Parallel inverter circulating current suppression method and system based on model predictive control

Abstract

This invention discloses a method and system for suppressing circulating current in parallel inverters based on model predictive control, relating to the field of circulating current suppression technology for parallel inverters. The method includes: real-time acquisition of the output current, output voltage, and carrier phase of each inverter module to identify the dominant circulating current source; extraction of the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and calculation of the circulating current deviation degree; determination of the circulating current evolution trend based on the rate of change of the circulating current deviation degree and historical circulating current trajectory; when the evolution trend is continuously deteriorating, using the circulating current deviation degree and evolution trend as constraints, predicting the future evolution trajectory of the circulating current deviation degree through a model predictive controller, and solving for the voltage compensation amount and carrier phase adjustment amount; and performing differentiated compensation control on the two modules in the dominant circulating current source. This invention achieves precise location and proactive suppression of the circulating current source, improving the response speed and control accuracy of circulating current suppression.

Description

Technical Field

[0001] This invention relates to the field of circulating current suppression technology for parallel inverters, and specifically to a method and system for suppressing circulating current in parallel inverters based on model predictive control. Background Technology

[0002] Due to issues such as differences in output filters, asynchronous carrier phases, line impedance mismatches, and inconsistent control parameters among parallel inverter modules, circulating currents can occur between the modules. Circulating currents refer to the currents flowing between parallel inverter modules that do not flow to the load. Their presence can lead to increased inverter losses, increased stress on power devices, and distortion of output current. In severe cases, it can even trigger overcurrent protection, threatening the safe operation of the system.

[0003] Existing circulating current suppression methods for parallel inverters mainly fall into two categories: hardware and software solutions. Hardware solutions increase circulating current impedance by adding isolation transformers or coupling inductors, but this increases system size, weight, and cost. Software solutions primarily employ master-slave control, droop control, or coordinated control strategies to suppress circulating current by adjusting the output voltage amplitude and phase of each module. However, traditional methods typically adjust all modules uniformly, failing to identify the dominant circulating current source, resulting in low regulation efficiency and potentially causing unnecessary disturbances to modules with smaller circulating currents. Secondly, they lack the ability to judge the evolution trend of circulating currents, failing to distinguish between naturally converging transient circulating currents and continuously deteriorating fault circulating currents, and applying control actions even when the circulating current is naturally converging, causing unnecessary adjustments. Thirdly, existing methods use fixed-parameter PI controllers, which are difficult to adapt to the nonlinear dynamic characteristics of circulating currents, resulting in lag in control response and limited suppression effectiveness.

[0004] Therefore, there is an urgent need for a circulation suppression method that can identify the dominant circulation source, determine the circulation evolution trend, and achieve precise compensation. Summary of the Invention

[0005] This invention addresses the technical problems in existing technologies, such as low regulation efficiency due to failure to identify the dominant circulating current source, unnecessary regulation due to lack of ability to judge the evolution trend of circulating current, and difficulty of fixed parameter controllers to adapt to the nonlinear dynamic characteristics of circulating current. It provides a circulating current suppression method and system for parallel inverters based on model predictive control.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a method for suppressing circulating current in parallel inverters based on model predictive control, comprising: The system collects the output current, output voltage, and carrier phase of each inverter module in real time, calculates the circulating current value between any two modules, and identifies the module pair with the largest circulating current amplitude as the dominant circulating current source. Extract the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and calculate the deviation between the circulating current value of the dominant circulating current source and the average circulating current value of the system as the circulating current deviation. Based on the rate of change of the circulation deviation and the historical circulation trajectory, the circulation evolution trend of the dominant circulation source is determined, and the circulation evolution trend includes natural convergence and continuous deterioration; When the circulating current evolution trend is continuously deteriorating, the circulating current deviation and the circulating current evolution trend are used as constraints. Combined with the physical constraint that the sum of the output currents of each module is equal to the load current, the model prediction controller predicts the future circulating current deviation trajectory. With the goal of minimizing the future circulating current deviation peak, the voltage compensation amount and carrier phase adjustment amount of the two modules in the dominant circulating current source are solved. Based on the voltage compensation amount and carrier phase adjustment amount, differentiated compensation control is performed on the two modules in the dominant circulating current source.

[0007] Secondly, the present invention provides a circulating current suppression system for parallel inverters based on model predictive control, comprising: The data acquisition module is used to collect the output current, output voltage and carrier phase of each inverter module in real time, calculate the circulating current value between any two modules, and identify the module pair with the largest circulating current amplitude as the dominant circulating current source. The extraction and calculation module is used to extract the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and to calculate the deviation between the circulating current value of the dominant circulating current source and the average circulating current value of the system as the circulating current deviation. The judgment module is used to determine the circulation evolution trend of the dominant circulation source based on the rate of change of the circulation deviation and the historical circulation trajectory. The circulation evolution trend includes natural convergence and continuous deterioration. The prediction and solution module is used to predict the future trajectory of the circulating current deviation when the circulating current evolution trend is continuously deteriorating, taking the circulating current deviation and the circulating current evolution trend as constraints, and combining the physical constraint that the sum of the output currents of each module is equal to the load current, through the model prediction controller, and with the goal of minimizing the peak value of the future circulating current deviation, to solve the voltage compensation amount and carrier phase adjustment amount of the two modules in the dominant circulating current source; The differential compensation control module is used to perform differential compensation control on the two modules in the dominant circulating current source according to the voltage compensation amount and the carrier phase adjustment amount.

[0008] The beneficial effects of this invention are: Compared to existing technologies, this invention first identifies the dominant circulating current source by comparing circulating current amplitudes, achieving precise targeting of circulating current suppression and avoiding unnecessary adjustments to non-faulty modules. Secondly, by constructing the circulating current deviation and analyzing its rate of change and historical trajectory, it accurately determines whether the circulating current is in a state of natural convergence or continuous deterioration. During natural convergence, no intervention is required, reducing redundant control actions. Thirdly, for continuously deteriorating conditions, a model predictive controller is used to proactively solve for voltage compensation and carrier phase adjustment with the goal of minimizing the future peak value of the circulating current deviation, overcoming the lag in response of traditional PI control. Finally, differentiated compensation is implemented for the two modules in the dominant circulating current source, blocking the circulating current path at its source. This invention achieves an organic combination of precise location of the circulating current source, intelligent identification of its evolution trend, and proactive suppression, improving the circulating current suppression effect and dynamic response performance of parallel inverter systems. Attached Figure Description

[0009] Figure 1 A schematic flowchart of the parallel inverter circulating current suppression method based on model predictive control provided by the present invention; Figure 2 This is a schematic diagram of the circulating current suppression system for parallel inverters based on model predictive control provided by the present invention.

[0010] In the attached diagram, the components represented by each number are as follows: The system includes a data acquisition module 11, an extraction and calculation module 12, a judgment module 13, a prediction and solution module 14, and a differential compensation control module 15. Detailed Implementation

[0011] Example 1, as Figure 1 As shown, this embodiment of the invention provides a method for suppressing circulating current in parallel inverters based on model predictive control, including: S10: Real-time acquisition of output current, output voltage and carrier phase of each inverter module, calculation of circulating current value between any two modules, and identification of the module pair with the largest circulating current amplitude as the dominant circulating current source; First, the output current, output voltage, and carrier phase of each inverter module are collected in real time. In a parallel inverter system, the output terminals of multiple inverter modules are connected in parallel to a common load, and circulating current may exist between the modules. To accurately obtain the system operating status, the instantaneous values ​​of the output current, output voltage, and carrier phase angle of each inverter module are collected synchronously at a preset sampling frequency.

[0012] Among them, the output current and output voltage reflect the real-time power output status of each module, and the carrier phase angle reflects the phase position of the pulse width modulation signal of each module. The above three types of parameters are the basic data for calculating circulating current and analyzing the causes of circulating current.

[0013] Secondly, the circulating current value between any two modules is calculated. For inverter systems operating in parallel, the circulating current is essentially the inter-module circulating current caused by differences in the amplitude or phase of the output voltage of each module. For any two modules i and j, the instantaneous value of the circulating current is equal to half the difference between the output current of module i and the output current of module j. To assess the severity of the circulating current, the root mean square (RMS) value of the instantaneous circulating current within a preset time window is calculated as the circulating current amplitude between module i and module j. The RMS value can effectively characterize the effective energy level of the AC circulating current within the time window.

[0014] Finally, the module pair with the largest circulating current amplitude is identified as the dominant circulating current source. All module pairs in the system are traversed, and the circulating current amplitude between each pair is calculated. The module pair with the largest circulating current amplitude is marked as the dominant circulating current source. The dominant circulating current source is the module pair with the most severe circulating current in the system and is also the key target for circulating current suppression. Traditional methods uniformly adjust all modules, but circulating current usually occurs concentrated between specific module pairs. Applying control to other modules not only has limited effect but may also introduce new disturbances. By identifying the dominant circulating current source, control resources can be focused on the module pair that most needs intervention, achieving precise regulation and improving circulating current suppression efficiency. Specifically, the starting point of this step is that the circulating current distribution in a parallel inverter system is non-uniform, and the module pair with the largest circulating current amplitude is usually the source of circulating current generation. Identifying and prioritizing the treatment of the dominant circulating current source can fundamentally suppress system circulating current.

[0015] Specifically, the output current, output voltage, and carrier phase of each inverter module are collected in real time. The circulating current value between any two modules is calculated, and the module pair with the largest circulating current amplitude is identified as the dominant circulating current source, including: The instantaneous values ​​of the output current, output voltage, and carrier phase angle of each inverter module are synchronously collected at a preset sampling frequency. For any two modules i and j, calculate the instantaneous value of the circulating current based on the instantaneous values ​​of their output currents. The instantaneous value of the circulating current is equal to half the difference between the output current of module i and the output current of module j. Calculate the root mean square value of the instantaneous value of the circulation within the preset time window, and use it as the circulation amplitude between module i and module j; Traverse all module pairs and mark the module pair with the largest circulation amplitude as the dominant circulation source.

[0016] First, the instantaneous values ​​of the output current, output voltage, and carrier phase angle of each inverter module are synchronously acquired at a preset sampling frequency. The preset sampling frequency should be set according to the inverter switching frequency and circulating current dynamic characteristics; for example, it can be set to 10 kHz to 20 kHz to ensure complete capture of the instantaneous changes in the circulating current. Synchronous acquisition means that the parameters of all modules are sampled simultaneously at the same sampling time to ensure time alignment of the data from each module during subsequent circulating current calculations. The instantaneous output voltage value reflects the voltage output state of the module, and the carrier phase angle reflects the phase position of the module's pulse width modulation signal; both are important bases for analyzing the causes of circulating current.

[0017] Secondly, for any two modules i and j, calculate the instantaneous value of the circulating current based on their instantaneous output current values. Circulating current refers to the current flowing between parallel inverter modules that does not flow to the load. Let the instantaneous output current of module i be ij. i The instantaneous output current value of module j is i j Then the instantaneous value of the circulating current between module i and module j is equal to half the difference between the output current of module i and the output current of module j, that is (i i -i j The physical meaning of this formula is that the circulating current is zero when the output currents of the two modules are exactly equal, and the greater the difference, the greater the circulating current. Through the above calculation, the instantaneous value of the circulating current between any two modules at each sampling time can be obtained.

[0018] Furthermore, the root mean square (RMS) value of the instantaneous circulating current values ​​within a preset time window is calculated as the circulating current amplitude between module i and module j. The instantaneous circulating current values ​​fluctuate rapidly over time, and a single instantaneous value cannot fully reflect the severity of the circulating current. The RMS value is the square root of the average of the squared instantaneous circulating current values ​​within the time window, effectively characterizing the effective energy level of the AC circulating current within that time window. The length of the preset time window can be set according to the rate of change of the circulating current, for example, it can be set to 0.01 seconds to 0.1 seconds, covering several switching cycles. The formula for calculating the RMS value is: the circulating current amplitude equals the square root of the sum of the squared instantaneous circulating current values ​​within the time window divided by the number of sampling points. The larger the circulating current amplitude, the more severe the circulating current between module i and module j.

[0019] Finally, iterate through all module pairs and mark the module pair with the largest circulating current amplitude as the dominant circulating current source. Assume there are N inverter modules in the system, then there are N×(N-1) / 2 module pairs. Calculate the circulating current amplitude for each module pair and select the module pair with the largest circulating current amplitude. This module pair is the one with the most severe circulating current in the system and is marked as the dominant circulating current source.

[0020] Circulating currents typically occur between specific module pairs. Identifying the dominant circulating current source allows control resources to be focused on the module pairs that most require intervention, avoiding the inefficiency and additional disturbances caused by uniform adjustments to all modules. Through the above steps, the source of the circulating current can be accurately located, providing a clear target for subsequent differentiated compensation control.

[0021] S20: Extract the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and calculate the deviation between the circulating current value of the dominant circulating current source and the average circulating current value of the system as the circulating current deviation. Secondly, the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source are extracted. The dominant circulating current source is the module pair with the most severe circulating current in the system, and its circulating current mainly stems from the difference in output characteristics between the two modules. Voltage deviation refers to the difference in output voltage amplitude between the two modules in the dominant circulating current source, reflecting the degree of voltage amplitude mismatch between the modules. The greater the voltage amplitude difference, the greater the reactive component in the circulating current. Phase deviation refers to the difference in phase angle of the output voltage between the two modules, reflecting the degree of phase inconsistency between the modules. The greater the phase difference, the greater the active component in the circulating current. Carrier phase difference refers to the difference in the phase angle of the pulse width modulation carrier between the two modules. Asynchronous carrier phases will lead to high-frequency circulating current components. Extracting these three deviations can comprehensively quantify the degree of difference between the two modules in the dominant circulating current source, providing a basis for the calculation of subsequent control quantities.

[0022] Furthermore, the deviation between the circulation value of the dominant circulation source and the system circulation mean is calculated as the circulation deviation degree. This circulation deviation degree reflects the degree to which the severity of the circulation from the dominant circulation source deviates from the system average level. The larger the deviation degree, the more prominent the circulation problem of the dominant circulation source, requiring priority intervention. Through the circulation deviation degree, the circulation suppression problem can be transformed into a problem of adjusting the deviation degree; that is, by using compensatory control, the circulation value of the dominant circulation source can be made closer to the system circulation mean, thereby reducing the overall circulation level of the system.

[0023] Specifically, the starting point of this step is that the goal of circulating current suppression is not to reduce all circulating currents to zero, but to control the circulating currents to a level comparable to the system average, avoiding the harm caused by excessive local circulating currents. By extracting voltage deviation, phase deviation, carrier phase difference, and circulating current deviation, quantitative state inputs can be provided for subsequent judgment of circulating current evolution trends and model predictive control.

[0024] Specifically, the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source are extracted, and the deviation between the circulating current value of the dominant circulating current source and the average circulating current of the system is calculated as the circulating current deviation, including: Calculate the difference in output voltage amplitude between the two modules in the dominant circulating current source as the voltage deviation; Calculate the difference in the output voltage phase angle between the two modules in the dominant circulating current source, and use it as the phase deviation; Calculate the difference in carrier phase angle between the two modules in the dominant circulating current source, and use it as the carrier phase difference; Calculate the arithmetic mean of the circulating current values ​​for all modules, and use it as the system circulating current mean. The difference between the circulation value of the dominant circulation source and the average circulation value of the system is calculated as the circulation deviation.

[0025] First, the difference in output voltage amplitude between the two modules in the dominant circulating current source is calculated as the voltage deviation. Specifically, let the two modules in the dominant circulating current source be module A and module B, and their output voltage amplitudes be U and U, respectively. A and U B Then the voltage deviation ΔU equals U A Subtract U B The absolute value of the voltage deviation reflects the degree of mismatch between the output voltage amplitudes of the two modules. The greater the amplitude difference, the greater the reactive circulating current component generated between the modules. The sign of the voltage deviation indicates the direction of the amplitude deviation, i.e., which module has a higher voltage amplitude.

[0026] Next, the difference in output voltage phase angles between the two modules in the dominant circulating current source is calculated as the phase deviation. Let the output voltage phase angles of module A and module B be θ. A and θ B Then the phase deviation Δθ equals θ A Subtract θ B The absolute value of the phase deviation reflects the degree of inconsistency between the phases of the output voltages of the two modules. The greater the phase difference, the greater the active circulating current component generated between the modules. The sign of the phase deviation indicates whether the phase is leading or lagging.

[0027] Then, the difference in carrier phase angle between the two modules in the dominant circulating current source is calculated as the carrier phase difference. The carrier phase refers to the initial phase angle of the triangular carrier in the pulse width modulation signal of each inverter module. Let the carrier phase angles of module A and module B be φ... A and φ B Then the carrier phase difference Δφ is equal to φ A Subtract φ B The absolute value of the carrier phase difference is the main cause of high-frequency circulating current. When the carrier phases are out of sync, high-frequency switching frequency components will circulate between modules, increasing harmonic losses.

[0028] Further, the arithmetic mean of the circulating current values ​​of all module pairs is calculated as the system circulating current mean. Specifically, assuming there are N inverter modules in the system, the total number of module pairs is N multiplied by N minus 1 and then divided by 2. The circulating current amplitude of each module pair is calculated, and the sum of the circulating current amplitudes of all module pairs is divided by the total number of module pairs to obtain the system circulating current mean. This system circulating current mean reflects the average circulating current level of the entire parallel inverter system and can be used as a benchmark for judging the severity of the dominant circulating current source.

[0029] Finally, the difference between the circulation value of the dominant circulation source and the system circulation mean is calculated as the circulation deviation. The circulation deviation reflects the degree to which the severity of the dominant circulation source deviates from the system average level. When the circulation deviation is positive, it indicates that the circulation of the dominant circulation source is higher than the system average level and requires compensatory control to suppress it; when the circulation deviation is negative, it indicates that the circulation of the dominant circulation source is lower than the system average level and does not require priority treatment.

[0030] By using the circulation deviation, the circulation suppression problem can be quantified as an adjustment target for the deviation. That is, by using compensation control, the circulation deviation can be made to approach zero, thereby causing the circulation value of the dominant circulation source to fall back to the system average level.

[0031] S30: Based on the rate of change of the circulation deviation and the historical circulation trajectory, determine the circulation evolution trend of the dominant circulation source, wherein the circulation evolution trend includes natural convergence and continuous deterioration; Furthermore, based on the rate of change of circulation deviation and historical circulation trajectories, the circulation evolution trend of the dominant circulation source is determined. The circulation evolution trend refers to the direction of development of the circulation deviation of the dominant circulation source within a future time window, including two states: natural convergence and continuous deterioration.

[0032] Natural convergence refers to the circulation deviation showing an automatic decreasing trend, indicating that the system itself has a certain circulation suppression capability and the circulation problem is mitigating on its own without the need for external control intervention. Continuous deterioration refers to the circulation deviation showing a continuous increasing trend, indicating that the circulation problem is intensifying and, if not intervened in time, may evolve into a serious circulation failure, threatening the safe operation of the system.

[0033] The rate of change of circulation deviation reflects how fast and in what direction the circulation deviation changes over time. A positive rate of change indicates that the circulation deviation is increasing and the circulation is deteriorating; a negative rate of change indicates that the circulation deviation is decreasing and the circulation is converging. Historical circulation trajectories refer to the changes in circulation deviation over several past sampling times, reflecting the trend and acceleration characteristics of the circulation deviation. Relying solely on the current circulation deviation value is insufficient to predict its future direction. For example, if the circulation deviation is large but in a rapid decreasing phase, the circulation may converge naturally without intervention; conversely, if the circulation deviation is small but in a rapid increasing phase, the circulation may continue to deteriorate. Therefore, a comprehensive judgment requires considering both the rate of change and historical trajectories.

[0034] Specifically, the purpose of this step in determining the circulation evolution trend is to distinguish between scenarios requiring intervention and those that do not. When the circulation converges naturally, applying control is not only unnecessary but may also introduce new disturbances, causing system instability. When the circulation continues to worsen, compensatory control must be applied promptly to prevent further expansion of the circulation. By determining the circulation evolution trend, on-demand control can be achieved, unnecessary adjustments can be avoided, and the efficiency and stability of the control system can be improved.

[0035] Specifically, based on the rate of change of the circulation deviation and the historical circulation trajectory, the circulation evolution trend of the dominant circulation source is determined, including: Collect the circulation deviation of the N sampling times preceding the current time to form a deviation time series; Calculate the first difference of the deviation time series as the deviation rate of change series; Calculate the first difference of the deviation rate change sequence as the deviation change acceleration; When the rate of change of deviation is positive and the acceleration of change of deviation is positive, it is judged as a continuous deterioration trend; when the rate of change of deviation is negative and the acceleration of change of deviation is negative, it is judged as a natural convergence trend.

[0036] First, the circulation deviation is collected for the N sampling times preceding the current time, forming a deviation time series. Specifically, let the current sampling time be k, and the circulation deviation values ​​are collected for N sampling times from time k-N+1 to time k, forming a deviation time series ΔI. k-N+1 ,ΔI k-N+2 ,...,ΔI k The value of N can be set according to the dynamic characteristics of the circulation, for example, it can be set to 5 to 10, which can reflect the changing trend of the circulation without introducing too much historical noise. This deviation time series constitutes the historical trajectory data for analyzing the evolution trend of the circulation.

[0037] Secondly, the first-order difference of the deviation time series is calculated as the deviation rate sequence. The first-order difference refers to the difference between the circulation deviations at two adjacent sampling times, reflecting the speed and direction of the circulation deviation change over time. Specifically, for time t, the deviation rate v t Equal to the current circulation deviation ΔI t Subtract the circulation deviation ΔI from the previous moment t-1 When the rate of change is positive, it indicates that the circulation deviation is increasing and the circulation is deteriorating; when the rate of change is negative, it indicates that the circulation deviation is decreasing and the circulation is converging. By calculating the rate of change of deviation sequentially for all adjacent time points, the sequence of deviation rates can be obtained.

[0038] Furthermore, the first difference of the deviation rate change sequence is calculated as the deviation change acceleration. The deviation change acceleration refers to the rate of change of the rate of change, reflecting the acceleration characteristic of the circulation deviation change trend. Specifically, for time t, the deviation change acceleration a... t Equal to the rate of change v at the current moment t Subtract the rate of change v from the previous moment t-1 When the acceleration is positive, it indicates that the rate of change of the circulation deviation is accelerating, that is, the rate of circulation deterioration or convergence is increasing; when the acceleration is negative, it indicates that the rate of change is slowing down.

[0039] Specifically, when both the rate of change and the acceleration of the deviation are positive, it is determined to be a continuously deteriorating trend. A positive rate of change indicates that the circulation deviation is increasing, and the circulation is tending to worsen; a positive acceleration indicates that the rate of increase in circulation deviation is accelerating, that is, the deterioration trend is accelerating. The combination of these two factors indicates that the circulation problem is continuously and rapidly worsening. If not intervened in time, it will evolve into a serious circulation failure. Therefore, it is determined to be a continuously deteriorating trend, and compensatory control measures are required.

[0040] Furthermore, when both the rate of change and the acceleration of the deviation are negative, it is determined to be a natural convergence trend. A negative rate of change indicates that the circulation deviation is decreasing, and the circulation is tending to converge; a negative acceleration indicates that the rate of decrease in circulation deviation is accelerating, that is, the convergence trend is accelerating. The combination of these two factors indicates that the circulation is rapidly and naturally converging, and the system's own regulatory effect has gradually alleviated the circulation problem, without the need for external control intervention; therefore, it is determined to be a natural convergence trend.

[0041] In summary, the dual judgment mechanism based on the rate of change and acceleration of change can accurately identify the evolution trend of the circulation, avoid applying unnecessary control actions when the circulation naturally converges, and prevent the impact of control disturbances on system stability; at the same time, it ensures that compensation control is triggered in a timely manner when the circulation continues to deteriorate, preventing the circulation from expanding further.

[0042] S40: When the circulating current evolution trend is continuously deteriorating, the circulating current deviation and the circulating current evolution trend are used as constraints. Combined with the physical constraint that the sum of the output currents of each module is equal to the load current, the evolution trajectory of the future circulating current deviation is predicted by the model prediction controller. With the goal of minimizing the peak value of the future circulating current deviation, the voltage compensation amount and carrier phase adjustment amount of the two modules in the dominant circulating current source are solved. When the circulating current evolution trend is determined to be continuously deteriorating, it indicates that the circulating current problem of the dominant circulating current source is intensifying. If not intervened in time, it will evolve into a serious fault, requiring proactive compensation control. At this point, with the circulating current deviation and the circulating current evolution trend as constraints, combined with the physical constraint that the sum of the output currents of each module equals the load current, the model predictive controller predicts the future evolution trajectory of the circulating current deviation. With the goal of minimizing the peak value of the future circulating current deviation, the voltage compensation and carrier phase adjustment of the two modules in the dominant circulating current source are solved.

[0043] Among them, the model predictive controller is a model-based feedforward control method. Its core idea is to use a predictive model to predict the trajectory of the system state change over a period of time in the future at the current moment, and then obtain the optimal control sequence through optimization and apply the first control quantity to the system.

[0044] Specifically, the inputs to the model predictive controller include state variables such as the current circulation deviation, circulation evolution trend, voltage deviation, phase deviation, and carrier phase difference, as well as candidate voltage compensation and carrier phase adjustment values. Based on the current state and candidate compensation values, the predictive model can recursively calculate the predicted circulation deviation values ​​for each time in the future preset time domain, forming the circulation deviation evolution trajectory.

[0045] Two important constraints must be met during the prediction process. The first is the constraint on the circulating current deviation and its evolution trend; the predicted circulating current deviation trajectory should match the currently observed evolution trend to ensure the prediction's rationality. The second is the physical constraint that the sum of the output currents of all modules equals the load current. Regardless of the compensation control applied, the sum of the output currents of all inverter modules must equal the load current. This constraint is a fundamental physical law of parallel inverter systems, used to ensure that compensation control does not disrupt the system's power balance.

[0046] Furthermore, the optimization objective is to minimize the peak value of the future circulation deviation. The peak value of the circulation deviation reflects the maximum severity that the circulation may reach over a future period. By minimizing this peak value, it is ensured that the circulation is effectively suppressed throughout the entire prediction time domain, avoiding excessive circulation in local periods. The optimization solver is used to select the optimal set from candidate voltage compensation and carrier phase adjustment values, while satisfying the above constraints, so as to minimize the peak value of the corresponding circulation deviation evolution trajectory.

[0047] Specifically, the voltage compensation amount is used to adjust the output voltage amplitude and phase of the two modules in the dominant circulating current source to reduce voltage and phase deviations; the carrier phase adjustment amount is used to adjust the carrier phase of the two modules to reduce the carrier phase difference. Through the coordinated adjustment of these two compensation amounts, the cause of circulating current can be eliminated at its source.

[0048] The prediction model is pre-built through the following steps: Collect historical operating data of the parallel inverter system under continuously deteriorating operating conditions. The historical operating data includes the output current, output voltage, carrier phase, circulating current value of each module, as well as the applied voltage compensation amount and carrier phase adjustment amount. The training sample set is constructed by taking the current state and the applied compensation amount in the historical operation data as input and the actual circulation value of the next moment in the historical operation data as output. A prediction model is constructed and trained using a supervised learning algorithm. The training objective is to minimize the mean square error between the predicted circulation value and the actual circulation value until convergence is achieved, thus obtaining the trained prediction model.

[0049] First, historical operating data of the parallel inverter system under continuously deteriorating operating conditions were collected. Continuously deteriorating operating conditions refer to an operating state where the circulating current deviation continuously increases and the circulating current problem gradually intensifies. Data under this condition can cover the complete evolution process of circulating current from mild to severe, enabling the predictive model to learn the dynamic laws of circulating current deterioration. Historical operating data includes the output current, output voltage, carrier phase, circulating current value, and applied voltage compensation and carrier phase adjustment for each module. Output current and output voltage reflect the real-time power output status of the module, carrier phase reflects the phase position of the pulse width modulation signal, circulating current value is the target output of the predictive model, and voltage compensation and carrier phase adjustment are the control input variables of the model. All data were recorded in chronological order, with consistent sampling time intervals.

[0050] Secondly, a training sample set is constructed using the current state and applied compensation amount from historical operating data as input, and the actual circulating current value at the next time step from historical operating data as output. For each sampling time t, the current state variables, including the output current, output voltage, carrier phase, and circulating current value of each module, as well as the voltage compensation amount and carrier phase adjustment amount applied at that time, are combined as the input feature vector. The actual circulating current value at the next time step t+1 is used as the output label. Multiple input-output pairs are extracted from historical operating data using a sliding time window, and each input-output pair constitutes a training sample. The set of all samples constitutes the training sample set.

[0051] The sliding time window refers to a method of moving point by point along the time axis to extract continuous data segments. It represents a data processing method that extracts input-output pairs of historical running data in sequence at fixed step sizes. The step size of the time window is set according to the sampling time interval. For example, when the sampling period is 0.1 milliseconds, the sliding step size can be set to 1 sampling point, so that two adjacent training samples are closely connected in time and fully cover the continuous process of circulation evolution.

[0052] Furthermore, a prediction model is constructed and trained using a supervised learning algorithm. The training objective is to minimize the mean squared error between the predicted and actual circulation values ​​until convergence, resulting in a trained prediction model. Optionally, the prediction model can employ architectures such as Long Short-Term Memory networks, temporal convolutional networks, or fully connected neural networks. Its input layer has the same number of nodes as the dimension of the input feature vector, and its output layer has only one node, outputting the predicted circulation value for the next time step.

[0053] For example, taking a Long Short-Term Memory (LSTM) network as an example, the specific architecture and parameters of the prediction model are as follows. The number of nodes in the input layer is equal to the dimension of the input feature vector. The input feature vector includes the output current, output voltage, carrier phase, circulating current value, and applied voltage compensation and carrier phase adjustment for each module. Assuming there are N inverter modules in the system, the output current and output voltage each contain N values, the carrier phase contains N values, and the circulating current value contains N×(N-1) / 2 module pairs of circulating current values, plus 2 values ​​each for voltage compensation and carrier phase adjustment. The number of nodes in the input layer is determined based on the actual number of modules. For example, when N equals 3, the number of nodes in the input layer is approximately 3+3+3+3+2+2, which equals 16. The number of hidden units in both LSM layers is set to 64. Each LSM layer is followed by a Dropout layer with a dropout rate of 0.2 to suppress overfitting. The output layer is a fully connected layer with 1 node, outputting the predicted circulating current value for the next time step. A linear activation function is used.

[0054] During training, key hyperparameters included a learning rate of 0.001, 100 training epochs, and a batch size of 32. Mean squared error was used as the loss function, and Adam was employed as the optimizer. The training sample set was divided into training and validation sets in an 8:2 ratio. After each training epoch, the model performance was evaluated using the validation set. Training was stopped when the loss function value on the validation set no longer decreased for 10 consecutive epochs. The model parameters at the minimum loss on the validation set were saved, resulting in the trained prediction model. This prediction model can predict the circulation value at the next moment based on the current system state and the applied compensation, providing a basis for the model predictive controller to recursively predict the future circulation evolution trajectory.

[0055] Specifically, when the circulating current evolution trend is continuously deteriorating, the evolution trajectory of the future circulating current deviation is predicted by a model predictive controller, constrained by the circulating current deviation and the circulating current evolution trend. With the objective of minimizing the peak value of the future circulating current deviation, the voltage compensation and carrier phase adjustment of the two modules in the dominant circulating current source are solved, including: Collect the circulation deviation of the M sampling times before the current time to form a deviation time series, and calculate the variance of the deviation time series as the circulation fluctuation intensity; Calculate the correlation coefficients between each pair of voltage deviation, phase deviation, and carrier phase difference, and use them as the module differential coupling degree; The prediction model inside the model prediction controller uses the current circulation deviation and circulation evolution trend as input to predict the initial circulation deviation evolution trajectory corresponding to different candidate compensation amounts applied in the future preset time domain. Using the circulation fluctuation intensity as a fluctuation correction factor, the predicted value at each moment in the initial circulation deviation evolution trajectory is multiplied by the fluctuation correction factor to obtain the corrected circulation deviation evolution trajectory. Using the module difference coupling degree as a coupling correction factor, the peak value of the circulation deviation in the optimization objective of the optimization solver inside the model prediction controller is multiplied by the coupling correction factor to obtain the corrected optimization objective; By optimizing the solver and minimizing the modified optimization objective, the optimal voltage compensation and carrier phase adjustment are selected from the candidate compensation values.

[0056] First, the circulation deviation is collected for the M sampling times preceding the current moment, forming a deviation time series. The variance of the deviation time series is then calculated as the circulation fluctuation intensity. The circulation fluctuation intensity reflects the severity of the fluctuation in circulation deviation within the recent time window. A larger variance indicates more severe fluctuations in circulation deviation, a more unstable system operation, and higher uncertainty in the prediction model; a smaller variance indicates smoother fluctuations in circulation deviation, a more stable system operation, and higher reliability of the prediction model. The value of M can be set according to the dynamic characteristics of the circulation, for example, it can be set to 10 to 20.

[0057] Secondly, the correlation coefficients between each pair of voltage deviation, phase deviation, and carrier phase difference are calculated as the module difference coupling degree. Voltage deviation, phase deviation, and carrier phase difference are three difference dimensions between two modules in the dominant circulating current source, and there may be coupling relationships among them. For example, voltage deviation and phase deviation may increase or decrease simultaneously. The correlation coefficient is used to quantify the degree of linear correlation between two variables, and its value ranges from -1 to 1.

[0058] Calculate the correlation coefficients between voltage deviation and phase deviation, voltage deviation and carrier phase difference, and phase deviation and carrier phase difference. Take the average or maximum absolute value of the three correlation coefficients as the module differential coupling degree. The higher the module differential coupling degree, the stronger the correlation between the various deviation quantities, and the more comprehensive the coordinated adjustment of multiple deviation quantities needs to be considered during control. The lower the module differential coupling degree, the more independent the various deviation quantities are, and they can be adjusted separately.

[0059] Specifically, the calculation steps for the correlation coefficient are as follows: First, the voltage deviation sequence, phase deviation sequence, and carrier phase difference sequence of the two modules in the dominant circulating current source are collected at T consecutive sampling times within a preset time window. Let the time window contain T sampling points, and the voltage deviation sequence be ΔU1, ΔU2, ..., ΔU... T The phase deviation sequence is Δθ1, Δθ2, ..., Δθ T The carrier phase difference sequence is Δφ1, Δφ2, ..., Δφ T The value of T can be set according to the dynamic characteristics of the circulation, for example, it can be set to 50 to 100.

[0060] Secondly, calculate the correlation coefficient ρ between voltage deviation and phase deviation. Uθ The calculation formula is: ρ Uθ It equals the covariance of voltage deviation and phase deviation divided by the product of the standard deviation of voltage deviation and the standard deviation of phase deviation. The covariance reflects the consistency of the trends of the two variables, while the standard deviation reflects their respective dispersion relative to the mean. When ρ Uθ When ρ is positive, it indicates that the voltage deviation and phase deviation change in the same direction, either increasing or decreasing simultaneously; when ρ is positive... Uθ When |ρ is negative, it indicates that the two change in opposite directions; Uθ The closer the correlation is to 1, the higher the degree of linear correlation between the two; the closer the correlation is to 0, the less linear correlation there is between the two.

[0061] Similarly, calculate the correlation coefficient ρ between voltage deviation and carrier phase difference. Uφ And the correlation coefficient ρ between phase deviation and carrier phase difference θφ The calculation formula is the same as above, only the corresponding variable sequence is replaced.

[0062] Finally, the absolute values ​​of the three correlation coefficients are averaged, i.e., the module difference coupling degree is equal to (|ρ Uθ |+|ρ Uφ |+|ρ θφ|) / 3. Alternatively, the maximum absolute value of the three correlation coefficients can be taken as the module difference coupling degree, that is, the module difference coupling degree equals max|ρUθ|,|ρUφ|,|ρθφ|. The two methods can be selected according to actual needs. Taking the average value can comprehensively reflect the overall coupling degree of the three dimensions, while taking the maximum value can highlight the impact of the strongest coupling relationship on control. For example, if the correlation coefficient between voltage deviation and phase deviation is 0.85, the correlation coefficient between voltage deviation and carrier phase difference is 0.12, and the correlation coefficient between phase deviation and carrier phase difference is 0.08, then the average value is (0.85+0.12+0.08) / 3≈0.35, and the maximum value is 0.85. When the module difference coupling degree is high, it indicates that there is a strong synergistic relationship between the various deviations. During control, multiple deviations need to be adjusted simultaneously to avoid the deterioration of other deviations caused by a single adjustment.

[0063] Furthermore, using the prediction model within the model predictive controller, and taking the current circulating current deviation and its evolution trend as input, the model predicts the initial circulating current deviation evolution trajectory corresponding to different candidate compensation amounts applied within a preset time domain. Specifically, the prediction model can recursively calculate the predicted circulating current deviation values ​​for multiple future time steps based on the current state and the applied control variables. For each candidate combination of voltage compensation and carrier phase adjustment, the prediction model outputs a corresponding initial circulating current deviation evolution trajectory, where each point represents the predicted circulating current deviation value at the corresponding time.

[0064] Specifically, taking into account the physical constraint that the sum of the output currents of all modules equals the load current, the model predictive controller predicts the evolution trajectory of the future circulating current deviation, including: Obtain real-time measurements of the load current; Add a summation constraint term to the prediction model to ensure that the sum of the predicted output currents of each module deviates from the load current within a preset range. Based on the prediction model with summation constraints, the predicted value of the initial circulation deviation at each time point within the preset future time domain is calculated recursively.

[0065] First, obtain the real-time measurement value of the load current. The load current refers to the actual current value of the load connected to the parallel inverter system, which can be obtained in real time by installing a current sensor at the common load terminal. The real-time measurement value of the load current reflects the total output current demand of the system at the current moment and is the physical constraint target of the sum of the output currents of all parallel inverter modules.

[0066] Secondly, a summation constraint term is added to the prediction model to ensure that the deviation between the sum of the predicted output currents of each module and the load current does not exceed a preset range. When the prediction model recursively calculates the circulating current predictions for future times, it also predicts the output currents of each module. The summation constraint term is a penalty term added to the loss function or constraint conditions of the prediction model. When the deviation between the sum of the predicted output currents of each module and the load current exceeds a preset range, a penalty gradient is generated, guiding the prediction results to meet physical constraints. The preset range can be set according to the system accuracy requirements, for example, it can be set to ±1% of the load current. By adding the summation constraint term, it is ensured that the sum of the module output currents corresponding to the circulating current predictions output by the prediction model conforms to the basic physical laws of parallel inverter systems, that is, the sum of the output currents of all modules must equal the load current.

[0067] Optionally, the summation constraint term can be implemented by adding a penalty term to the loss function of the prediction model. This penalty term is equal to the squared deviation between the sum of the module output currents and the load current multiplied by a penalty coefficient. Let the sum of the predicted output currents of each module be I. sum The measured load current is I. load Then the total constraint penalty term is α×(I sum -I load ) 2 , where α is the penalty coefficient, which can be set to 100. When I sum with I load When the deviation exceeds a preset range, such as ±1% of the load current, the penalty term generates a large loss value, guiding the model to prioritize satisfying the summation constraint during training and prediction. During recursive prediction, for each future time step, the predicted circulating current deviation is calculated, and the sum of the output currents of each module at that time is also calculated. The summation constraint is checked; if not, the predicted circulating current value is fine-tuned based on the deviation direction.

[0068] Finally, based on the prediction model with summation constraints, the initial circulating current deviation prediction values ​​for each time point within the preset future time domain are recursively calculated. During the recursive process, the prediction model with summation constraints checks whether the sum of the output currents of the corresponding modules satisfies the summation constraint for each future time point's circulating current prediction value. If not, the prediction value is corrected or a penalty is added. The recursive calculation refers to starting from the current time point and using the prediction model to progressively calculate the circulating current deviation prediction values ​​for the next time point, the next two time points, and up to the end point of the preset future time domain.

[0069] For example, if the current time is k, and the prediction time domain is 10 time steps, then the predicted circulating current deviation values ​​at times k+1, k+2, ..., k+10 are calculated sequentially, forming the initial circulating current deviation evolution trajectory. This initial circulating current deviation evolution trajectory reflects the possible future development trend of the circulating current deviation under the physical constraint that the sum of the module output currents equals the load current, providing a predictive basis that conforms to physical laws for subsequent optimization solutions. By introducing a summation constraint term, invalid predictions caused by violations of physical laws can be avoided, improving the reliability and accuracy of model predictive control.

[0070] Secondly, the intensity of circulation fluctuations is used as a fluctuation correction factor. The predicted values ​​at each time point in the initial circulation deviation evolution trajectory are multiplied by the fluctuation correction factor to obtain the corrected circulation deviation evolution trajectory.

[0071] The fluctuation correction factor is calculated as follows: the fluctuation correction factor equals 1 plus the circulation fluctuation intensity. The circulation fluctuation intensity is the variance of the deviation time series, and its value is greater than or equal to 0. When the circulation fluctuation intensity is 0, the fluctuation correction factor equals 1, and no correction is made to the predicted trajectory; when the circulation fluctuation intensity is greater than 0, the fluctuation correction factor is greater than 1, and the predicted circulation deviation values ​​at each time point of the predicted trajectory are proportionally amplified. The amplified predicted trajectory reflects the more severe degree of circulation deviation that may be reached when the system uncertainty is high, guiding the optimization solver to adopt a more conservative control strategy, that is, to apply a stronger compensation amount to ensure that the circulation can still be effectively suppressed in the worst case.

[0072] Specifically, the intensity of circulation fluctuations reflects the degree of uncertainty in the system. Greater fluctuation intensity results in lower confidence levels for the prediction model, necessitating amplification and correction of the predicted values ​​to ensure sufficient safety margin in the control strategy. Conversely, smaller fluctuation intensity leads to higher confidence levels for the prediction model, requiring a correspondingly smaller correction magnitude. This step, by multiplying by a fluctuation correction factor, ensures that the corrected predicted trajectory reflects the impact of system uncertainty on circulation evolution, preventing under-control due to overly optimistic predictions.

[0073] Furthermore, using the module difference coupling degree as a coupling correction factor, the peak value of the circulating current deviation in the optimization objective of the optimization solver within the model predictive controller is multiplied by the coupling correction factor to obtain the corrected optimization objective. The original optimization objective is to minimize the future peak value of the circulating current deviation. When the module difference coupling degree is high, it indicates a strong coupling relationship between voltage deviation, phase deviation, and carrier phase difference. Compensation in a single dimension may cause deterioration in other dimensions. Therefore, a higher penalty weight needs to be applied to the peak value of the circulating current deviation to guide the optimization solver to choose a more conservative control strategy. When the module difference coupling degree is low, the deviations are relatively independent, and the penalty weight can be appropriately reduced. This step, by multiplying by the coupling correction factor, allows the optimization objective to adaptively adjust according to the degree of coupling of module differences, improving the rationality of the optimization solution.

[0074] Furthermore, by optimizing the solver and using the minimization of the modified optimization objective as the criterion, the optimal voltage compensation amount and carrier phase adjustment amount are selected from the candidate compensation amounts.

[0075] The range and step size of the candidate compensation values ​​are set according to the system's rated parameters and control accuracy requirements. The voltage compensation value ranges from -0.1 times the rated voltage to +0.1 times the rated voltage, with a step size set to 0.01 times the rated voltage. For example, when the rated voltage is 400 volts, the voltage compensation value ranges from -40 volts to +40 volts, with a step size of 4 volts. The carrier phase adjustment value ranges from -30 degrees to +30 degrees, with a step size set to 2 degrees. Within these ranges, the voltage compensation value and carrier phase adjustment value are discretized according to the step size, forming a candidate compensation value combination grid. The optimization solver traverses all candidate combinations in this grid, calculates the corrected optimization target value corresponding to each combination, and selects the optimal solution from them.

[0076] The optimization solver iterates through all candidate combinations of voltage compensation and carrier phase adjustment, calculates the corrected circulating current deviation evolution trajectory for each candidate combination, calculates the corrected optimization objective value, and selects the candidate combination that minimizes the optimization objective as the optimal solution output.

[0077] Specifically, the optimization solver uses a quadratic programming algorithm to solve for the optimal control sequence that minimizes the cost function, including: Construct a cost function, which includes the sum of squares of the predicted circulating current deviation at each time point in the future preset time domain, and a penalty term for the rate of change of voltage compensation and carrier phase adjustment. The optimal control sequence that minimizes the cost function is solved using a quadratic programming algorithm. The initial voltage compensation amount and the initial carrier phase adjustment amount at the current moment are extracted from the optimal control sequence as candidate compensation amounts.

[0078] First, a cost function is constructed. This cost function includes the sum of squares of the corrected predicted circulation current deviation values ​​at each time point within a preset future time domain, and a penalty term for the rate of change of voltage compensation and carrier phase adjustment. The preset future time domain refers to the prediction range over several time steps from the current time, for example, 10 time steps. The sum of squares term is the sum of the squares of the corrected predicted circulation current deviation values ​​at each time point within the time domain. This term reflects the circulation current suppression effect; a smaller sum of squares indicates a lower predicted circulation current deviation and a better suppression effect.

[0079] The rate-of-change penalty term is the sum of the squares of the rates of change of voltage compensation and carrier phase adjustment, where the rate of change refers to the difference between the compensation at the current moment and the compensation at the previous moment. The rate-of-change penalty term is used to suppress drastic fluctuations in the compensation amount, preventing shocks to the system caused by sudden changes in the compensation amount. The two terms are balanced by weighting coefficients; for example, the weighting coefficient for the circulating current deviation sum of squares term can be set to 1, and the weighting coefficient for the rate-of-change penalty term can be set to 0.1.

[0080] Secondly, a quadratic programming algorithm is used to find the optimal control sequence that minimizes the cost function. Quadratic programming is a mathematical method for solving convex quadratic optimization problems. Its standard form is that the objective function is quadratic and the constraints are linear equality or inequality. Specifically, the sum of squares and rate of change penalty terms in the cost function can both be expressed in quadratic form, the range of voltage compensation and carrier phase adjustment can be expressed as linear inequality constraints, and the physical constraint that the sum of the output currents of all modules equals the load current can be expressed as linear equality constraints. Therefore, this optimization problem can be transformed into a quadratic programming problem, which can be solved by calling a quadratic programming solver. The quadratic programming algorithm can converge to the global optimum in polynomial time, has high computational efficiency, and is suitable for online control applications. The solution is an optimal control sequence that includes the voltage compensation and carrier phase adjustment at each time point within a preset future time domain.

[0081] Furthermore, the initial voltage compensation and initial carrier phase adjustment at the current moment are extracted from the optimal control sequence as candidate compensation values. Model predictive control employs a rolling optimization strategy, which means that at each sampling moment, the optimal control sequence for the future time domain is solved, but only the first control variable in the sequence is applied to the system; at the next moment, sampling is repeated and the solution is recalculated. Therefore, the voltage compensation and carrier phase adjustment corresponding to the current moment are extracted from the optimal control sequence as initial candidate compensation values. These candidate compensation values ​​will subsequently undergo coupling correction and other processing before being finally applied to the two modules in the dominant circulating current source.

[0082] Through the above steps, the optimized solver can quickly solve for the control quantity that achieves the optimal circulation suppression effect while satisfying the system's physical constraints, and at the same time avoids drastic fluctuations in the compensation quantity, ensuring the smoothness of the control process. Furthermore, the model predictive controller considers the intensity of circulation fluctuations and the degree of module differential coupling when solving for the compensation quantity, enabling the solution to adapt to the dynamic and coupling characteristics of the system, thus improving the robustness and adaptability of circulation suppression.

[0083] Specifically, the circulating current suppression problem is essentially a dynamic optimization problem, requiring the selection of the optimal control variable to achieve rapid convergence of the circulating current while satisfying the system's physical constraints. Traditional PI control only adjusts based on the current error, resulting in a delayed response and difficulty in predicting future trends. Model predictive control, on the other hand, can predict future evolution trajectories and apply control in advance, achieving proactive suppression and significantly improving the dynamic performance of circulating current suppression. This step achieves differentiated compensation for the dominant circulating current source by solving for the voltage compensation and carrier phase adjustment, i.e., applying control only to the two modules with the most severe circulating current, avoiding the disturbances and resource waste caused by uniform adjustment of all modules in traditional methods.

[0084] S50: Based on the voltage compensation amount and carrier phase adjustment amount, perform differentiated compensation control on the two modules in the dominant circulating current source.

[0085] Finally, based on the voltage compensation and carrier phase adjustment obtained from the model predictive controller, differentiated compensation control is applied to the two modules in the dominant circulating current source. Differentiated compensation control means applying control only to the module with the most severe circulating current, without adjusting other modules, thereby achieving precise intervention and avoiding the additional disturbances and wasted computational resources caused by the traditional method of uniformly adjusting all modules.

[0086] During the execution of compensation control, the amplitude of the dominant circulating current source must be continuously monitored. When the circulating current amplitude drops below the preset acceptable threshold, it indicates that the circulating current has been effectively suppressed, and compensation control should be stopped to avoid over-adjustment. The preset acceptable threshold can be set according to the system's allowable range for circulating current, for example, it can be set to 2% of the rated current. If the circulating current amplitude rises again above the threshold, the model predictive controller is re-triggered to perform a new round of compensation calculation and control.

[0087] The circulating current distribution in a parallel inverter system is non-uniform. The module pair with the largest circulating current amplitude is usually the source of the circulating current. Prioritizing the treatment of this source can effectively suppress the overall circulating current of the system without adjusting all modules. Through the above-mentioned differentiated compensation control, voltage compensation and carrier phase adjustment are applied only to the module pair with the most severe circulating current, while the original control of other modules remains unchanged. This achieves precise suppression of circulating current, avoids unnecessary system disturbances, reduces the complexity and computational burden of the control system, and improves the efficiency of circulating current suppression and dynamic response speed.

[0088] Specifically, based on the voltage compensation amount and carrier phase adjustment amount, differentiated compensation control is performed on the two modules in the dominant circulating current source, including: The voltage compensation is superimposed onto the voltage reference values ​​of the two modules in the dominant circulating current source; The carrier phase adjustment is superimposed on the carrier phases of the two modules in the dominant circulating current source, respectively; During the execution of compensation control, the circulation amplitude of the dominant circulation source is continuously monitored. When the circulation amplitude drops below the preset qualified threshold, compensation control is stopped.

[0089] First, the voltage compensation is superimposed onto the voltage reference values ​​of the two modules in the dominant circulating current source. Specifically, let the two modules in the dominant circulating current source be module A and module B, and their original voltage reference values ​​are the first voltage reference value and the second voltage reference value, respectively. The voltage compensation obtained by the model predictive controller is the first voltage compensation value and the second voltage compensation value, where the first voltage compensation value is the voltage compensation applied to module A, and the second voltage compensation value is the voltage compensation applied to module B. The voltage reference value of module A is then updated to the first voltage reference value plus the first voltage compensation value, and the voltage reference value of module B is updated to the second voltage reference value plus the second voltage compensation value.

[0090] The sign of the voltage compensation determines the adjustment direction: if the first voltage compensation is positive, the output voltage amplitude of module A is increased; if the first voltage compensation is negative, the output voltage amplitude of module A is decreased. By adjusting the voltage amplitudes of the two modules, the voltage deviation can be reduced, thereby suppressing the reactive circulating current caused by the voltage amplitude difference.

[0091] Secondly, the carrier phase adjustment is superimposed on the carrier phases of the two modules in the dominant circulating current source. Let the original carrier phases of module A and module B be the first carrier phase and the second carrier phase, respectively. The carrier phase adjustment obtained by the model predictive controller is the first carrier phase adjustment and the second carrier phase adjustment, where the first carrier phase adjustment is applied to module A, and the second carrier phase adjustment is applied to module B. The carrier phase of module A is updated to the first carrier phase plus the first carrier phase adjustment, and the carrier phase of module B is updated to the second carrier phase plus the second carrier phase adjustment.

[0092] The sign of the carrier phase adjustment determines the direction of the carrier phase lead or lag. By adjusting the carrier phase of the two modules, the carrier phase difference can be reduced, thereby suppressing the high-frequency circulating current caused by carrier asynchrony.

[0093] Furthermore, during the execution of compensation control, the circulating current amplitude of the dominant circulating current source is continuously monitored. When the circulating current amplitude drops below a preset acceptable threshold, compensation control is stopped. The preset acceptable threshold is a critical value set according to the system's allowable range for circulating current; for example, it can be set to 2% of the rated current or 0.1 amperes.

[0094] Specifically, during the execution of compensation control, the circulating current amplitude between the two modules in the dominant circulating current source is calculated in real time and compared with a preset qualified threshold. When the circulating current amplitude remains below the qualified threshold, it indicates that the circulating current has been effectively suppressed and the system has returned to normal operation. At this point, compensation control is stopped to avoid introducing new disturbances or causing unnecessary energy loss due to over-adjustment. If the circulating current amplitude rises again above the threshold after compensation is stopped, the model predictive controller is re-triggered to solve for a new compensation amount and perform differentiated compensation control. Through this stopping mechanism, the control system only actively intervenes when the circulating current exceeds the limit and remains silent when the circulating current is normal, achieving on-demand control and improving the system's operating efficiency.

[0095] In summary, the embodiments of this application have at least the following technical effects: This invention first acquires real-time data on the output current, output voltage, and carrier phase of each module to calculate the circulating current value and identify the module pair with the largest circulating current amplitude as the dominant circulating current source. This achieves precise location of the circulating current source, avoiding the inefficiency and additional disturbances caused by uniform adjustment of all modules in traditional methods. Second, it extracts the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source and constructs a circulating current deviation degree by combining the deviation between the circulating current value and the system circulating current mean, providing a quantitative characterization of the circulating current state for subsequent control. Third, it judges the circulating current evolution trend based on the rate of change of the circulating current deviation degree and the historical circulating current trajectory, distinguishing between natural convergence and continuous deterioration. During natural convergence, no control action is required, avoiding unnecessary adjustments. Finally, when the circulating current evolution trend is continuously deteriorating, the future circulating current deviation trajectory is predicted by the model predictive controller, with the circulating current deviation and evolution trend as constraints. The voltage compensation and carrier phase adjustment are solved with the goal of minimizing the future circulating current deviation peak value. Differential compensation control is performed on the two modules in the dominant circulating current source, realizing accurate prediction and advance suppression of the circulating current.

[0096] This invention solves the problems of unclear source location of circulating current, lack of judgment of evolution trend, and lag in control response in the prior art, and improves the circulating current suppression effect and dynamic response performance of parallel inverter systems.

[0097] Example 2, as Figure 2 As shown, based on the same inventive concept as the model predictive control-based parallel inverter circulating current suppression method provided in Embodiment 1, this embodiment of the invention also provides a model predictive control-based parallel inverter circulating current suppression system, comprising: The data acquisition module 11 is used to acquire the output current, output voltage and carrier phase of each inverter module in real time, calculate the circulating current value between any two modules, and identify the module pair with the largest circulating current amplitude as the dominant circulating current source. The extraction and calculation module 12 is used to extract the voltage deviation, phase deviation and carrier phase difference between the two modules in the dominant circulating current source, and to calculate the deviation between the circulating current value of the dominant circulating current source and the average circulating current value of the system as the circulating current deviation degree. The judgment module 13 is used to judge the circulation evolution trend of the dominant circulation source based on the rate of change of the circulation deviation and the historical circulation trajectory. The circulation evolution trend includes natural convergence and continuous deterioration. The prediction and solution module 14 is used to predict the future trajectory of the circulating current deviation by means of the circulating current deviation and the circulating current evolution trend as constraints, combined with the physical constraint that the sum of the output currents of each module is equal to the load current, when the circulating current evolution trend is continuously deteriorating, and to solve the voltage compensation amount and carrier phase adjustment amount of the two modules in the dominant circulating current source. The differential compensation control module 15 is used to perform differential compensation control on the two modules in the dominant circulating current source according to the voltage compensation amount and the carrier phase adjustment amount.

[0098] Specifically, the data acquisition module 11 is used for: Specifically, the output current, output voltage, and carrier phase of each inverter module are collected in real time. The circulating current value between any two modules is calculated, and the module pair with the largest circulating current amplitude is identified as the dominant circulating current source, including: The instantaneous values ​​of the output current, output voltage, and carrier phase angle of each inverter module are synchronously collected at a preset sampling frequency. For any two modules i and j, calculate the instantaneous value of the circulating current based on the instantaneous values ​​of their output currents. The instantaneous value of the circulating current is equal to half the difference between the output current of module i and the output current of module j. Calculate the root mean square value of the instantaneous value of the circulation within the preset time window, and use it as the circulation amplitude between module i and module j; Traverse all module pairs and mark the module pair with the largest circulation amplitude as the dominant circulation source.

[0099] Specifically, the extraction calculation module 12 is used for: Specifically, the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source are extracted, and the deviation between the circulating current value of the dominant circulating current source and the average circulating current of the system is calculated as the circulating current deviation, including: Calculate the difference in output voltage amplitude between the two modules in the dominant circulating current source as the voltage deviation; Calculate the difference in the output voltage phase angle between the two modules in the dominant circulating current source, and use it as the phase deviation; Calculate the difference in carrier phase angle between the two modules in the dominant circulating current source, and use it as the carrier phase difference; Calculate the arithmetic mean of the circulating current values ​​for all modules, and use it as the system circulating current mean. The difference between the circulation value of the dominant circulation source and the average circulation value of the system is calculated as the circulation deviation.

[0100] Specifically, the judgment module 13 is used for: Specifically, based on the rate of change of the circulation deviation and the historical circulation trajectory, the circulation evolution trend of the dominant circulation source is determined, including: Collect the circulation deviation of the N sampling times preceding the current time to form a deviation time series; Calculate the first difference of the deviation time series as the deviation rate of change series; Calculate the first difference of the deviation rate change sequence as the deviation change acceleration; When the rate of change of deviation is positive and the acceleration of change of deviation is positive, it is judged as a continuous deterioration trend; when the rate of change of deviation is negative and the acceleration of change of deviation is negative, it is judged as a natural convergence trend.

[0101] Specifically, the prediction and solution module 14 is used for: Specifically, when the circulating current evolution trend is continuously deteriorating, the evolution trajectory of the future circulating current deviation is predicted by a model predictive controller, constrained by the circulating current deviation and the circulating current evolution trend. With the objective of minimizing the peak value of the future circulating current deviation, the voltage compensation and carrier phase adjustment of the two modules in the dominant circulating current source are solved, including: Collect the circulation deviation of the M sampling times before the current time to form a deviation time series, and calculate the variance of the deviation time series as the circulation fluctuation intensity; Calculate the correlation coefficients between each pair of voltage deviation, phase deviation, and carrier phase difference, and use them as the module differential coupling degree; The prediction model inside the model prediction controller uses the current circulation deviation and circulation evolution trend as input to predict the initial circulation deviation evolution trajectory corresponding to different candidate compensation amounts applied in the future preset time domain. Using the circulation fluctuation intensity as a fluctuation correction factor, the predicted value at each moment in the initial circulation deviation evolution trajectory is multiplied by the fluctuation correction factor to obtain the corrected circulation deviation evolution trajectory. Using the module difference coupling degree as a coupling correction factor, the peak value of the circulation deviation in the optimization objective of the optimization solver inside the model prediction controller is multiplied by the coupling correction factor to obtain the corrected optimization objective; By optimizing the solver and minimizing the modified optimization objective, the optimal voltage compensation and carrier phase adjustment are selected from the candidate compensation values.

[0102] The prediction model is pre-built through the following steps: Collect historical operating data of the parallel inverter system under continuously deteriorating operating conditions. The historical operating data includes the output current, output voltage, carrier phase, circulating current value of each module, as well as the applied voltage compensation amount and carrier phase adjustment amount. The training sample set is constructed by taking the current state and the applied compensation amount in the historical operation data as input and the actual circulation value of the next moment in the historical operation data as output. A prediction model is constructed and trained using a supervised learning algorithm. The training objective is to minimize the mean square error between the predicted circulation value and the actual circulation value until convergence is achieved, thus obtaining the trained prediction model.

[0103] Specifically, the optimization solver uses a quadratic programming algorithm to solve for the optimal control sequence that minimizes the cost function, including: Construct a cost function, which includes the sum of squares of the predicted circulating current deviation at each time point in the future preset time domain, and a penalty term for the rate of change of voltage compensation and carrier phase adjustment. The optimal control sequence that minimizes the cost function is solved using a quadratic programming algorithm. The initial voltage compensation amount and the initial carrier phase adjustment amount at the current moment are extracted from the optimal control sequence as candidate compensation amounts.

[0104] Specifically, taking into account the physical constraint that the sum of the output currents of all modules equals the load current, the model predictive controller predicts the evolution trajectory of the future circulating current deviation, including: Obtain real-time measurements of the load current; Add a summation constraint term to the prediction model to ensure that the sum of the predicted output currents of each module deviates from the load current within a preset range. Based on the prediction model with summation constraints, the predicted value of the initial circulation deviation at each time point within the preset future time domain is calculated recursively.

[0105] Specifically, the differential compensation control module 15 is used for: Specifically, based on the voltage compensation amount and carrier phase adjustment amount, differentiated compensation control is performed on the two modules in the dominant circulating current source, including: The voltage compensation is superimposed onto the voltage reference values ​​of the two modules in the dominant circulating current source; The carrier phase adjustment is superimposed on the carrier phases of the two modules in the dominant circulating current source, respectively; During the execution of compensation control, the circulation amplitude of the dominant circulation source is continuously monitored. When the circulation amplitude drops below the preset qualified threshold, compensation control is stopped.

Claims

1. A method for circulating current suppression of parallel inverters based on model predictive control, characterized in that, The method includes: The system collects the output current, output voltage, and carrier phase of each inverter module in real time, calculates the circulating current value between any two modules, and identifies the module pair with the largest circulating current amplitude as the dominant circulating current source. Extract the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and calculate the deviation between the circulating current value of the dominant circulating current source and the average circulating current value of the system as the circulating current deviation. Based on the rate of change of the circulation deviation and the historical circulation trajectory, the circulation evolution trend of the dominant circulation source is determined, and the circulation evolution trend includes natural convergence and continuous deterioration; When the circulating current evolution trend is continuously deteriorating, the circulating current deviation and the circulating current evolution trend are used as constraints. Combined with the physical constraint that the sum of the output currents of each module is equal to the load current, the model prediction controller predicts the future circulating current deviation trajectory. With the goal of minimizing the future circulating current deviation peak, the voltage compensation amount and carrier phase adjustment amount of the two modules in the dominant circulating current source are solved. Based on the voltage compensation amount and carrier phase adjustment amount, differentiated compensation control is performed on the two modules in the dominant circulating current source.

2. The model predictive control based parallel inverter circulating current mitigation method of claim 1, wherein, The system collects the output current, output voltage, and carrier phase of each inverter module in real time, calculates the circulating current value between any two modules, and identifies the module pair with the largest circulating current amplitude as the dominant circulating current source, including: The instantaneous values ​​of the output current, output voltage, and carrier phase angle of each inverter module are synchronously collected at a preset sampling frequency. For any two modules i and j, calculate the instantaneous value of the circulating current based on the instantaneous values ​​of their output currents. The instantaneous value of the circulating current is equal to half the difference between the output current of module i and the output current of module j. Calculate the root mean square value of the instantaneous value of the circulation within the preset time window, and use it as the circulation amplitude between module i and module j; Traverse all module pairs and mark the module pair with the largest circulation amplitude as the dominant circulation source.

3. The model predictive control based parallel inverter circulating current mitigation method of claim 1, wherein, Extract the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and calculate the deviation between the circulating current value of the dominant circulating current source and the system circulating current mean value as the circulating current deviation, including: Calculate the difference in output voltage amplitude between the two modules in the dominant circulating current source as the voltage deviation; Calculate the difference in the output voltage phase angle between the two modules in the dominant circulating current source, and use it as the phase deviation; Calculate the difference in carrier phase angle between the two modules in the dominant circulating current source, and use it as the carrier phase difference; Calculate the arithmetic mean of the circulating current values ​​for all modules, and use it as the system circulating current mean. The difference between the circulation value of the dominant circulation source and the average circulation value of the system is calculated as the circulation deviation.

4. The model predictive control based parallel inverter circulating current mitigation method of claim 1, wherein, Based on the rate of change of the circulation deviation and the historical circulation trajectory, the circulation evolution trend of the dominant circulation source is determined, including: Collect the circulation deviation of the N sampling times preceding the current time to form a deviation time series; Calculate the first difference of the deviation time series as the deviation rate of change series; Calculate the first difference of the deviation rate change sequence as the deviation change acceleration; When the rate of change of deviation is positive and the acceleration of change of deviation is positive, it is judged as a continuous deterioration trend; when the rate of change of deviation is negative and the acceleration of change of deviation is negative, it is judged as a natural convergence trend.

5. The model predictive control based parallel inverter circulating current mitigation method of claim 1, wherein, When the circulating current evolution trend is continuously deteriorating, the evolution trajectory of the future circulating current deviation is predicted by a model predictive controller, constrained by the circulating current deviation and the circulating current evolution trend. With the objective of minimizing the peak value of the future circulating current deviation, the voltage compensation and carrier phase adjustment of the two modules in the dominant circulating current source are solved, including: Collect the circulation deviation of the M sampling times before the current time to form a deviation time series, and calculate the variance of the deviation time series as the circulation fluctuation intensity; Calculate the correlation coefficients between each pair of voltage deviation, phase deviation, and carrier phase difference, and use them as the module differential coupling degree; The prediction model inside the model prediction controller uses the current circulation deviation and circulation evolution trend as input to predict the initial circulation deviation evolution trajectory corresponding to different candidate compensation amounts applied in the future preset time domain. Using the circulation fluctuation intensity as a fluctuation correction factor, the predicted value at each moment in the initial circulation deviation evolution trajectory is multiplied by the fluctuation correction factor to obtain the corrected circulation deviation evolution trajectory. Using the module difference coupling degree as a coupling correction factor, the peak value of the circulation deviation in the optimization objective of the optimization solver inside the model prediction controller is multiplied by the coupling correction factor to obtain the corrected optimization objective; By optimizing the solver and minimizing the modified optimization objective, the optimal voltage compensation and carrier phase adjustment are selected from the candidate compensation values.

6. The model predictive control based parallel inverter circulating current mitigation method of claim 5, wherein, The prediction model is pre-built through the following steps: Collect historical operating data of the parallel inverter system under continuously deteriorating operating conditions. The historical operating data includes the output current, output voltage, carrier phase, circulating current value of each module, as well as the applied voltage compensation amount and carrier phase adjustment amount. The training sample set is constructed by taking the current state and the applied compensation amount in the historical operation data as input and the actual circulation value of the next moment in the historical operation data as output. A prediction model is constructed and trained using a supervised learning algorithm. The training objective is to minimize the mean square error between the predicted circulation value and the actual circulation value until convergence is achieved, thus obtaining the trained prediction model.

7. The model predictive control based parallel inverter circulating current mitigation method of claim 5, wherein, The optimization solver uses a quadratic programming algorithm to find the optimal control sequence that minimizes the cost function, including: Construct a cost function, which includes the sum of squares of the predicted circulating current deviation at each time point in the future preset time domain, and a penalty term for the rate of change of voltage compensation and carrier phase adjustment. The optimal control sequence that minimizes the cost function is solved using a quadratic programming algorithm. The initial voltage compensation amount and the initial carrier phase adjustment amount at the current moment are extracted from the optimal control sequence as candidate compensation amounts.

8. The model predictive control based parallel inverter circulating current mitigation method of claim 5, wherein, Based on the physical constraint that the sum of the output currents of all modules equals the load current, a model predictive controller is used to predict the evolution trajectory of future circulating current deviation, including: Obtain real-time measurements of the load current; Add a summation constraint term to the prediction model to ensure that the sum of the predicted output currents of each module deviates from the load current within a preset range. Based on the prediction model with summation constraints, the predicted value of the initial circulation deviation at each time point within the preset future time domain is calculated recursively.

9. The model predictive control based parallel inverter circulating current mitigation method of claim 1, wherein, Based on the voltage compensation amount and carrier phase adjustment amount, differentiated compensation control is performed on the two modules in the dominant circulating current source, including: The voltage compensation is superimposed onto the voltage reference values ​​of the two modules in the dominant circulating current source; The carrier phase adjustment is superimposed on the carrier phases of the two modules in the dominant circulating current source, respectively; During the execution of compensation control, the circulation amplitude of the dominant circulation source is continuously monitored. When the circulation amplitude drops below the preset qualified threshold, compensation control is stopped.

10. A model predictive control based circulating current suppression system for parallel connected inverters, characterized in that, The method for suppressing circulating current in a parallel inverter based on model predictive control as described in any one of claims 1-9 includes: The data acquisition module is used to collect the output current, output voltage and carrier phase of each inverter module in real time, calculate the circulating current value between any two modules, and identify the module pair with the largest circulating current amplitude as the dominant circulating current source. The extraction and calculation module is used to extract the voltage deviation, phase deviation, and carrier phase difference between the two modules in the dominant circulating current source, and to calculate the deviation between the circulating current value of the dominant circulating current source and the average circulating current value of the system as the circulating current deviation. The judgment module is used to determine the circulation evolution trend of the dominant circulation source based on the rate of change of the circulation deviation and the historical circulation trajectory. The circulation evolution trend includes natural convergence and continuous deterioration. The prediction and solution module is used to predict the future trajectory of the circulating current deviation when the circulating current evolution trend is continuously deteriorating, taking the circulating current deviation and the circulating current evolution trend as constraints, and combining the physical constraint that the sum of the output currents of each module is equal to the load current, through the model prediction controller, and with the goal of minimizing the peak value of the future circulating current deviation, to solve the voltage compensation amount and carrier phase adjustment amount of the two modules in the dominant circulating current source; The differential compensation control module is used to perform differential compensation control on the two modules in the dominant circulating current source according to the voltage compensation amount and the carrier phase adjustment amount.

Images