mode variable causing chattering into a high-order derivative. Therefore, the super-twisting observer is used to establish a nonlinear sliding mode function to overcome the problem of inaccurate estimation of nonlinear systems by traditional linear observers, and to achieve accurate estimation of states and disturbances through nonlinear sliding mode characteristics.

[0071] Among them, the super-twisting observer is a high-order sliding mode observer, which can quickly, smoothly and without overshoot estimate the system state through a unique nonlinear structure. The super-twisting observer can estimate the state variables X 1 and the extended state X 2 in the second-order state space equation to obtain the first observed state and the second observed state. Further, the first observed state and the second observed state are discretized, which aims to facilitate digital implementation and form an algorithm that can be iteratively calculated within each sampling period, which can ensure good estimation performance under various operating conditions. The derivative of the reference value of the system state is subtracted from the estimate of the extended state, and then subtracted from the system tracking error modulated by the error coefficient, and finally divided by the input gain to obtain the control input of the super-twisting observer output.

[0072] On this basis, according to the feedback control principle, in order to make the tracking error converge to 0, the control input is designed as follows:

[0073] ;

[0074] wherein, represents the state X 1 reference value, may represent the tracking error, is the robust control term of k t Generally take 1 / T s , K r is a normal number.

[0075] It should be noted that, may be an ideal instruction calculated in the control algorithm, which exists in the digital world of the controller.

[0076] Because .

[0077] This definition can establish the mapping relationship between the mathematical control quantity U and the active front-end rectifier input voltage and in the physical world. Further, by inverse transformation to solve the actual voltage, since the control algorithm gives the ideal , and U has a clear mathematical relationship with the active front-end rectifier input voltage and , through matrix inversion, the input voltage of the active front-end rectifier can be obtained from the ideal control quantity:

[0078] .

[0079] Step 104: Based on the input voltage, the optimal voltage vector is obtained by using the finite set model predictive control framework, and the active front-end rectifier is controlled based on the switching state corresponding to the optimal voltage vector.

[0080] Specifically, the finite set model predictive control framework is used to realize the direct decision and execution of the optimal switching vector. The core of this control strategy is to fully utilize the inherent characteristics of the limited switching state of the three-level rectifier, and to form a determined finite search set with 27 possible switching combinations; in each control period, all candidate switching vectors are traversed, and the key controlled quantities, such as active power and reactive power, that each vector will produce at the next sampling time are predicted based on the discretized mathematical model.

[0081] Further, the predicted key controlled variables are sent into a pre-designed value function for evaluation, which can assign a score representing the control effect to each candidate vector by quantifying the tracking error (e.g. the deviation of the predicted power from the reference value) and taking into account the switching frequency constraints and other limitations; by comparing all scores, the switching state that minimizes the value function is selected as the optimal voltage vector for the current period. Further, the optimal switching group is applied to the power switch tube of the rectifier, and the state is maintained throughout the sampling period until the next period re-optimization.

[0082] In the embodiment, the method calculates the active power and the reactive power of the power grid based on the rectifier voltage model and the instantaneous power theory; calculates the power gradient according to the voltage model and the power value; after converting the power gradient into a second-order state space equation, the input of the super-helix observer is predicted and controlled, and the input voltage of the rectifier is obtained in combination with the grid-side voltage; based on the input voltage, the optimal voltage vector is obtained by using the finite set model predictive control, and the rectifier is controlled by the corresponding switching state. Through the organic combination of power gradient modeling, super-helix observer disturbance compensation and finite set predictive control, the rectifier overcomes the influence of parameter mismatch, realizes high-sinusoidal current, fast and stable DC voltage and accurate power tracking control.

[0083] In one embodiment, as shown in FIG. 1, the control input of the active front-end rectifier is predicted by using a super-helix observer, which includes the following steps: Figure 3

[0084] Step 301: A super-helix observer is used to establish a nonlinear sliding mode function to estimate the state variables X 1 and the extended state X 2 in the second-order state space equation, to obtain the first observed state and the second observed state;

[0085] Specifically, by establishing a nonlinear sliding mode function, the observer can estimate the state variables X 1 and the extended state X 2 in the second-order state space equation, to obtain the first observed state Z 1 and the second observed state Z 2. Its continuous-time dynamics are described by the following formula:

[0086] ;

[0087] The method introduces the fractional power term of the state error and the sign function to form a nonlinear sliding surface, so that the observer dynamics can quickly converge.

[0088] Step 302: Discretize the first observed state and the second observed state, and tune the super-helix observer so that the second observed state approximates the extended state X ​2;

[0089] Specifically, the continuous-time observer dynamics are discretized to obtain the discrete form shown in the following formula:

[0090] ;

[0091] This discrete algorithm enables the observer to perform each sampling period. T s Internally, based on current measurements X ( k ) and control input U ( k Iteratively update its state. X 1 and expansion state X 2 Observational state values Z 1 and Z 2.

[0092] By tuning the observer gain k 1 and k 2 (Usually based on the system's dynamic range and sampling period, the optimal range is determined through stability analysis or simulation), ensuring that the discretized superspiral observer not only remains stable but also rapidly forces the state estimation error to change within a finite time. X ( k )− Z 1( k Its derivative converges to the zero neighborhood. Once the superhelical observer is well tuned, its internal dynamics will drive the second observation state. Z 2. Approximates the extended state with high precision (within the accuracy range of the discrete system). X 2.

[0093] Step 303: Set system status The control input of the superspiral observer is obtained by subtracting the estimate of the extended state from the derivative of the reference value, then subtracting the system tracking error modulated by the error coefficient, adding the robust control term of the system, and finally dividing by the input gain.

[0094] Specifically, the system state X 1 Reference value R ∗ ( k The derivative of +1) minus the expansion state provided in real time by the superspiral observer X Estimator of 2 Z 2( k +1), its purpose is to actively offset all identified uncertainties and disturbances in the system through a feedforward approach.

[0095] Further subtract the instantaneous tracking error of the system after the error coefficient modulation. , the feedback term ensures that the system output can accurately track the reference trajectory; plus the robust control term of the system , which is a nonlinear correction term based on the sliding mode principle, used to suppress unmodeled dynamics and residual disturbances; the algebraic sum of all the above terms divided by the estimated value of the input gain of the system (online updated by fast gradient method), complete the normalization processing of the control quantity, so as to obtain the control input of the final output of the supercoil observer , which integrates disturbance compensation, feedback error adjustment and nonlinear robust correction, constitutes the core instruction to realize high-performance control of the active front-end rectifier.

[0096] In one embodiment, the predictive control method further comprises using a fast gradient method to adaptively update the input gain of the supercoil observer, comprising:

[0097] When the input gain of the supercoil observer does not match, the second-order state space equation is expressed as:

[0098] ;

[0099] In the formula, , , , , , P g represents the active power, Q g represents the reactive power, represents the filter resistance of the system, represents the filter inductance, , represents the grid-side voltage axis component, represents the grid-side voltage axis component, represents the grid-side angular frequency, represents the estimated value of the input gain ;

[0100] Calculate the normal unit vector of the state variable X 1, take the inner product of the transpose vector of the normal unit vector and the extended state X 2, and construct the error equation of the fast gradient algorithm based on the inner product result;

[0101] Based on the error equation, estimate the input gain of the supercoil observer, and update the input gain of the supercoil observer based on the estimated input gain.

[0102] Specifically, when the input gain estimation value preset inside the controller is not equal to the actual valueb When mismatch exists, the second-order state-space equations characterizing the system power dynamics and lumped disturbances are reformulated as follows: This formula can represent the parameter mismatch term ( b - ) U This is a significant source of observation and control errors. To achieve... The algorithm calculates the state variables for self-correction. X 1 (i.e., power vector) The normal unit vector of ) The normal unit vector is perpendicular to the current power operating point.

[0103] Furthermore, the normal unit vector S Transpose and expansion states X 2. Perform inner product operations S T X 2, can be obtained ,in, , can represent the control input vector U In the state normal unit vector S Projection components in the direction. The error equation for the fast gradient algorithm is constructed based on the inner product result: F s = .

[0104] Through the error equation The core purpose of processing the original inner product result is to... Mixed signals (including known nonlinear couplings) and some known input gains Calculated partial effects Subtracting known or estimated quantities from the input gain mismatch yields a pure result. b - The error components contributed by ) .

[0105] Based on this result, the error equation is configured to calculate the inner product of zero and the result after removing some known quantities. The square of the difference, i.e. F ( b )=(0− F s ) 2 =[( - b ()( )] 2 The error function F ( b When it reaches its minimum value (zero), it must satisfy the following condition. = 0, under the condition of continuous excitation (i.e. = 0), that is, means = 0. b The minimum point of this function corresponds to the correct b value.

[0106] Based on the error equation , the input gain of the super-spiral observer is estimated and updated online. According to the error equation, the gradient ( F ) of the input gain estimate value is obtained by derivation operation, which is obtained by deriving the error function F ( ) with respect to the estimate value , and can be expressed as: . Wherein, is the effective excitation signal, wherein is the projection component of the control input vector U in the direction of the state method normal unit vector S . The gradient F ( ) can be expressed as the direction and amplitude that should be adjusted to reduce the estimation error.

[0107] Since the true parameter b is unknown, in actual calculation, the F s output of the observer is used as a feedback signal to replace it, so that the gradient in practice is . Further, using the fast gradient method, the input gain estimate value can be estimated as:

[0108] ;

[0109] Wherein, θ is the adaptive step size, λ and ρ are adjustable coefficients.

[0110] Further, by calculating the gradient F ( ) of the error function and using the discrete form of the fast gradient method:

[0111] ;

[0112] Wherein, is the adaptive step factor, which itself will be dynamically adjusted according to the gradient norm (by the coefficient​ Driven increase, by (Controlled decay), continuously fine-tuning within the normalized gradient direction. This allows it to converge quickly to the true input gain. b This ensures that the superspiral observer is even when the system filter inductor is in operation. L g Even when the situation is unknown or changes, it can still maintain accurate disturbance estimation capabilities.

[0113] In one embodiment, the predictive control method further includes:

[0114] An actuator is used to trim the control input of the active front-end rectifier output from the superspiral observer. Based on the trimmed control input and the grid-side voltage, the input voltage of the active front-end rectifier is obtained.

[0115] Specifically, a neural network is used to fit the execution deviation of the actuator to obtain an estimated value of the execution deviation. The estimated value of the execution deviation is then subtracted from the control input of the active front-end rectifier output by the superspiral observer to obtain the trimmed control input.

[0116] Specifically, in a practical control system, the ideal control input output of the superspiral observer is... When actually executed by an actuator (such as a power switching device), a saturation effect may occur due to physical limitations, that is, when the desired input exceeds the upper limit of the actuator's output. U max or lower limit U min At that time, the actuator will trim it, resulting in the actual output. U ( k ) and expected output Γ( k ) produces an execution deviation Ψ( k ), execution deviation Ψ( k )as follows:

[0117] ;

[0118] in, m This is the scaling factor between the expected input and the actual input when the system is not saturated; it is usually 1.

[0119] In one embodiment, to actively compensate for this nonlinear saturation effect, this application introduces a neural network to fit the execution deviation online. The neural network is a two-layer neural network, and the neural network is represented by the following formula:

[0120] ;

[0121] In the formula, V Tdenotes the fixed and known first layer weights of the neural network, denotes the second layer weights of the neural network, denotes the known activation function, , denotes the reference value of active power, denotes the reference value of reactive power;

[0122] The update rate of the second layer weights of the neural network is set, based on the update rate, the second layer weights of the neural network are updated in a first-order Euler discrete manner, and the execution deviation estimation value is calculated based on the updated neural network;

[0123] The update rate of the second layer weights of the neural network is configured as:

[0124] ;

[0125] In the formula, is a constant matrix, is a normal number, E denotes the tracking error of the system, b denotes the input gain.

[0126] Specifically, the neural network used in practice is an ideal neural network, which is represented by the formula , wherein V T denotes the fixed weight matrix of the first layer of the neural network, which remains unchanged after network initialization; denotes the adjustable weight matrix of the second layer of the neural network, which is a key parameter for online learning of the network; σ (⋅) denotes a selected activation function, in this application, a Sigmoid function is specifically used, which has a smooth saturation characteristic and can effectively simulate the saturation behavior of an actuator. The network input vector is composed of the system power reference value and the actual power state vector .

[0127] To ensure that the neural network can dynamically track the time-varying execution deviation characteristics, a specific adaptive update rate is designed for the second layer weights , which has a clear physical meaning, that is, the first term on the right side of the equation constitutes a gradient descent learning term based on Lyapunov stability, wherein is a constant matrix, and is the learning rate of the neural network, which can control the speed of weight update, and the term drives the neural network weight to adjust in the direction of reducing the system tracking error E ; the second term may be a specially designed damping term, which can be a normal numberk nn The damping strength is adjusted to prevent the weight matrix from having excessive numerical fluctuations during the adaptive process, and to ensure numerical stability and convergence of the learning process.

[0128] In an actual digital control system, the continuous-time update rate can be converted into a discrete form that can be directly programmed and implemented by using a first-order Euler discretization method:

[0129] ;

[0130] So that the weight parameters can be updated based on the latest state in each sampling period Ts The execution deviation estimation value ( k +1) at the current time can be obtained in real time through the forward calculation of the neural network based on the updated weight This estimation value will be directly used in the feedforward compensation link of the control input to effectively suppress the nonlinear effects of the actuator saturation.

[0131] In one embodiment, the active front-end rectifier system parameters are as follows: the AC grid output frequency is set to 50 Hz, the DC bus voltage reference value is set to 500 V, the control system sampling time is set to 100 μs, the DC side capacitance value is set to 2700 μF, the grid-side filter inductance is set to 10 mH, the series equivalent resistance of the grid-side filter inductance is set to 1 Ω, and the load resistance value is set to 40 Ω.

[0132] Figure 4 The horizontal axis is the time axis, and the vertical axis is the instantaneous value of the three-phase current, which shows the quality of the AC side current waveform, and the displayed AC side three-phase current waveform presents high sinusoidal and symmetry; Figure 5 The horizontal axis is the time axis, and the vertical axis is the DC voltage value, which reflects the dynamic response of the DC side voltage, and the presented dynamic process of the DC side output voltage shows that it can quickly converge to the 500 V reference value after a short startup stage, with small overshoot and short regulation time; Figure 6 The horizontal axis is the time axis, and the vertical axis is the power value, which presents the active / reactive power tracking process, and the displayed grid side active / reactive power response curves show that both can smoothly and accurately track the given reference value, with very small steady-state fluctuations, realizing accurate power decoupling control; Figure 7 The horizontal axis is the harmonic order, and the vertical axis is the harmonic content rate, which shows the current harmonic distribution characteristics, and further shows that the amplitudes of each harmonic component are suppressed at a low level, effectively improving the grid current quality.

[0133] The simulation results collectively show that the method described in the application has superior comprehensive performance in dynamic response, steady-state accuracy, and power quality.

[0134] Based on the same concept, the application also provides an active front-end rectifier configured to apply the predictive control method described above.

[0135] Specifically, the active front-end rectifier protected by the application calculates instantaneous active power and reactive power in the stationary αβ coordinate system based on the collected grid voltage and current signals; uses a super-spiral observer, a nonlinear estimator, to capture power dynamics and observe all model uncertainties and external disturbances as a collective extended state in real time; adaptively updates the key parameters (input gain) of the observer online through the fast gradient method to completely eliminate the prior dependence on the physical parameters of the system (such as the filter inductance); integrates a neural network compensator to fit and compensate for nonlinear execution deviations caused by voltage saturation limits of power switching tubes. The controller converts the processed control instructions into optimal switching vectors and directly drives the power switching devices (such as IGBTs) of the rectifier under the framework of limited set model predictive control.

[0136] Through the collaborative design of software and hardware, the active front-end rectifier entity can continuously exhibit excellent steady-state accuracy, fast dynamic response, and strong anti-interference ability in actual industrial applications, especially in complex working conditions with parameter drift and model uncertainty, thereby becoming a high-performance and high-reliability power conversion device.

[0137] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) manner. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (for example, floppy disk, hard disk, magnetic tape), optical media (for example, DVD), or solid state disk (SSD) and the like.

[0138] Each of the embodiments in the specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment. The above is only the preferred embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of predictive control of an active front-end rectifier, characterized by, The prediction control method comprises: According to the voltage model of the three-level active front-end rectifier, the active power and the reactive power of the power grid are calculated according to the instantaneous power theory; According to the voltage model, the power gradient is calculated in combination with the active power and the reactive power; An extended state is introduced, which is a set of uncertain factors. The power gradient is converted into a second-order state space equation through the extended state, the control input of the active front-end rectifier is predicted by using a super-helix observer, and the input voltage of the active front-end rectifier is obtained according to the predicted control input in combination with the grid-side voltage. Based on the input voltage, an optimal voltage vector is obtained by using a finite set model prediction control framework, and the active front-end rectifier is controlled based on the switching state corresponding to the optimal voltage vector.

2. The predictive control method of an active front-end rectifier according to claim 1, characterized in that, The prediction control method further comprises using a fast gradient method to adaptively update the input gain of the super-helix observer, comprising: When there is no match between the input gain of the super-helix observer, the second-order state space equation is represented as: ； wherein , , , , , P g denotes the active power, Q g denotes the reactive power, denotes the filter resistance of the system, denotes the filter inductance, , denotes the α-axis component of the rectifier input voltage, denotes the β-axis component of the rectifier input voltage, denotes the grid-side voltage α-axis component, denotes the grid-side voltage β-axis component, denotes the grid-side angular frequency, denotes an estimate of the input gain . Calculate state variables X The normal unit vector of 1, the transpose of the normal unit vector and the expansion state X 2. Perform the inner product and construct the error equation for the fast gradient algorithm based on the inner product result; Based on the error equation, the input gain of the super-helix observer is estimated, and the input gain of the super-helix observer is updated based on the estimated input gain.

3. The prediction control method of the active front-end rectifier according to claim 2, wherein The error equation of the fast gradient algorithm is constructed based on the inner product result, comprising: Part of the known quantities in the inner product result are removed by the following formula: wherein denotes the state variable X 1, a normal unit vector, denotes the grid-side angular frequency, , The error equation is configured to calculate the square of the difference between zero and the inner product result after removing part of the known quantities.

4. The prediction control method of the active front-end rectifier according to claim 3, wherein Based on the error equation, the input gain of the super-helix observer is estimated, comprising the following steps: According to the error equation, a gradient of the input gain estimate value is obtained, and the input gain estimate value is adaptively updated using a fast gradient method in a discrete form based on the gradient of the input gain estimate value .

5. The method of predictive control of an active front-end rectifier according to claim 1, characterized in that, The prediction control method further comprises: The control input of the active front-end rectifier output by the super-helix observer is clipped by using an actuator, and the input voltage of the active front-end rectifier is obtained based on the clipped control input in combination with the grid-side voltage; Wherein, the execution deviation of the actuator is fitted by using a neural network to obtain an execution deviation estimation value, and the control input of the active front-end rectifier output by the super-helix observer is subtracted by the execution deviation estimation value to obtain the clipped control input.

6. The predictive control method of an active front-end rectifier according to claim 5, characterized in that, The neural network is a two-layer neural network, and the neural network is represented by the following formula: ； wherein V T denotes the first layer weights of the neural network which are fixed and known, denotes the second layer weights of the neural network, denotes a known activation function, , , P g denotes a reference value for the active power, Q g denotes a reference value for the reactive power.

7. The method of predictive control of an active front-end rectifier according to claim 6, characterized in that, The prediction control method further comprises: The update rate of the second layer weight of the neural network is set, the second layer weight of the neural network is updated by using a first-order Euler discrete method based on the update rate, and the execution deviation estimation value is calculated based on the updated neural network.

8. The prediction control method of the active front-end rectifier according to claim 7, wherein The update rate of the second layer weight of the neural network is configured as: ； wherein is a constant matrix, is a normal number, E denotes the tracking error of the system, b denotes the input gain.

9. The prediction control method of the active front-end rectifier according to claim 1, wherein The control input of the active front-end rectifier is predicted by using a super-helix observer, comprising: A nonlinear sliding function is established for the state variables in the second-order state space equation using the supercoiling observer X 1 and the extended state X 2 are estimated to obtain a first observed state and a second observed state discretizing the first and second observation states, tuning the super-twisting observer to cause the second observation state to approximate the extended state X 2, The derivative of the reference value of the state variable is subtracted from the estimate of the extended state, then subtracted from the system tracking error modulated by the error coefficient, plus the robust control item of the system, and finally divided by the input gain to obtain the control input of the super-spiral observer output.

10. An active front-end rectifier characterized by The active front-end rectifier is configured to apply the predictive control method as claimed in any of claims 1 to 9.

Images