Wind power converter current optimization method and system based on model predictive control

Abstract

The embodiment of the application relates to the technical field of wind power converter optimization, and provides a wind power converter current optimization method and system based on model predictive control, which firstly constructs a system state vector capable of comprehensively reflecting real-time state of a power grid, and innovatively introduces fuzzy reasoning, takes normalized voltage amplitude, voltage unbalance degree and other key state quantities as inputs, and generates adaptive weight factors in real time and dynamically through the fuzzy reasoning, so as to online adjust a cost function of model predictive control. In this way, multiple control targets such as current tracking accuracy and switching loss can be online weighted and optimized under different power grid conditions, and finally, accurate and robust control of the converter current under a complex power grid environment is realized, so that the adaptability of the wind power converter to the power grid and the reliability of grid-connected operation are significantly enhanced.

Description

Technical Field

[0001] This invention relates to the field of wind power converter optimization technology, and in particular to a method and system for optimizing wind power converter current based on model predictive control. Background Technology

[0002] With the transformation of the global energy structure, wind power, as an important renewable energy source, faces higher demands on the stability, reliability, and power quality of the power system due to its large-scale grid connection. As the key interface between wind turbines and the grid, the performance of the wind turbine converter directly affects the grid connection capability and grid adaptability of wind farms. Especially in complex grid environments, such as voltage dips, frequency fluctuations, and harmonic injection, accurately and quickly controlling the converter current to ensure the continuous and stable operation of wind turbines and meet stringent grid guidelines has become a core challenge for wind power grid connection technology. Therefore, developing efficient and robust wind turbine converter current optimization methods is of great significance for improving the overall performance and grid friendliness of wind power systems.

[0003] Traditional wind power converter current control schemes, especially those based on Model Predictive Control (MPC), while exhibiting advantages in fast response and multi-objective control, often suffer from performance limitations imposed by the weighting factors in the cost function. Existing MPC methods typically tune and fix the cost function weights offline. This static design is essentially a single-point optimization, its effectiveness highly dependent on the typical operating conditions used in the design. However, the actual operating environment of wind power converters is highly dynamic and variable, creating a fundamental contradiction between the dynamic changes in control objective priorities and the fixed weighting factors. Specifically, MPC methods based on fixed weighting factors cannot achieve the optimal trade-off between control objectives under dynamically changing conditions such as grid voltage amplitude, voltage imbalance, and frequency deviation. This leads to suboptimal control performance when dealing with complex grid faults, and may even fail to effectively meet the dynamic response requirements of grid guidelines, thus affecting the stable operation of wind farms and grid connection reliability. Summary of the Invention

[0004] The present invention aims to solve at least one of the problems existing in the prior art, and provides a method and system for optimizing the current of wind power converters based on model predictive control.

[0005] One aspect of the present invention provides a method for optimizing the current of a wind power converter based on model predictive control, the method comprising: Based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current in the power grid connected to the wind power converter, a system state vector is constructed. The system state vector includes the normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current. Based on the system state vector, the external d-axis current reference and the external q-axis current reference are dynamically estimated by grid guidelines to obtain the final dq-axis current reference. Fuzzy reasoning is performed based on the system state vector to obtain the adaptive weight factor vector; Based on the current dq-axis current, the current dq-axis voltage, and the current switching state, predict the future current state to obtain a set of all predicted currents; Based on the final dq-axis current reference, the adaptive weight factor vector, and the current switching state, an adaptive cost function evaluation and optimal vector selection are performed on the set of all predicted currents to obtain the optimal switching state.

[0006] Optionally, based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current in the power grid connected to the wind power converter, a system state vector is constructed, including: The instantaneous values ​​of the three-phase voltages are used to estimate the grid synchronization angle and frequency to obtain the estimated synchronization angle, grid frequency, and static voltage. Voltage components in a coordinate system; Based on the estimated synchronization angle, the instantaneous values ​​of the three-phase voltage, the instantaneous values ​​of the three-phase converter current, and the static... The voltage components in the coordinate system are subjected to synchronous coordinate system transformation and sequence component extraction to obtain the dq-axis voltage, dq-axis current, positive sequence voltage amplitude, and negative sequence voltage amplitude. The system state vector is constructed based on the dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency.

[0007] Optionally, based on the dq-axis current, positive-sequence voltage magnitude, negative-sequence voltage magnitude, and grid frequency, a system state vector is constructed, including: The positive sequence voltage amplitude is normalized based on the effective value of the rated line voltage to obtain the normalized voltage amplitude. Calculate the voltage imbalance based on the positive-sequence voltage amplitude and the negative-sequence voltage amplitude; The frequency deviation is obtained by subtracting the grid frequency from the rated frequency. The normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current are vectorized and assembled to obtain the system state vector.

[0008] Optionally, the voltage imbalance is calculated based on the positive-sequence voltage magnitude and the negative-sequence voltage magnitude, including: Calculate the voltage imbalance using the following formula: ; in, For voltage imbalance, The magnitude of the negative sequence voltage. This is the positive sequence voltage amplitude. It is a very small positive number.

[0009] Optionally, based on the system state vector, grid-guided dynamic estimation of the external d-axis current reference and the external q-axis current reference is performed to obtain the final dq-axis current reference, including: Operating conditions are determined and segmented calculations are performed based on the normalized voltage amplitude in the system state vector to obtain the reactive current increment required by the power grid guidelines. The external q-axis current reference and the reactive current increment required by the grid guidelines are superimposed to obtain the final q-axis current reference. Based on the converter's rated maximum output current and the final q-axis current reference, the external d-axis current reference is subjected to power current reference limiting to obtain the final d-axis current reference. The final q-axis current reference and the final d-axis current reference are combined to obtain the final dq-axis current reference.

[0010] Optionally, based on the converter's rated maximum output current and the final q-axis current reference, a power current reference limiting is applied to the external d-axis current reference to obtain the final d-axis current reference, including: The power current reference is limited to the external d-axis current reference using the following formula: ; ; in, The rated maximum output current of the converter, For the final q-axis current reference, This represents the maximum allowable value for the d-axis current. For external d-axis current reference, This serves as the final d-axis current reference.

[0011] Optionally, fuzzy inference is performed based on the system state vector to obtain an adaptive weight factor vector, including: Extract the normalized voltage magnitude and voltage imbalance from the system state vector; The membership degree of the extracted normalized voltage amplitude and voltage imbalance is calculated to obtain the voltage amplitude membership vector and the imbalance membership vector. Based on the fuzzy rule set and the output membership function, fuzzy reasoning and output aggregation are performed on the voltage amplitude membership vector and the unbalance membership vector to obtain the aggregated fuzzy output set of the d-axis weight, the aggregated fuzzy output set of the q-axis weight, and the aggregated fuzzy output set of the switching loss weight. The aggregated fuzzy output sets of the d-axis weights, the aggregated fuzzy output sets of the q-axis weights, and the aggregated fuzzy output sets of the switching loss weights are defuzzified to obtain the adaptive weight factor vector.

[0012] Optionally, future current states are predicted based on the current dq-axis current, the current dq-axis voltage, and the current switching state to obtain a set of all predicted currents, including: Based on the current dq-axis current, the current dq-axis voltage, and the current switching state, for each possible switching state, the discretized converter mathematical model is used to predict the dq-axis current at the next moment, so as to obtain the set of all predicted currents.

[0013] Optionally, based on the final dq-axis current reference, the adaptive weighting factor vector, and the current switching state, an adaptive cost function evaluation and optimal vector selection are performed on the set of all predicted currents to obtain the optimal switching state, including: Based on the final dq-axis current reference and the current switching state, performance indicators are calculated for the set of all predicted currents to obtain the d-axis current tracking error set, the q-axis current tracking error set, and the switching number variation set. Based on the adaptive weight factor vector, the total cost aggregation based on adaptive weight is performed on the d-axis current tracking error set, the q-axis current tracking error set, and the switching number change set to obtain the total cost set of all candidate vectors; Optimal switching state decision is made on the total value set of all candidate vectors to obtain the optimal switching state.

[0014] Another aspect of the present invention provides a wind power converter current optimization system based on model predictive control, the wind power converter current optimization system based on model predictive control comprising: The system state vector construction module is used to construct a system state vector based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current in the power grid connected to the wind power converter. The system state vector includes normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current. The current reference dynamic estimation module is used to perform grid-guided dynamic estimation of the external d-axis current reference and the external q-axis current reference based on the system state vector, so as to obtain the final dq-axis current reference. The fuzzy inference module is used to perform fuzzy inference based on the system state vector to obtain an adaptive weight factor vector; The future current state prediction module is used to predict the future current state based on the current dq-axis current, the current dq-axis voltage, and the current switching state, so as to obtain a set of all predicted currents. The optimal vector selection module is used to perform adaptive cost function evaluation and optimal vector selection on the set of all predicted currents based on the final dq-axis current reference, adaptive weight factor vector, and the current switching state, so as to obtain the optimal switching state.

[0015] Compared with existing technologies, this invention provides a wind power converter current optimization method and system based on model predictive control. It first constructs a system state vector that comprehensively reflects the real-time state of the power grid, and innovatively introduces fuzzy inference. Using key state variables such as normalized voltage amplitude and voltage imbalance as inputs, it generates adaptive weighting factors in real-time and dynamically through fuzzy inference, thereby adjusting the cost function of model predictive control online. In this way, it is possible to perform online trade-off optimization of multiple control objectives such as current tracking accuracy and switching losses under different power grid operating conditions, ultimately achieving accurate and robust control of converter current in complex power grid environments, thus significantly enhancing the adaptability of wind power converters to the power grid and the reliability of grid-connected operation. Attached Figure Description

[0016] One or more embodiments are illustrated by way of example with the corresponding pictures in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0017] Figure 1 A flowchart of a wind power converter current optimization method based on model predictive control according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the data flow of a wind power converter current optimization method based on model predictive control according to an embodiment of the present invention; Figure 3 This is a block diagram of a wind power converter current optimization system based on model predictive control according to an embodiment of the present invention. Detailed Implementation

[0018] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0019] As indicated in the specification and claims of this invention, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0020] While this invention makes various references to certain modules in systems according to embodiments of the invention, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0021] This invention uses flowcharts to illustrate the specific operational steps included in the method according to embodiments of the invention. It should be understood that the preceding or following operational steps are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously, as needed. Furthermore, other operations can be added to these processes, or one or more operations can be removed from these processes.

[0022] In the technical solution of this invention, a method for optimizing the current of a wind power converter based on model predictive control is proposed. Figure 1 This is a flowchart of a wind power converter current optimization method based on model predictive control according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the data flow in a wind power converter current optimization method based on model predictive control according to an embodiment of the present invention. Figure 1 and Figure 2 According to an embodiment of the present invention, a wind power converter current optimization method based on model predictive control includes the following steps: S1, constructing a system state vector based on the instantaneous values ​​of the three-phase voltage and the three-phase converter current of the power grid connected to the wind power converter; the system state vector includes normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current; S2, based on the system state vector, performing grid-guided dynamic estimation of the current reference driven by the external d-axis current reference and the external q-axis current reference to obtain the final dq-axis current reference; S3, performing fuzzy inference based on the system state vector to obtain an adaptive weight factor vector; S4, predicting the future current state based on the current dq-axis current, the current dq-axis voltage, and the current switching state to obtain a set of all predicted currents; S5, based on the final dq-axis current reference, the adaptive weight factor vector, and the current switching state, performing adaptive cost function evaluation and optimal vector selection on the set of all predicted currents to obtain the optimal switching state.

[0023] Specifically, S1 constructs a system state vector based on the instantaneous values ​​of the three-phase voltage and the instantaneous current of the three-phase converter in the grid connected to the wind turbine converter. The system state vector includes normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current. When connected to the grid, the wind turbine converter needs to cope with various complex grid conditions, such as voltage sags, frequency fluctuations, and voltage imbalances. A system state vector containing normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current can reflect the current operational health of the grid connected to the wind turbine converter and the electrical state of the wind turbine converter itself in real time and accurately. This provides the controller with necessary and sufficient input information, enabling the wind turbine converter to dynamically adjust its control objectives and strategies according to actual operating conditions. This ensures that the grid guidelines are met under various operating conditions, optimizes its own operating performance, and achieves stable grid connection and grid-friendly operation of the wind turbine.

[0024] In specific implementation, S1 includes: estimating the grid synchronization angle and frequency of the instantaneous three-phase voltage values ​​to obtain the estimated synchronization angle, grid frequency, and stationary voltage. Voltage components in a coordinate system. That is, by processing the instantaneous values ​​of the three-phase voltage acquired in real time, the synchronous rotation angle and actual operating frequency of the current power grid are identified. The synchronous angle is a key parameter for achieving coordinate transformation, while the power grid frequency is an important indicator of power grid operational stability. Simultaneously, this process also converts the original three-phase voltage signal into a static signal. The voltage component in the coordinate system provides a standard input format for subsequent sequence component extraction and dq-axis transformation.

[0025] The S1 further includes: based on the estimated synchronization angle, analyzing the instantaneous values ​​of the three-phase voltage, the instantaneous values ​​of the three-phase converter current, and the static... The voltage components in the coordinate system are transformed to a synchronous coordinate system and their sequence components are extracted to obtain the dq-axis voltage, dq-axis current, positive-sequence voltage amplitude, and negative-sequence voltage amplitude. In other words, after obtaining an accurate synchronization angle estimate, the measured instantaneous values ​​of the three-phase voltage, the instantaneous values ​​of the three-phase converter current, and the previously obtained static components are transformed using the Park transformation. Voltage components in a coordinate system, from the stationary reference frame, i.e., at rest The coordinate system is transformed to a dq coordinate system that rotates synchronously with the power grid to obtain the dq-axis voltage and current. The dq-axis current is a key variable for the decoupling control of active and reactive power in the wind power converter. Furthermore, to more precisely characterize the grid voltage condition, especially in handling grid voltage imbalances, the positive-sequence and negative-sequence voltage amplitudes are extracted from the dq-axis voltage. The positive-sequence voltage amplitude reflects the voltage level under normal grid operation, while the negative-sequence voltage amplitude quantifies the degree of grid voltage asymmetry, providing guidance for negative-sequence current compensation in wind power converters under unbalanced grid conditions.

[0026] S1 further includes constructing a system state vector based on dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency. This transforms the complex transient behaviors of the power grid and wind turbine converter into quantifiable characteristics, enabling the controller to accurately determine the current operating conditions, including the health of the grid voltage, frequency stability, and the power output mode of the wind turbine converter itself. By integrating key parameters such as dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency, the operating status of the power system and the interactive characteristics of the grid can be comprehensively perceived. For example, the positive-sequence voltage amplitude reflects the voltage level during normal grid operation, the negative-sequence voltage amplitude directly indicates the degree of voltage imbalance, the grid frequency reflects the power balance of the grid, and the dq-axis current quantifies the actual active and reactive power output of the converter. The accurate acquisition and assembly of these parameters is fundamental to enabling adaptive control and performance optimization of the wind turbine converter under various complex operating conditions, such as voltage dips, frequency fluctuations, or voltage imbalances, ensuring the continuous and stable operation of the wind turbine and compliance with stringent grid guidelines.

[0027] Specifically, based on the dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency, a system state vector is constructed. This includes normalizing the positive-sequence voltage amplitude based on the effective value of the rated line voltage to obtain a normalized voltage amplitude. This step aims to transform the key grid parameter, the positive-sequence voltage amplitude, into a standard dimensionless form. The normalized voltage amplitude is obtained by calculating the ratio of the measured positive-sequence voltage amplitude to the phase voltage amplitude corresponding to the preset effective value of the rated line voltage (usually referring to the RMS value of the line voltage when the power system is operating normally). This normalization process eliminates numerical differences between power systems of different voltage levels, making the voltage level universal and comparable in the control algorithm, facilitating a unified evaluation and judgment of the voltage state in subsequent fuzzy inference and other stages.

[0028] Based on the dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency, a system state vector is constructed. This also includes calculating the voltage imbalance based on the positive-sequence and negative-sequence voltage amplitudes. Voltage imbalance is an important indicator for measuring grid voltage quality and symmetry, especially in wind power grid-connected systems, where a large voltage imbalance can cause problems such as internal current loop coupling and increased heat generation in the converter. Therefore, in the technical solution of this invention, the positive-sequence and negative-sequence voltage amplitudes are used to quantify grid asymmetry. Specifically, calculating the voltage imbalance based on the positive-sequence and negative-sequence voltage amplitudes includes calculating the voltage imbalance using the following formula: ; in, For voltage imbalance, The magnitude of the negative sequence voltage. This is the positive sequence voltage amplitude. It is a very small positive number.

[0029] Based on the dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency, a system state vector is constructed. This includes subtracting the grid frequency from the rated frequency to obtain the frequency deviation. Grid frequency is one of the core indicators of power system operational stability, and its deviation reflects the supply and demand balance of active power in the power system. This step aims to calculate the frequency deviation by comparing the real-time monitored grid frequency with the preset rated frequency. The magnitude and direction of the frequency deviation can indicate whether there is a power deficit or surplus in the grid, providing direct guidance for wind power converters to provide frequency support or adjust their active power output to help maintain grid frequency stability.

[0030] After obtaining the normalized voltage amplitude, voltage imbalance, and frequency deviation, a system state vector is constructed based on the dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency. This also includes vectorizing and assembling the normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current to obtain the system state vector.

[0031] Specifically, S2, based on the system state vector, performs grid-guided dynamic estimation of the external d-axis current reference and external q-axis current reference to obtain the final dq-axis current reference. This ensures that the wind turbine converter strictly adheres to grid guidelines under any grid operating conditions, especially when facing abnormal situations such as voltage dips and frequency fluctuations, while avoiding exceeding its rated carrying capacity. Traditional control strategies often use fixed current reference values, which cannot provide sufficient flexibility to cope with rapidly changing grid environments, potentially leading to the wind turbine converter's inability to respond to grid events in a timely manner, or even triggering a chain reaction. This embodiment, through dynamic estimation of the current reference, can adjust the active and reactive power output of the wind turbine converter in real time. It not only prioritizes grid stability requirements (e.g., injecting reactive current during voltage dips) but also optimizes active power output to ensure economy, and ensures that all operations are performed within the safe operating range of the wind turbine converter, thereby significantly improving the grid friendliness and operational reliability of the wind turbine unit.

[0032] In specific implementation, S2 includes: performing operating condition judgment and segmented calculation based on the normalized voltage amplitude in the system state vector to obtain the reactive current increment required by the grid guidelines. During this process, the normalized voltage amplitude of the current grid is acquired in real time. This normalized voltage amplitude, as a core component of the system state vector, accurately reflects the health status of the grid voltage. The controller can use the normalized voltage amplitude to determine the current grid operating condition (e.g., whether it is undervoltage, overvoltage, or normal) according to preset grid guidelines. For different operating conditions, grid guidelines typically have a series of segmented or continuous provisions requiring wind turbine converters to inject or absorb a specific amount of reactive current to support the grid voltage. For example, under voltage dip conditions, grid guidelines may require the reactive current injected by the wind turbine converter to be proportional to the voltage dip depth. Through this operating condition judgment and segmented calculation, the reactive current increment urgently needed by the current grid can be dynamically determined to enhance the grid's voltage support capability.

[0033] S2 further includes: superimposing instructions on the external q-axis current reference and the reactive current increment required by the grid guidelines to obtain the final q-axis current reference. That is, after obtaining the reactive current increment required by the grid guidelines, it is effectively superimposed on the external q-axis current reference provided by the upper-level control system (such as a reactive power optimization or voltage control module). The external q-axis current reference typically represents the reactive power demand under ideal or normal operating conditions. The final q-axis current reference obtained through superposition will simultaneously take into account both the normal reactive power demand and the mandatory reactive power support requirements of the grid guidelines under abnormal operating conditions. This means that while providing basic reactive power support, the wind power converter can quickly and accurately respond to the grid's emergency needs, such as rapidly injecting additional reactive current to boost voltage during voltage dips.

[0034] S2 further includes: limiting the external d-axis current reference based on the converter's rated maximum output current and the final q-axis current reference to obtain the final d-axis current reference. The physical characteristics of the wind power converter determine that its total output current has a rated maximum value, i.e., the converter's rated maximum output current. When the final q-axis current reference becomes larger due to grid guidelines, the available d-axis current (corresponding to active power) output capability of the wind power converter will decrease accordingly to avoid the total current exceeding its rated maximum output current. Therefore, in the technical solution of this invention, the external d-axis current reference provided by the upper-level control system is limited by considering the converter's rated maximum output current and the determined final q-axis current reference. Specifically, the external d-axis current reference is limited using the following formula: ; ; in, The rated maximum output current of the converter, For the final q-axis current reference, This represents the maximum allowable value for the d-axis current. For external d-axis current reference, This serves as the final d-axis current reference. By using the above formula to limit the active current reference of the external d-axis current, it is ensured that the output of the d-axis current of the wind power converter is reasonably limited while meeting the reactive current requirements. This ensures that the total current of the wind power converter does not exceed its rated maximum output current, thereby protecting the equipment and maintaining the safe and stable operation of the system.

[0035] S2 further includes: integrating the final q-axis current reference and the final d-axis current reference to obtain the final dq-axis current reference. After the above dynamic adjustment and limiting processing, the obtained final d-axis current reference and final q-axis current reference are integrated into a complete final dq-axis current reference vector. This vector is the direct input of the model predictive control algorithm, which comprehensively reflects the power target from the upper control, the real-time changing grid guidance requirements, and the physical operating limits of the converter.

[0036] Specifically, S3 involves fuzzy inference based on the system state vector to obtain an adaptive weight factor vector. It should be understood that wind power converters face complex and variable grid conditions, and the priority requirements for control objectives (such as current tracking accuracy, switching losses, and grid support responsibilities) also change dynamically. If pre-set fixed weights are used, the MPC controller cannot achieve optimal control performance under all operating conditions, potentially leading to poor performance under critical conditions or even failure to meet grid guidelines. This embodiment, by introducing fuzzy inference, can intelligently adjust the weights of various control objectives in the cost function online based on real-time changes in the system state vector (especially grid health status), thereby achieving flexible and adaptive control strategies. This means that during normal grid operation, the focus may be more on optimizing efficiency (reducing switching losses), while during abnormal situations such as voltage drops or imbalances in the grid, the priority of current tracking accuracy and grid support objectives can be automatically increased, ensuring that the wind power converter can dynamically adapt to various operating conditions and improving the overall adaptability and operational reliability of the wind power system.

[0037] In specific implementation, S3 includes: extracting the normalized voltage amplitude and voltage imbalance from the system state vector. The normalized voltage amplitude and voltage imbalance are directly related to the stability and symmetry of the grid voltage, playing a decisive role in determining the current grid operating conditions and prioritizing the wind power converter control strategy. For example, the normalized voltage amplitude indicates whether the grid is undervoltage or overvoltage, while the voltage imbalance reveals the degree of grid asymmetry. Both require the wind power converter to make different responses to comply with grid guidelines.

[0038] S3 further includes: calculating the membership degrees of the extracted normalized voltage amplitude and voltage imbalance to obtain voltage amplitude membership vectors and imbalance membership vectors. During this process, a predefined membership function is used for fuzzification. The membership function maps the precise input value to one or more fuzzy sets (such as "normal voltage," "low voltage," "low imbalance," and "high imbalance"), and assigns a membership value between 0 and 1 to each fuzzy set. For example, voltage amplitude might have fuzzy sets such as "normal," "slight drop," and "severe drop," while voltage imbalance might have fuzzy sets such as "low," "medium," and "high." Through this calculation, the membership function generates a corresponding membership vector for each input value, including the normalized voltage amplitude and voltage imbalance, to indicate its degree of belonging to each fuzzy set.

[0039] S3 further includes: based on the fuzzy rule set and output membership function, performing fuzzy inference and output aggregation on the voltage amplitude membership vector and the imbalance membership vector to obtain the aggregated fuzzy output set of the d-axis weight, the aggregated fuzzy output set of the q-axis weight, and the aggregated fuzzy output set of the switching loss weight. The offline-designed fuzzy rule set consists of a series of "IF-THEN" statements, which associate the input fuzzy variables (including the voltage amplitude membership vector and the imbalance membership vector) with the output fuzzy variables (including the d-axis weight, the q-axis weight, and the switching loss weight). For example, a fuzzy rule might be "IF voltage is 'slightly dropped' AND imbalance is 'low' THEN q-axis weight is 'medium-high' AND d-axis weight is 'medium' AND switching loss weight is 'medium-low'". Traversing all activated fuzzy rules, and combining the input membership vectors including the voltage amplitude membership vector and the imbalance membership vector, the activation strength of each fuzzy rule is determined through fuzzy logic operations (such as minimum value, product, etc.). Subsequently, using a preset output membership function, the result of each fuzzy rule is mapped to the corresponding output fuzzy set. Finally, through a certain aggregation method (such as the maximum value method, the sum method, etc.), the fuzzy output sets of all fuzzy rules are merged to obtain the aggregated fuzzy output sets of the d-axis weight, q-axis weight, and switching loss weight, respectively.

[0040] S3 further includes defuzzifying the aggregated fuzzy output sets of the d-axis weights, q-axis weights, and switching loss weights to obtain an adaptive weight factor vector. It should be understood that after completing fuzzy inference and output aggregation, the resulting outputs of the d-axis weights, q-axis weights, and switching loss weights—that is, the aggregated fuzzy output sets of the d-axis weights, q-axis weights, and switching loss weights—are still fuzzy sets. To enable these fuzzy results to be practically applied in the subsequent cost function, they need to be defuzzified to convert them back to precise values. Commonly used defuzzification methods include the centroid method and the center of area method. These defuzzification methods calculate a representative precise value based on the shape and membership distribution of the fuzzy output sets. By performing defuzzification operations on the three aggregated fuzzy output sets—namely, the aggregated fuzzy output set of the d-axis weight, the aggregated fuzzy output set of the q-axis weight, and the aggregated fuzzy output set of the switching loss weight—three precise values ​​are obtained. These three precise values ​​together constitute the adaptive weight factor vector, which serves as the key input for the adaptive cost function evaluation in the next step, enabling online dynamic adjustment of the MPC control weights.

[0041] Specifically, S4 predicts future current states based on the current dq-axis current, current dq-axis voltage, and current switching state to obtain a set of all predicted currents. It should be understood that the core idea of ​​Model Predictive Control (MPC) is prediction and optimization; that is, by accurately predicting the future behavior of the power system, evaluating the consequences of each possible control action, and selecting the optimal control action that minimizes the cost function value. Therefore, an accurate and comprehensive prediction of future current states provides reliable input for subsequent adaptive cost function evaluation, enabling the controller to predict the impact of different switching state combinations on the wind power converter current in each sampling period. This is crucial for wind power converters to maintain current tracking accuracy, meet grid guidelines, and optimize switching losses in rapidly changing grid environments. By traversing all possible switching states and predicting the future currents they each cause, the controller can make the decision that best fits the current objectives and constraints among a range of potential outcomes, thereby significantly improving the speed, accuracy, and overall control performance of the wind power converter response.

[0042] Specifically, S4 aims to obtain all potential future current states based on the electrical model of the wind power converter, starting from the currently known power system state, by simulating the impact of each possible switching state on the wind power converter current in the next sampling period.

[0043] For example, based on the current dq-axis current, the current dq-axis voltage, and the current switching state, future current states are predicted to obtain a set of all predicted currents. This includes: based on the current dq-axis current, the current dq-axis voltage, and the current switching state, for each possible switching state, using a discretized converter mathematical model to predict the dq-axis current at the next moment, to obtain a set of all predicted currents.

[0044] In specific implementation, S4 first obtains the dq-axis current and dq-axis voltage at the current moment. These are the instantaneous measurements of the electrical output of the wind power converter and the power grid environment in which it is located. They represent the actual starting point of the power system at the current moment k. At the same time, the switching state at the current moment is obtained. Then, for each possible switching state, the discretized converter mathematical model is used to predict the dq-axis current at the next moment, i.e., the next sampling moment k+1, so as to obtain the set of all predicted currents. Taking a typical three-phase two-level voltage source converter as an example, it usually has eight basic switching states (including six effective vectors and two zero vectors). For each possible switching state, the converter will output the corresponding dq-axis voltage components, which will drive the converter output current to change. Discretizing this continuous-time model, usually using the Euler forward difference method, can yield a current prediction model. By traversing all eight possible switching states of the converter and substituting the current prediction model for each switching state, the corresponding predicted current can be calculated, thus obtaining eight different combinations of switching states and predicted currents. These combinations ultimately constitute the set of all predicted currents.

[0045] Specifically, in step S5, based on the final dq-axis current reference, the adaptive weighting factor vector, and the current switching state, an adaptive cost function evaluation and optimal vector selection are performed on the set of all predicted currents to obtain the optimal switching state. That is, through a quantifiable and optimizable evaluation mechanism, a switching state that achieves the best overall performance under the current operating conditions is selected from all possible future control actions. Here, it should be understood that the previous step has predicted the future currents resulting from all possible switching states, but these predictions themselves do not provide a decision. Therefore, in the technical solution of this invention, by introducing an adaptive cost function, these prediction results are compared with the desired control objective (final dq-axis current reference), and combined with an adaptive weight generated by fuzzy inference that reflects the current control priority, each candidate switching state is scored. This systematic evaluation and selection ensures that the controller does not merely passively track the reference value, but actively and proactively selects a control behavior that can achieve the most important current objective at the lowest cost (such as the lowest current tracking error and the lowest switching loss) in each control cycle, thereby ultimately achieving high-performance, high-efficiency, and high-reliability operation of the wind power converter in a complex and variable power grid environment.

[0046] In specific implementation, S5 includes: calculating performance indicators for the set of all predicted currents based on the final dq-axis current reference and the current switching state, to obtain the d-axis current tracking error set, the q-axis current tracking error set, and the switching frequency change set. This step quantifies the performance of each candidate switching state (corresponding to each predicted current) across various performance dimensions. For each predicted current pair in the set of all predicted currents obtained in the previous step, i.e., each combination of switching state and predicted current, the deviation between it and the final dq-axis current reference is calculated, i.e., the current tracking errors of the d-axis and q-axis. These errors are usually expressed in absolute or squared form to quantify the accuracy of control. At the same time, to evaluate the economy of control actions (i.e., switching losses), the switching frequency change is also calculated. This is achieved by comparing the difference between the candidate switching state that generates the predicted current and the current switching state. For a three-phase bridge arm, the switching frequency change is usually calculated as the sum of the number of switches in the three bridge arms that undergo state changes (from on to off or from off to on). Repeat this process for all predicted currents (i.e., all candidate switch states) to finally obtain the d-axis current tracking error set, the q-axis current tracking error set, and the switch count change set. Each element in the set corresponds one-to-one with a candidate switch state.

[0047] S5 further includes: based on the adaptive weight factor vector, performing total cost aggregation on the d-axis current tracking error set, the q-axis current tracking error set, and the switching frequency change set, to obtain the total cost set of all candidate vectors. That is, after obtaining each independent performance index, the adaptive weight factor vector generated by fuzzy inference aggregates these performance indices into a single, comprehensive total cost. The design of the total cost function is one of the core aspects of model predictive control. It reflects the degree of importance given to different control objectives under the current operating condition through linear weighting. For example, when the grid voltage drops, fuzzy inference increases the weights, making the cost function more sensitive to the q-axis current tracking error, thus prioritizing the accurate injection of reactive current. By performing total cost aggregation on the performance index set corresponding to each candidate switching state based on adaptive weights, the total cost set of all candidate vectors is calculated. Each value in this set represents the total cost incurred in selecting the corresponding switching state.

[0048] S5 further includes: making an optimal switching state decision on the total cost set of all candidate vectors to obtain the optimal switching state. In this process, the total cost set of all candidate vectors calculated in the previous step is traversed, and the minimum value is found as the minimum total cost. The candidate switching state that generates this minimum total cost value is considered the optimal switching state within the current control cycle. This optimal switching state will be sent to the wind power converter's drive circuit at the next sampling time to actually control the on / off state of the power semiconductors, thereby guiding the wind power converter's operating state towards cost minimization. This process is repeated cyclically, with prediction, evaluation, and optimal decision-making performed once in each sampling cycle to achieve rolling optimization control of the wind power converter current.

[0049] In summary, the model predictive control-based wind power converter current optimization method according to embodiments of the present invention is elucidated. It first constructs a system state vector that comprehensively reflects the real-time state of the power grid, and innovatively introduces fuzzy inference. Using key state variables such as normalized voltage amplitude and voltage imbalance as inputs, it generates adaptive weighting factors in real-time and dynamically through fuzzy inference, thereby adjusting the cost function of model predictive control online. In this way, multiple control objectives, such as current tracking accuracy and switching losses, can be optimized online under different power grid operating conditions. Ultimately, it achieves accurate and robust control of the converter current in complex power grid environments, thereby significantly enhancing the adaptability of the wind power converter to the power grid and the reliability of its grid-connected operation.

[0050] This invention also provides a wind power converter current optimization system based on model predictive control.

[0051] Figure 3 This is a block diagram of a wind power converter current optimization system based on model predictive control according to an embodiment of the present invention. Figure 3As shown, the model predictive control-based wind power converter current optimization system 300 according to an embodiment of the present invention includes: a system state vector construction module 310, used to construct a system state vector based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current of the power grid connected to the wind power converter; the system state vector includes normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current; and a current reference dynamic estimation module 320, used to perform grid-guided dynamic estimation of the external d-axis current reference and the external q-axis current reference based on the system state vector to obtain the final dq-axis current. The system includes a flow reference; a fuzzy inference module 330, which performs fuzzy inference based on the system state vector to obtain an adaptive weight factor vector; a future current state prediction module 340, which predicts the future current state based on the current dq-axis current, the current dq-axis voltage, and the current switching state to obtain a set of all predicted currents; and an optimal vector selection module 350, which performs adaptive cost function evaluation and optimal vector selection on the set of all predicted currents based on the final dq-axis current reference, the adaptive weight factor vector, and the current switching state to obtain the optimal switching state.

[0052] The specific implementation method of the wind power converter current optimization system based on model predictive control provided in this embodiment of the invention can be found in the wind power converter current optimization method based on model predictive control provided in this embodiment of the invention, and will not be repeated here.

[0053] The model predictive control-based wind power converter current optimization system 300 according to embodiments of the present invention can be implemented in various wireless terminals, such as servers with model predictive control-based wind power converter current optimization algorithms. In one possible implementation, the model predictive control-based wind power converter current optimization system 300 according to embodiments of the present invention can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the model predictive control-based wind power converter current optimization system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the model predictive control-based wind power converter current optimization system 300 can also be one of many hardware modules of the wireless terminal.

[0054] Alternatively, in another example, the model predictive control-based wind power converter current optimization system 300 and the wireless terminal can also be separate devices, and the model predictive control-based wind power converter current optimization system 300 can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0055] Those skilled in the art will understand that the above embodiments are specific implementations of the present invention, and in practical applications, various changes can be made in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A method for optimizing the current of a wind power converter based on model predictive control, characterized in that, The wind power converter current optimization method based on model predictive control includes: Based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current in the power grid connected to the wind power converter, a system state vector is constructed. The system state vector includes the normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current. Based on the system state vector, the external d-axis current reference and the external q-axis current reference are dynamically estimated by grid guidelines to obtain the final dq-axis current reference. Fuzzy reasoning is performed based on the system state vector to obtain the adaptive weight factor vector; Based on the current dq-axis current, the current dq-axis voltage, and the current switching state, predict the future current state to obtain a set of all predicted currents; Based on the final dq-axis current reference, the adaptive weight factor vector, and the current switching state, an adaptive cost function evaluation and optimal vector selection are performed on the set of all predicted currents to obtain the optimal switching state.

2. The wind power converter current optimization method based on model predictive control according to claim 1, characterized in that, Based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current in the power grid connected to the wind power converter, a system state vector is constructed, including: The instantaneous values ​​of the three-phase voltages are used to estimate the grid synchronization angle and frequency to obtain the estimated synchronization angle, grid frequency, and static voltage. Voltage components in a coordinate system; Based on the estimated synchronization angle, the instantaneous values ​​of the three-phase voltage, the instantaneous values ​​of the three-phase converter current, and the static... The voltage components in the coordinate system are subjected to synchronous coordinate system transformation and sequence component extraction to obtain the dq-axis voltage, dq-axis current, positive sequence voltage amplitude, and negative sequence voltage amplitude. The system state vector is constructed based on the dq-axis current, positive-sequence voltage amplitude, negative-sequence voltage amplitude, and grid frequency.

3. The wind power converter current optimization method based on model predictive control according to claim 2, characterized in that, Based on the dq-axis current, positive-sequence voltage magnitude, negative-sequence voltage magnitude, and grid frequency, a system state vector is constructed, including: The positive sequence voltage amplitude is normalized based on the effective value of the rated line voltage to obtain the normalized voltage amplitude. Calculate the voltage imbalance based on the positive-sequence voltage amplitude and the negative-sequence voltage amplitude; The frequency deviation is obtained by subtracting the grid frequency from the rated frequency. The normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current are vectorized and assembled to obtain the system state vector.

4. The wind power converter current optimization method based on model predictive control according to claim 3, characterized in that, Based on the positive-sequence voltage magnitude and the negative-sequence voltage magnitude, the voltage imbalance is calculated, including: Calculate the voltage imbalance using the following formula: ； in, For voltage imbalance, The magnitude of the negative sequence voltage. This is the positive sequence voltage amplitude. It is a very small positive number.

5. The wind power converter current optimization method based on model predictive control according to claim 1, characterized in that, Based on the system state vector, dynamic estimation of the current reference driven by grid guidelines is performed on the external d-axis current reference and the external q-axis current reference to obtain the final dq-axis current reference, including: Operating conditions are determined and segmented calculations are performed based on the normalized voltage amplitude in the system state vector to obtain the reactive current increment required by the power grid guidelines. The external q-axis current reference and the reactive current increment required by the grid guidelines are superimposed to obtain the final q-axis current reference. Based on the converter's rated maximum output current and the final q-axis current reference, the external d-axis current reference is subjected to power current reference limiting to obtain the final d-axis current reference. The final q-axis current reference and the final d-axis current reference are combined to obtain the final dq-axis current reference.

6. The wind power converter current optimization method based on model predictive control according to claim 5, characterized in that, Based on the converter's rated maximum output current and the final q-axis current reference, the external d-axis current reference is subjected to power current reference limiting to obtain the final d-axis current reference, including: The power current reference is limited to the external d-axis current reference using the following formula: ； ； in, The rated maximum output current of the converter, For the final q-axis current reference, This represents the maximum allowable value for the d-axis current. For external d-axis current reference, This serves as the final d-axis current reference.

7. The wind power converter current optimization method based on model predictive control according to claim 1, characterized in that, Fuzzy inference is performed based on the system state vector to obtain an adaptive weight factor vector, including: Extract the normalized voltage magnitude and voltage imbalance from the system state vector; The membership degree of the extracted normalized voltage amplitude and voltage imbalance is calculated to obtain the voltage amplitude membership vector and the imbalance membership vector. Based on the fuzzy rule set and the output membership function, fuzzy reasoning and output aggregation are performed on the voltage amplitude membership vector and the unbalance membership vector to obtain the aggregated fuzzy output set of the d-axis weight, the aggregated fuzzy output set of the q-axis weight, and the aggregated fuzzy output set of the switching loss weight. The aggregated fuzzy output sets of the d-axis weights, the aggregated fuzzy output sets of the q-axis weights, and the aggregated fuzzy output sets of the switching loss weights are defuzzified to obtain the adaptive weight factor vector.

8. The wind power converter current optimization method based on model predictive control according to claim 1, characterized in that, Based on the current dq-axis current, the current dq-axis voltage, and the current switching state, future current states are predicted to obtain a set of all predicted currents, including: Based on the current dq-axis current, the current dq-axis voltage, and the current switching state, for each possible switching state, the discretized converter mathematical model is used to predict the dq-axis current at the next moment, so as to obtain the set of all predicted currents.

9. The wind power converter current optimization method based on model predictive control according to claim 1, characterized in that, Based on the final dq-axis current reference, the adaptive weighting factor vector, and the current switching state, an adaptive cost function is evaluated and the optimal vector is selected for the set of all predicted currents to obtain the optimal switching state, including: Based on the final dq-axis current reference and the current switching state, performance indicators are calculated for the set of all predicted currents to obtain the d-axis current tracking error set, the q-axis current tracking error set, and the switching number variation set. Based on the adaptive weight factor vector, the total cost aggregation based on adaptive weight is performed on the d-axis current tracking error set, the q-axis current tracking error set, and the switching number change set to obtain the total cost set of all candidate vectors; Optimal switching state decision is made on the total value set of all candidate vectors to obtain the optimal switching state.

10. A wind power converter current optimization system based on model predictive control, characterized in that, The model predictive control-based wind power converter current optimization system includes: The system state vector construction module is used to construct a system state vector based on the instantaneous values ​​of the three-phase voltage and the instantaneous values ​​of the three-phase converter current in the power grid connected to the wind power converter. The system state vector includes normalized voltage amplitude, voltage imbalance, frequency deviation, and dq-axis current. The current reference dynamic estimation module is used to perform grid-guided dynamic estimation of the external d-axis current reference and the external q-axis current reference based on the system state vector, so as to obtain the final dq-axis current reference. The fuzzy inference module is used to perform fuzzy inference based on the system state vector to obtain an adaptive weight factor vector; The future current state prediction module is used to predict the future current state based on the current dq-axis current, the current dq-axis voltage, and the current switching state, so as to obtain a set of all predicted currents. The optimal vector selection module is used to perform adaptive cost function evaluation and optimal vector selection on the set of all predicted currents based on the final dq-axis current reference, adaptive weight factor vector, and the current switching state, so as to obtain the optimal switching state.

Images