A co-split dc bus multi-three-level inverter cooperative control method and system

Abstract

The application discloses a kind of co-split direct current bus multi-three-level inverter collaborative control method and system, belong to power electronic multi-motor drive technical field. Including: S1, establish unified coordination window;S2, calculate the natural midpoint charge of each module;S3, construct the feasible region of each module's correction charge;S4, predict the residual disturbance charge of common midpoint;S5, based on the unbalanced state of upper and lower capacitor charge, obtain target total midpoint charge, calculate active correction charge demand;S6, obtain each module target correction charge in the total reachable range distribution;S7, generate gate sequence output under the condition of PWM physical constraint;S8, residual error is obtained, and the observation value of residual disturbance charge is obtained from residual error and the prediction of next coordination window residual disturbance charge is updated;S9, execute system abnormal diagnosis.The application can realize the collaborative control of multiple inverter common midpoint, and the instruction is strong in executable, robust, with fault degradation operation ability.

Description

Technical Field

[0001] This invention relates to the field of power electronics multi-motor drive technology, and in particular to a method and system for coordinated control of multiple three-level inverters with a common split DC bus. Background Technology

[0002] In rail transit, mining machinery, ship propulsion, and industrial multi-motor equipment, a common architecture is to use multiple three-level inverters sharing a single split DC bus, with each inverter driving an independent AC motor. Multiple inverters act together at a common midpoint through switching redundancy vectors, and the voltage deviation between the upper and lower capacitors is determined by the sum of the midpoint currents of all inverters.

[0003] The existing control schemes have the following shortcomings: First, they lack system-level coordination mechanisms using the common midpoint charge as a unified physical quantity, and the midpoint regulation capability of each inverter is not uniformly quantified in conjunction with PWM physical constraints; Second, they fail to distinguish between the natural midpoint charge generated by the basic PWM itself and the correction charge that needs to be actively compensated, so the natural midpoint charge that can be mutually canceled between modules will still trigger redundant vector actions, causing unnecessary switching; Third, they lack a charge conservation and distribution mechanism among multiple modules, which can easily lead to issuing unexecutable control commands to modules with high modulation, insufficient redundancy time, or faults; Fourth, they lack closed-loop correction methods based on the measured residual of the common midpoint for various unmodeled disturbances such as dead zones, sampling communication delays, and device parameter deviations, and in some schemes, the prediction module directly outputs the switching gate signal, which poses an engineering safety risk. Summary of the Invention

[0004] The purpose of this invention is to provide a collaborative control method and system for multiple three-level inverters with a common split DC bus, thereby solving the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention provides a method for coordinated control of multiple three-level inverters with a common split DC bus, comprising the following steps: S1. Establish a unified coordination window and obtain the basic PWM sequence of each module, the predicted value of phase current, and the historical data of the common bus at the start of the coordination window. S2. Calculate the natural midpoint charge of each module based on the planned basic PWM sequence and the predicted phase current. S3. Based on the point charge sensitivity in the redundant state and the PWM physical constraints, construct the modified charge feasible region for each module. S4. Predict the remaining perturbation charge at the common midpoint; S5. Based on the unbalanced state of the upper and lower capacitor charges, the target total midpoint charge is obtained, and combined with the natural midpoint charge of all modules and the predicted value of the remaining disturbance charge, the active correction charge requirement is calculated. S6. Obtain the total reachability of the system from the feasible domain of the corrected charge of each module, and allocate the target corrected charge of each module within the total reachability. S7. Each module converts the target correction charge into a complementary action time increment of the equivalent redundant state, and generates a gate sequence output under PWM physical constraints. S8. Calculate the residual using the measured current at the common midpoint and the reconstructed midpoint current from the actual switching states of each module. Obtain the observed value of the remaining disturbance charge from the residual and update the prediction of the remaining disturbance charge in the next coordination window using the observed value. S9. Perform system anomaly diagnosis: If an anomaly is detected, perform hierarchical rollback control; if no anomaly is detected, proceed to the next coordination window for cyclic execution.

[0006] Preferably, the unified coordination window in S1 is defined as follows: ; , ; In the formula, The PWM period; The number of PWM cycles contained in a single coordination window; Number the coordination window; For the first One coordination window; , These represent the start and end times of the coordination window, respectively.

[0007] Preferably, the charge imbalance state of the upper and lower capacitors, the target total midpoint charge, and the active correction charge requirement in S5 are calculated according to the following formula: ; ; ; In the formula, For the first The capacitors above and below the window are in an unbalanced charge state. , These are the capacitance values ​​for the upper and lower capacitors, respectively. , The first upper and lower capacitor voltages at the window; The target at the end of the window is unbalanced; Let be the convergence coefficient, and ; The total midpoint charge of the target in this window; For the first Module No. The natural midpoint charge of the window; For the first Predicted residual perturbation charge at the window's common midpoint; To proactively adjust charge requirements.

[0008] Preferably, the equivalent redundancy state in S7 consists of positive and negative redundancy states where the output voltage vectors in the same stationary two-phase coordinate system are different, but the midpoint currents are different; the midpoint charge sensitivity of the redundancy state pair and the midpoint charge correction of the single module satisfy the following: ; ; In the formula, Midpoint charge sensitivity; , These are the midpoint currents corresponding to the positive and negative redundant states, respectively. For the first Module correction of point charge; For the first Module No. The set of all redundant state pairs for each PWM cycle; This is the time increment for the redundant state pairs; when each redundant state pair undergoes complementary time transition, the total time of action of the positive and negative redundant states remains unchanged. PWM physical constraints include at least minimum effective pulse width, dead time, and legal commutation constraints. Furthermore, PWM physical constraints may also include one or more of the following: update prohibition region, timer resolution, modulation, overmodulation, device fault shielding, and common-mode voltage limit.

[0009] Preferably, the update formulas for the instantaneous residual at the common midpoint, the observed residual perturbation charge, and the perturbation prediction for the next window in S8 are as follows: ; ; ; In the formula, The instantaneous residual at the common midpoint; The measured current at the common midpoint; For the first The module reconstructs the midpoint current based on the actual switching state; These are the observations of the remaining perturbation charge; These are the filter coefficients; This is a safety limiting function.

[0010] Preferably, the total reachable range of the system is obtained by summing the feasible regions of the corrected charges of each module: ; In the formula, , The first The module corrects the lower and upper limits of the feasible charge region; When the active correction charge demand exceeds the total reachability of the system, the active correction charge demand is projected onto the total reachability. Incomplete correction charge Deferred to the next coordination window; if the correction cannot be completed in multiple consecutive windows, a regulation limitation, load derating, or midpoint deviation warning will be triggered. The target correction charge for each module is solved using constraint optimization: ; ; In the formula, For the first The module's correction point charge; This represents the total reachable range of the system. The projected distributable charge; This is a limited projection function; Module weights; This is the penalty coefficient; The cost function represents at least one of the following: additional commutation times, switching losses, capacitor RMS current increment, device junction temperature, and task rate of change. For the first The module corrects the charge feasible region.

[0011] Preferably, the graded rollback control includes: switching to deterministic residual estimation and continuing to perform system-level charge allocation when the prediction confidence is insufficient; rolling back to the local midpoint balance control of each module when communication is abnormal; removing the corrected charge feasible region of the faulty module and redistributing the correction task among the remaining normal modules when a module fails; and rolling back to the local midpoint balance control of each module when a common sensor is abnormal or the feasible region calculation fails. The residual disturbance charge prediction model only outputs the predicted value and prediction confidence of the residual disturbance charge, and does not directly output the switching state or gate signal; the gate sequence is generated by the local controller of each module according to the PWM physical constraints.

[0012] This invention also provides a shared-split DC bus multi-three-level inverter collaborative control system, comprising: a shared-split DC bus, M three-level inverters, M independent AC motors, upper and lower capacitor voltage detection units, a common neutral point current detection unit, phase current detection units for each module, local controllers for each module, a communication synchronization unit, and a system coordination controller. ; The common split DC bus includes an upper capacitor and a lower capacitor connected in series. The upper capacitor is connected between the common positive bus terminal P and the common midpoint terminal O, and the lower capacitor is connected between the common midpoint terminal O and the common negative bus terminal N. Each three-level inverter is connected to P, O, and N on the DC side at the same time. Each three-level inverter is connected to an independent AC motor on the AC side. The AC sides of each inverter are not connected in parallel. The upper and lower capacitor voltage detection unit, the common midpoint current detection unit, and the phase current detection unit of each module respectively collect the corresponding voltage and current signals; Each module's local controller is used to generate the planned basic PWM sequence, receive the target corrected charge, and output a valid gate sequence; The communication synchronization unit is used for time base alignment of multiple modules and transmission of interactive data; The system coordination controller is used to establish a unified coordination window to complete charge aggregation, demand calculation, target correction charge allocation, residual prediction update, and anomaly classification backoff control.

[0013] Preferably, the system coordination controller is equipped with a natural charge aggregation module, a residual disturbance prediction module, a target charge calculation module, a total reachability construction module, a charge conservation and allocation module, a residual correction module, and a hierarchical backoff state machine; the residual disturbance prediction module only outputs the predicted value and confidence level of the residual disturbance charge, and does not output switch or gate control signals; The three-level inverter is any one of NPC, ANPC, T-type, or hybrid clamped three-level inverters, and has equivalent redundant switching states with the same static two-phase coordinate system voltage vector and different midpoint currents. The common midpoint current detection unit is an independent midpoint current sensor, which can be obtained either by summing the currents of each branch or by estimating the voltage and capacitance parameters of the upper and lower capacitors.

[0014] Preferably, the local controller of each module is also used to calculate the feasible domain of the natural midpoint charge and the corrected charge of the module, convert the target corrected charge into the time increment of the complementary effect of the equivalent redundant state, and after the window is executed, feed back the actual completed charge, the uncompleted charge and the module running status to the system coordination controller. The data transmitted by the communication synchronization unit includes window number, timestamp, data validity flag, module feasible region, target correction charge, and module execution status.

[0015] Therefore, the present invention employs the above-mentioned method and system for coordinated control of multiple three-level inverters with a common split DC bus, which has the following beneficial effects: 1. Using the common midpoint charge as a unified physical quantity to carry out multi-inverter collaborative control, quantify the adjustment capability of each module under the physical constraints of PWM, complete charge distribution within the total reachable range of the system, avoid issuing unexecutable control commands, and improve the reliability of command execution.

[0016] 2. Distinguish between natural midpoint charge and active correction charge, and prioritize the mutual cancellation effect of natural midpoint charge between different modules to reduce unnecessary switching of redundant states and reduce switching operation overhead.

[0017] 3. Closed-loop correction is achieved through measured residuals at the common midpoint, which can suppress unmodeled disturbances such as dead zone, sampling delay, and device parameter errors, thereby improving system robustness; the prediction module only outputs disturbance charge and does not directly output gate signal, reducing the risk of device runaway caused by prediction anomalies.

[0018] 4. The system adopts equivalent redundant state complementary transfer time to achieve midpoint adjustment, keeping the average volt-second integral of the base reference voltage of each motor unchanged and without altering the original motor control link; at the same time, it has a graded backoff mechanism, which can degrade operation when there is module failure or communication abnormality, and the system has good scalability and fault tolerance.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] Figure 1 This is an overall flowchart of a collaborative control method for multiple three-level inverters with a common split DC bus according to the present invention; Figure 2 This is a general framework diagram of a multi-three-level inverter collaborative control system with a common split DC bus according to the present invention; Figure 3 This is a schematic diagram illustrating the calculation principle of the feasible region of a point charge in a single module of the present invention. Figure 4 This is a diagram showing the input-output and timing relationship of the trans-inverter timing prediction model of the present invention; Figure 5 This is a diagram showing the relationship between the midpoint corrected charge conservation distribution and local execution in this invention; Figure 6 This is a schematic diagram of the time increment of the equivalent redundant state complementary action of the present invention; Figure 7 This is the common midpoint residual closed-loop and graded backoff control diagram of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0022] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0023] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0024] like Figures 1-7 As shown, this invention provides a collaborative control method and system for multiple three-level inverters on a common split DC bus. Each three-level inverter is a module, and there is a one-to-one correspondence between modules and three-level inverters. The module number is... The negative busbar end is fixed and denoted as N. This indicates the number of three-level inverters.

[0025] Current direction convention: The positive direction of each phase current is defined as flowing from the inverter bridge arm output to the motor; the positive direction of the inverter midpoint current is defined as flowing from the common midpoint O to the inverter; when a phase bridge arm is connected to the common midpoint O and the phase current is positive, that phase contributes positively to the midpoint current. Natural midpoint charge: Without applying the additional midpoint adjustment of this invention, only the planned basic PWM sequence is executed to coordinate the midpoint charge generated by the inverter modules within the window. Corrected charge feasible region: The set of midpoint corrected charges that the modules can achieve by redistributing the redundancy state action time under the premise of keeping the average volt-second integral of the voltage vector in the basic stationary two-phase coordinate system unchanged and satisfying all PWM physical constraints; different from the absolute charge feasible region that includes natural midpoint charge. Residual disturbance charge: The residual charge after deducting the midpoint charge obtained by each module according to the actual switching state and phase current reconstruction from the measured total charge at the common midpoint; characterizing the dead zone, device voltage drop, capacitor parameter error, sampling and communication delay, sensor bias, and the combined disturbance of the unmodeled midpoint branch.

[0026] This invention applies to a multi-three-level inverter collaborative drive system with a shared DC bus: each inverter shares a common P, O, and N bus on its DC side, while its AC side drives independent motors and they are not connected in parallel. This application differs from single-three-level inverter and AC-side parallel inverter solutions. The hardware architecture of a multi-three-level inverter collaborative control system with a shared DC bus provided in this embodiment is as follows: Figure 2 As shown, M three-level inverters are connected to a common positive bus terminal P, a common neutral point terminal O, and a common negative bus terminal N on their DC sides. Each inverter drives an independent AC motor on its AC side, and the AC sides of the inverters are not connected in parallel. The three-level inverter can be an NPC, ANPC, T-type, or hybrid clamped topology, as long as the topology has equivalent redundant switching states with the same output static two-phase coordinate system voltage vector and different midpoint currents. The method of this invention uses the common midpoint charge as a unified physical quantity and executes a complete control flow of "natural midpoint charge separation - construction of feasible charge domains for each module - total reachable range charge conservation allocation - equivalent redundant state volt-second conservation execution - common midpoint residual closed-loop correction - anomaly graded backoff". The prediction model of this invention is only used to estimate unmodeled residual disturbances and does not generate switching or gate signals. All PWM gate sequences are generated by the local controller after PWM physical constraint verification. The specific implementation steps are as follows: S1. Establish a unified coordination window ; , In the formula, The PWM period; The number of PWM cycles contained in a single coordination window; Number the coordination window; For the first One coordination window; , These represent the start and end times of the coordination window, respectively; within the coordination window... At the start time, acquire the planned basic PWM sequence, phase current prediction values, and historical data of the common bus for each module; this step must adhere to causal constraints: window Internal prediction and allocation calculations are only allowed to use the cutoff date. Historical measurement data that has ended cannot be used with windows. Measurements that can only be obtained after the end of the process are used in the prediction and allocation calculations of this window; the collected data also includes carrier phase or switching edge time, power direction, device temperature, fault status, window number, and data validity flag; the system preferably adopts common carrier synchronization and unified sampling trigger; when the carriers of each module are not synchronized, the data of multiple modules are aligned in the same integration window by relying on unified timestamps and actual switching edge records.

[0027] S2. Based on the planned basic PWM sequence and the predicted phase current, calculate the natural midpoint charge of each module. First, calculate the planned natural midpoint current. The formula is: ; In the formula, , , This is a status indicator for the connection of the three-phase bridge arms to the common midpoint O in the planned basic PWM sequence; it is set to 1 when the corresponding bridge arm is connected to O, and 0 otherwise. , , For the m-th module, predict the three-phase phase current; Module in The coordination window only executes the midpoint charge generated by the planned base PWM sequence. for: ; The sum of point charges in all modules in nature: ; It should be noted that the natural neutral point charge is calculated from the planned basic PWM sequence and the predicted phase current value; the actual reconstructed neutral point charge after the window ends is calculated using the actual switching state and the measured phase current, and the two data sources cannot be mixed; the natural neutral point charges of all modules are summed, and the positive and negative cancellation effect of the natural neutral point charges between different modules' electric and feedback operating conditions is given priority to reduce the need for subsequent active correction.

[0028] S3. Based on the point charge sensitivity in the redundant state and the PWM physical constraints, construct the modified charge feasible region for each module. When the set is a continuous interval or uses a safe and conservative interval, it is represented as , Indicates the first The module can achieve a midpoint correction charge range while maintaining the average volt-second integral of the base reference voltage. The PWM physical constraints must include at least minimum effective pulse width, dead zone, and legal commutation constraints; these three constraints are mandatory. Furthermore, the PWM physical constraints may also include one or more of the following: update prohibition region, timer resolution, modulation index, overmodulation, device fault shielding, and common-mode voltage limitation. The vector is composed of the total time increments of all modules. The constraint set consists of minimum effective pulse width, dead zone, update-prohibited region, legal commutation, timer resolution, modulation and overmodulation state, extra commutation times, device temperature, fault shielding, current prediction error, and optional common-mode voltage limit. Under the conservative continuous adjustment method, the minimum effective pulse width constraint for: ; Deformation yields time increment boundaries for: ; No. Modified charge feasible region of the module Defined as: ; The feasible region of absolute charge containing the point charge in nature for: ; Solving for the upper and lower boundaries: ; ; In the formula, For the first Point charge sensitivity vector in the module Transpose of; , The first The module corrects the lower and upper limits of the feasible charge region; , The first Module No. The PWM cycle number Lower and upper limits of the effect time increment of the group of redundant states There are two engineering implementation methods for the feasible region: The first type is the conservative continuous interval mode, which forces the duration of redundant states to meet the minimum pulse width constraint, outputting a closed feasible region with low computational complexity, suitable for real-time control. The second type is the discrete enumeration mode, which allows a set of redundant states to not appear at all within the PWM cycle. This requires legal sequence enumeration, segmented intervals, or mixed integer programming to obtain the complete feasible region. The aforementioned conservative closed interval can be used as the safe conservative feasible region for the discrete enumeration mode. The system-level allocation object of this invention is the modified charge feasible region. The point charge in nature has already been deducted from the active correction charge requirement and will not be counted again in the allocation calculation.

[0029] S4. Predicted value of residual disturbance charge at the common midpoint ; The k-th window, predicted based on historical data, is affected by non-ideal factors such as dead zone, device voltage drop, sampling delay, sensor bias, and unmodeled midpoint branches, resulting in disturbance charges. The input to the trans-inverter timing prediction model includes common bus characteristics, timing operation characteristics of each module, and module capability quality characteristics. The input cutoff... End of historical measured sequence ,as well as Planning characteristics Output the predicted value of the remaining perturbation charge. and prediction confidence : ; The lowest simplified implementation can omit the complex prediction model and directly use the low-pass extrapolation of the residuals from the previous window: ; In the formula, The filter coefficients are between 0 and 1. This is a safety limiting function.

[0030] The prediction model can employ any algorithm that satisfies causality, such as state-space model, Kalman filter, recursive least squares, sliding mode observer, or autoregressive model based on historical residuals. The observed residual perturbation charge values ​​are only used to update subsequent window predictions, and it is prohibited to modify the allocation instructions of windows that have already been executed in reverse.

[0031] S5, the unbalanced charge state of the upper and lower capacitors, the target total midpoint charge, and the active correction charge requirement are calculated according to the following formula: ; ; ; In the formula, For the first The capacitors above and below the window are in an unbalanced charge state. , These are the capacitance values ​​for the upper and lower capacitors, respectively. , The first upper and lower capacitor voltages at the window; The target at the end of the window is unbalanced; Let be the convergence coefficient, and ; The total midpoint charge of the target in this window; For the first Module No. The natural midpoint charge of the window; For the first Predicted residual perturbation charge at the window's common midpoint; To proactively correct charge demand; this charge imbalance model adapts to operating conditions with unequal upper and lower capacitor parameters and fluctuating total DC bus voltage; when = When the total DC bus voltage is stable, the model automatically degenerates into the traditional midpoint voltage difference-charge correspondence.

[0032] S6. The total reachable range of the system is obtained by correcting the charge feasible region of each module: The total reachable range of the system is obtained by summing the feasible regions of the corrected charges of each module: ; In the formula, For the total reachable range of the system; when actively adjusting the charge demand. Exceeding the total reachable range of the system The active correction charge requirement will be projected onto the total reachable range. Incomplete correction charge Deferred to the next coordination window; if the correction cannot be completed in multiple consecutive windows, a regulation limitation, load derating, or midpoint deviation warning will be triggered. The amplitude-limited projection function is used; the target correction charge of each module is solved through constraint optimization. ; ; In the formula, For the first The module's correction point charge; The projected distributable charge; Module weights; This is the penalty coefficient; The cost function represents at least one of the following: additional commutation times, switching losses, capacitor RMS current increment, device junction temperature, and task rate of change. For the first The module corrects the charge feasible region; in addition to quadratic programming, the allocation solution algorithm can also be replaced by linear programming, marginal cost sorting, projected gradient, ADMM, distributed consistency, and master-slave priority allocation.

[0033] S7, Each module will correct the target charge. Complementary action time increments converted to equivalent redundant states Under PWM physical constraints, a gate sequence output is generated; the equivalent redundancy states are positive and negative redundancy states with the same output static two-phase coordinate system voltage vector and different midpoint current; the redundancy states perform complementary time transitions, and the action time satisfies the following relationship: ; ; ; The midpoint charge sensitivity of the redundant state pair and the midpoint charge correction of the single module satisfy the following: ; ; In the formula, Midpoint charge sensitivity; , These are the midpoint currents corresponding to the positive and negative redundant states, respectively. For the first Module correction of point charge; For the first Module No. The set of all redundant state pairs for each PWM cycle; This is the time increment for redundant state pairs; when each redundant state pair undergoes complementary time transition, the total time of action of the positive and negative redundant states remains unchanged, i.e. In the formula, , These represent the base durations for the positive and negative redundant states, respectively. , The effective time after adjusting for time increment.

[0034] After the local controller obtains the action time increment, it arranges the redundant state sequence. The arrangement method can be a symmetrical sequence, a minimum commutation sequence, or a thermal stress equalization sequence. After the arrangement is completed, timer quantization, minimum pulse width verification, dead zone insertion, gate interlocking, and legal commutation check are executed sequentially, and then the gate drive signal is output. After the window execution is completed, the local controller completes two sets of charge reconstruction: (1) The actual midpoint charge of this module can be reconstructed based on the actual switching state and measured phase current. (2) Reconstructing the counterfactual fundamental charge based on measured phase currents and the saved original fundamental PWM sequence; Calculate the actual effective correction charge: ; This calculation is used to isolate phase current prediction errors, avoiding misjudging phase current prediction errors as correction effects brought about by redundant state time adjustments; the local controller feeds back the actual completed charge and incomplete charge to the system coordination controller for consistency evaluation in the next window. PWM physical constraints include at least minimum effective pulse width, dead time, and legal commutation constraints, and also include at least one of the following: update prohibition zone, timer resolution, modulation, overmodulation, device fault shielding, and common-mode voltage limit.

[0035] S8. Calculate the residual using the measured current at the common midpoint and the reconstructed midpoint current from the actual switching states of each module. Obtain the observed value of the remaining disturbance charge from the residual and update the prediction of the remaining disturbance charge in the next coordination window using the observed value. The instantaneous residual at the common midpoint, the observed residual perturbation charge, and the update formula for the next window perturbation prediction are as follows: ; ; ; In the formula, The instantaneous residual at the common midpoint; The measured current at the common midpoint; For the first The module reconstructs the midpoint current based on the actual switching state; These are the observations of the remaining perturbation charge; The remaining disturbance charge observations are used not only for updating predictions but also for anomaly diagnosis. They can identify communication asynchrony, switch status feedback errors, current sensor bias, module failure to execute instructions, and unmodeled midpoint branch anomalies. High-frequency measurement noise is suppressed by low-pass bias estimation and directly fed into the compensation instruction.

[0036] S9. Perform system anomaly diagnosis: If an anomaly is detected, perform tiered rollback control; if no anomaly is detected, proceed to the next coordination window for cyclical execution; tiered rollback control includes: When prediction confidence Below the preset confidence threshold When switching to deterministic residual estimation, apply a safety limit to the predicted disturbance value and continue system-level charge allocation; when communication is continuous When a coordination window fails and the window number, timestamp, or module validity status cannot be obtained, the system will revert to local midpoint balance control in each module. When a module fails, a device exceeds its temperature limit, or a fault shield is activated, the corrected charge feasible region of the faulty module is removed from the joint feasible region. If the faulty module still has a natural midpoint charge output, its natural midpoint charge contribution needs to be estimated or measured. This contribution should not be directly deleted, and the remaining correction task is redistributed within the feasible regions of the remaining normal modules. When a common sensor malfunctions or the feasible region calculation fails, the local midpoint balance control of each module is returned to normal. The remaining disturbance charge prediction model only outputs the predicted value and prediction confidence of the remaining disturbance charge, and does not directly output the switching state or gate signal. The gate sequence is generated by the local controller of each module based on the PWM physical constraints. Regardless of any abnormality in the prediction model or communication, the prediction model cannot bypass the PWM physical constraints to directly output the gate state, in order to avoid switching actions that violate the minimum pulse width, dead zone, and legal commutation constraints.

[0037] This invention also provides a common-split DC bus multi-three-level inverter collaborative control system for executing the aforementioned midpoint charge collaborative control method, comprising: a common-split DC bus, M three-level inverters, M independent AC motors, upper and lower capacitor voltage detection units, a common midpoint current detection unit, phase current detection units for each module, local controllers for each module, a communication synchronization unit, and a system coordination controller. ; The common split DC bus includes the upper capacitor connected in series. Lower capacitor upper capacitor Connect the lower capacitor between the common positive bus terminal P and the common midpoint terminal O. Connect the common neutral point O and the common negative bus N; each three-level inverter is connected to P, O and N on the DC side at the same time, and each three-level inverter is connected to an independent AC motor on the AC side. The AC sides of each inverter are not connected in parallel. The upper and lower capacitor voltage detection unit is used to collect the voltage of the upper capacitor. Lower capacitor voltage The common midpoint current detection unit is used to collect the measured current at the common midpoint. The common midpoint current detection unit can be an independent midpoint current sensor, or it can be obtained by summing the currents of each branch, or it can be estimated based on the voltage and parameters of the upper and lower capacitors; the phase current detection unit of each module is used to collect the three-phase phase current of the corresponding inverter; the local controller of each module is used to generate the planned basic PWM sequence, calculate the feasible region of the natural midpoint charge and the corrected charge of the module; receive the target corrected charge, convert it into the time increment of the complementary action of the equivalent redundant state, and generate a legal gate sequence according to the PWM physical constraints; after the window is executed, it feeds back the actual completed charge, the uncompleted charge and the module operating status to the system coordination controller; The communication synchronization unit is used to transmit interactive data such as window number, timestamp, data validity flag, module feasible region, target correction charge, and module execution status; it realizes time base alignment of multiple modules; the system supports common carrier synchronization, and can also achieve asynchronous module window alignment by relying on unified timestamp and switch edge recording; The system coordination controller is equipped with a natural charge aggregation module, a residual disturbance prediction module, a target charge calculation module, a total reachability construction module, a charge conservation allocation module, a residual correction module, and a hierarchical backoff state machine; it is used to establish a unified coordination window to complete charge aggregation, demand calculation, target corrected charge allocation, residual prediction update, and anomaly hierarchical backoff control; the residual disturbance prediction module only outputs the residual disturbance charge prediction value and confidence level, and does not output switch or gate control signals.

[0038] Furthermore, the three-level inverter can be any one of NPC, ANPC, T-type, or hybrid clamped three-level inverters; the three-level inverter has equivalent redundant switching states with the same static two-phase coordinate system voltage vector and different midpoint currents.

[0039] Example (M=3): A coordination window is set, and the planned natural midpoint charges of the three inverters are respectively... , , The total point charge in all modules is naturally equal to Remaining perturbation charge prediction ;Target total midpoint charge .

[0040] Calculate the active correction charge requirement: .

[0041] Each module corrects the feasible charge domain: ; ; .

[0042] Total reachable range of the system ; Distributable charge lies on the total reachable boundary. .

[0043] The target correction charge of each module is allocated. , , .

[0044] Take a set of redundant state pairs from module 1, and calculate the midpoint charge sensitivity. Calculate the increment of the action time. .

[0045] Before adjustment: Basic duration of positive redundancy state Negative redundancy state basic duration Minimum effective pulse width .

[0046] After adjustment: , Total action time maintained The duration of both states remains unchanged, and the duration of each state is greater than the minimum effective pulse width, satisfying the PWM physical constraints; the midpoint charge correction is completed while keeping the total volt-second integral of the equivalent output voltage vector constant.

[0047] After the window ends, assume that the actual correction charge is completed. , , The actual total corrected charge is -2.24 mC; the result of integrating the common midpoint residual is... It is sent to the prediction update stage for calculation in the next coordination window.

[0048] While retaining the core technical relationships of "natural contribution separation - modified charge capability boundary - common total constraint - equivalent volt-second conservation execution - measured residual closed loop", the local implementation methods can be replaced: (1) Three-level topology: NPC, ANPC, T-type, and hybrid clamping type, as long as they have the same static two-phase coordinate system voltage vector and different midpoint current equivalent redundant switching states; (2) Modified charge feasible region: one-dimensional closed interval, union of multiple intervals, polyhedron, offline lookup table, random set with confidence boundary; feasible region solution can be solved by analytical boundary, legal sequence enumeration, linear programming, quadratic programming, mixed integer programming, table lookup interpolation; (3) Residual perturbation prediction: Various causal prediction algorithms can also be simplified by using residual low-pass extrapolation; (4) Charge allocation algorithms: centralized planning, marginal cost sorting, projected gradient, ADMM, distributed consensus, master-slave priority allocation; (5) Local execution mode: positive and negative redundant small vector time transfer or virtual vector combination, as long as the average volt-second integral of the stationary two-phase coordinate system remains unchanged; (6) Common midpoint current acquisition: independent sensor, summation of current in each branch, and estimation of upper and lower capacitor voltages combined with capacitor parameters; (7) Synchronization scheme: common carrier synchronization, or unified timestamp + switch edge recording aligned window in asynchronous mode; (8) Constraint extension: When common-mode voltage, electromagnetic compatibility, and thermal stress have higher priority, the above constraints can be incorporated into the modified charge feasible region, and the midpoint balance and other performance indicators can be compromised through the cost function.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for coordinated control of multiple three-level inverters with a common split DC bus, characterized in that, Includes the following steps: S1. Establish a unified coordination window and obtain the basic PWM sequence of each module, the predicted value of phase current, and the historical data of the common bus at the start of the coordination window. S2. Calculate the natural midpoint charge of each module based on the planned basic PWM sequence and the predicted phase current. S3. Based on the point charge sensitivity in the redundant state and the PWM physical constraints, construct the modified charge feasible region for each module. S4. Predict the remaining perturbation charge at the common midpoint; S5. Based on the unbalanced state of the upper and lower capacitor charges, the target total midpoint charge is obtained, and combined with the natural midpoint charge of all modules and the predicted value of the remaining disturbance charge, the active correction charge requirement is calculated. S6. Obtain the total reachability of the system from the feasible domain of the corrected charge of each module, and allocate the target corrected charge of each module within the total reachability. S7. Each module converts the target correction charge into a complementary action time increment of the equivalent redundant state, and generates a gate sequence output under PWM physical constraints. S8. Calculate the residual using the measured current at the common midpoint and the reconstructed midpoint current from the actual switching states of each module. Obtain the observed value of the remaining disturbance charge from the residual and update the prediction of the remaining disturbance charge in the next coordination window using the observed value. S9. Perform system anomaly diagnosis: If an anomaly is detected, perform hierarchical rollback control; if no anomaly is detected, proceed to the next coordination window for cyclic execution.

2. The method for coordinated control of multiple three-level inverters with a common split DC bus according to claim 1, characterized in that, The unified coordination window in S1 is defined as: ； ， ； In the formula, The PWM period; The number of PWM cycles contained in a single coordination window; Number the coordination window; For the first One coordination window; , These represent the start and end times of the coordination window, respectively.

3. The method for coordinated control of multiple three-level inverters with a common split DC bus according to claim 1, characterized in that: The charge imbalance state of the upper and lower capacitors in S5, the target total midpoint charge, and the active correction charge requirement are calculated according to the following formula: ； ； ； In the formula, For the first The capacitors above and below the window are in an unbalanced charge state. , These are the capacitance values ​​for the upper and lower capacitors, respectively. , The first upper and lower capacitor voltages at the window; The target at the end of the window is unbalanced; Let be the convergence coefficient, and ; The total midpoint charge of the target in this window; For the first Module No. The natural midpoint charge of the window; For the first Predicted residual perturbation charge at the window's common midpoint; To proactively adjust charge requirements.

4. The method for coordinated control of multiple three-level inverters with a common split DC bus according to claim 1, characterized in that: In S7, the equivalent redundancy states are positive and negative redundancy states with the same output static two-phase coordinate system voltage vectors but different midpoint currents; the midpoint charge sensitivity of the redundancy state pair and the single-module correction of the midpoint charge satisfy: ； ； In the formula, Midpoint charge sensitivity; , These are the midpoint currents corresponding to the positive and negative redundant states, respectively. For the first Module correction of point charge; For the first Module No. The set of all redundant state pairs for each PWM cycle; This is the time increment for the redundant state pairs; when each redundant state pair undergoes complementary time transition, the total time of action of the positive and negative redundant states remains unchanged. PWM physical constraints include at least minimum effective pulse width, dead time, and legal commutation constraints. Furthermore, PWM physical constraints may also include one or more of the following: update prohibition region, timer resolution, modulation, overmodulation, device fault shielding, and common-mode voltage limit.

5. The method for coordinated control of multiple three-level inverters with a common split DC bus according to claim 3, characterized in that: The formulas for updating the instantaneous residual at the common midpoint, the observed residual perturbation charge, and the next window perturbation prediction in S8 are as follows: ； ； ； In the formula, The instantaneous residual at the common midpoint; The measured current at the common midpoint; For the first The module reconstructs the midpoint current based on the actual switching state; These are the observations of the remaining perturbation charge; These are the filter coefficients; This is a safety limiting function.

6. The method for coordinated control of multiple three-level inverters with a common split DC bus according to claim 5, characterized in that: The total reachable range of the system is obtained by summing the feasible regions of the corrected charges of each module: ； In the formula, , The first The module corrects the lower and upper limits of the feasible charge region; When the active correction charge demand exceeds the total reachability of the system, the active correction charge demand is projected onto the total reachability. Incomplete correction charge Deferred to the next coordination window; if the correction cannot be completed in multiple consecutive windows, a regulation limitation, load derating, or midpoint deviation warning will be triggered. The target correction charge for each module is solved using constraint optimization: ； ； In the formula, For the first The module's correction point charge; This represents the total reachable range of the system. The projected distributable charge; This is a limited projection function; Module weights; This is the penalty coefficient; The cost function represents at least one of the following: additional commutation times, switching losses, capacitor RMS current increment, device junction temperature, and task rate of change. For the first The module corrects the charge feasible region.

7. The method for coordinated control of multiple three-level inverters with a common split DC bus according to claim 1, characterized in that, The graded rollback control includes: switching to deterministic residual estimation and continuing system-level charge allocation when the prediction confidence is insufficient; rolling back to local midpoint balance control of each module when communication is abnormal; removing the corrected charge feasible region of the faulty module and redistributing the correction task among the remaining normal modules when a module fails; and rolling back to local midpoint balance control of each module when a common sensor is abnormal or the feasible region calculation fails. The residual disturbance charge prediction model only outputs the predicted value and prediction confidence of the residual disturbance charge, and does not directly output the switching state or gate signal; the gate sequence is generated by the local controller of each module according to the PWM physical constraints.

8. A multi-three-level inverter collaborative control system with a shared split DC bus, applied to the method described in any one of claims 1-7, characterized in that: include: It consists of a shared DC bus, M three-level inverters, M independent AC motors, upper and lower capacitor voltage detection units, a common neutral point current detection unit, phase current detection units for each module, local controllers for each module, a communication synchronization unit, and a system coordination controller. ; The common split DC bus includes an upper capacitor and a lower capacitor connected in series. The upper capacitor is connected between the common positive bus terminal P and the common midpoint terminal O, and the lower capacitor is connected between the common midpoint terminal O and the common negative bus terminal N. Each three-level inverter is connected to P, O, and N on the DC side at the same time. Each three-level inverter is connected to an independent AC motor on the AC side. The AC sides of each inverter are not connected in parallel. The upper and lower capacitor voltage detection unit, the common midpoint current detection unit, and the phase current detection unit of each module respectively collect the corresponding voltage and current signals; Each module's local controller is used to generate the planned basic PWM sequence, receive the target corrected charge, and output a valid gate sequence; The communication synchronization unit is used for time base alignment of multiple modules and transmission of interactive data; The system coordination controller is used to establish a unified coordination window to complete charge aggregation, demand calculation, target correction charge allocation, residual prediction update, and anomaly classification backoff control.

9. A collaborative control system for multiple three-level inverters with a common split DC bus according to claim 8, characterized in that: The system coordination controller is equipped with a natural charge aggregation module, a residual disturbance prediction module, a target charge calculation module, a total reachable range construction module, a charge conservation and allocation module, a residual correction module, and a hierarchical backoff state machine; the residual disturbance prediction module only outputs the predicted value and confidence level of the residual disturbance charge, and does not output switch or gate control signals; The three-level inverter is any one of NPC, ANPC, T-type, or hybrid clamped three-level inverters, and has equivalent redundant switching states with the same static two-phase coordinate system voltage vector and different midpoint currents. The common midpoint current detection unit is an independent midpoint current sensor, which can be obtained either by summing the currents of each branch or by estimating the voltage and capacitance parameters of the upper and lower capacitors.

10. A collaborative control system for multiple three-level inverters with a common split DC bus according to claim 8, characterized in that: The local controller of each module is also used to calculate the feasible region of the natural midpoint charge and the corrected charge of the module, convert the target corrected charge into the time increment of the complementary effect of the equivalent redundant state, and after the window is executed, it feeds back the actual completed charge, the uncompleted charge and the module running status to the system coordination controller. The data transmitted by the communication synchronization unit includes window number, timestamp, data validity flag, module feasible region, target correction charge, and module execution status.

Images