A three-level electric drive chain control system for a doubly-fed wind turbine generator

Abstract

The application discloses a three-level electrical transmission chain control system of a doubly-fed wind turbine, and relates to the technical field of wind power generation power electronic transmission control, which comprises a working condition sensing module, which is used for collecting the operation parameters of a three-level NPC converter of a doubly-fed wind turbine in real time, and constructing a characteristic vector, wherein the midpoint potential deviation, i.e. the difference between the upper bus capacitor voltage and the lower bus capacitor voltage, is a DC bus voltage, and a rotor electrical frequency; the application pairs P-type / N-type redundant small vectors to synthesize virtual small vectors, and pairs adjacent middle vectors to synthesize virtual middle vectors, so that the net injection of all candidate virtual vectors to the midpoint potential is zero under the default allocation coefficient, and the common-mode voltage suppression is realized by pre-eliminating high common-mode vectors, thereby avoiding the dilemma in the traditional scheme, i.e. sacrificing the midpoint regulation capacity in order to suppress the common mode or introducing high common mode in order to balance the midpoint.

Description

Technical Field

[0001] This invention relates to the field of power electronic drive control technology for wind power generation, specifically to a three-level electrical drive chain control system for doubly fed wind turbine generators. Background Technology

[0002] In doubly-fed induction generator (DFIG) wind power systems, the rotor-side converter is the core power conversion device for achieving variable-speed, constant-frequency operation. Traditional two-level converters are limited by device voltage stress, requiring series voltage equalization at medium to high power levels (e.g., above 3MW), resulting in high control complexity and limited reliability. The three-level neutral-point clamp (NPC) topology, by introducing a neutral point potential, reduces the blocking voltage of each power device to half of the DC bus voltage, making it naturally suitable for medium-voltage, high-power wind turbine drive systems and the mainstream solution for large-capacity DFIG units.

[0003] Three-level NPC converters present technical challenges in engineering applications. Firstly, because the DC bus in a three-level topology consists of two series capacitors, the midpoint potential is affected by load current, modulation strategies, and differences in device parameters, leading to low-frequency oscillations and DC bias. Excessive midpoint potential deviation can cause uneven voltage stress on power devices, increased output current harmonics, and in severe cases, even trigger overvoltage protection shutdown. Existing technologies often employ redundant small vector allocation for midpoint potential regulation, but this method is inherently coupled with common-mode voltage suppression; the selection of P-type and N-type small vectors simultaneously affects both the midpoint potential and the common-mode voltage level, making it difficult to simultaneously achieve both objectives. Secondly, the doubly-fed generator rotor winding is connected to the converter via slip rings. Common-mode voltage generates shaft voltage on the rotor shaft, forming shaft current through the bearings. Long-term operation can cause bearing electro-corrosion, shortening the unit's maintenance cycle. The common-mode voltage amplitude generated by large vectors and some zero vectors can reach half the DC bus voltage, which is a major cause of shaft current. Existing common-mode suppression schemes typically achieve this by eliminating high common-mode vectors, but this reduces the number of available vectors, thus affecting current tracking performance and midpoint regulation capability, creating a contradiction among the three. Thirdly, when the grid connection point voltage drops, a large inrush current is generated on the rotor side. Traditional control strategies usually employ discrete mode switching, i.e., switching directly from normal operation mode to fault ride-through mode. This step-like switching causes a sudden change in control input at the moment of switching, triggering a secondary drop in midpoint potential and electromagnetic torque impact, which is detrimental to the mechanical safety of the drive train. Simultaneously, if the active power recovery rate during the fault recovery phase is too fast, it will also cause drastic fluctuations in the midpoint potential and DC bus voltage; if the recovery is too slow, it will fail to meet the grid connection standard's requirements for recovery time.

[0004] Therefore, this application proposes a three-level electrical drive chain control system for doubly fed wind turbine generators to overcome the above-mentioned defects. Summary of the Invention

[0005] To address the aforementioned technical problems, a three-level electrical drive chain control system for doubly-fed wind turbines is provided. This technical solution resolves the issues of discontinuous modulation strategies, current surges, and torque oscillations in the three-level NPC converter of doubly-fed wind turbines during the switching of common-mode voltage suppression, midpoint potential balance, and low-voltage fault ride-through conditions.

[0006] To achieve the above objectives, the technical solution adopted by this invention includes: a condition sensing module, used to collect the operating parameters of the three-level NPC converter of the doubly-fed wind turbine in real time and construct a feature vector. in, The midpoint potential deviation is the upper bus capacitance. Voltage and lower bus capacitance voltage difference DC bus voltage, For rotor electrical frequency, For modulation ratio, The voltage per unit value at the grid connection point is used to determine the current operating condition based on the feature vector. The virtual vector synthesis module, connected to the operating condition sensing module, is used to remove the common-mode voltage amplitude from the 27 basic space voltage vectors of the three-level NPC converter. The vectors are used to obtain a set of retained vectors. Then, based on the set of retained vectors, virtual small vectors and virtual mid-vectors with zero net injection of midpoint potential are constructed to form a candidate virtual vector set. The midpoint balance adjustment module, connected to the virtual vector synthesis module, is used to adjust the midpoint potential based on the real-time midpoint potential deviation. Calculate the duty cycle allocation coefficients for P-type and N-type mini-vectors in the virtual mini-vectors. This ensures that the mean current of the virtual small vector is zero within the control cycle, while feedback correction counteracts the disturbance of the load current to the midpoint potential. The modulation execution module, connected to the virtual vector synthesis module and the midpoint balance adjustment module, is used to select the optimal virtual vector from the candidate virtual vector set and allocate it based on the duty cycle coefficient. Generate the PWM drive signal for the three-level NPC converter; The continuous transition control module, connected to the operating condition sensing module and the modulation execution module, is used to adaptively adjust the rotor current limiting value and active power recovery rate during the voltage dip and recovery process at the grid connection point, so that the midpoint potential deviation is maintained at a constant value during the transition process. Within.

[0007] Preferably, the virtual vector synthesis module removes the common-mode voltage amplitude. The vector, specifically: For the 27 basic space voltage vectors, the common-mode voltage of each vector is calculated using the following formula: in , , These are the voltages of the three-phase output terminals A, B, and C relative to the midpoint of the DC side, corresponding to the vector. Eliminate satisfied The six large vectors PPN, PNP, PNN, NPP, NPN, and NNP, and the common-mode voltage amplitude equal to The zero vectors PPP and NNN are retained, and the remaining 19 basic vectors constitute the reserved vector set, which includes 6 pairs of redundant P-type / N-type small vectors (POO and ONN, OPO and NON, OOP and NNO, PPO and OON, POP and ONO, OPP and NOO), 6 medium vectors (PON, OPN, NPO, NOP, ONP, PNO), and 1 zero vector OOO.

[0008] Preferably, the virtual vector synthesis module constructs virtual small vectors, specifically as follows: For each pair of redundant P-type / N-type small vectors in the retained vector set The virtual small vector is synthesized according to the following formula: in The duty cycle allocation coefficient for the P-type small vector is output by the midpoint balance adjustment module; The direction of the midpoint current corresponding to the P-type small vector is: The direction of the midpoint current corresponding to the N-type small vector is and The virtual small vector in one control cycle The average midpoint current within is: when hour, The net injection of the virtual small vector to the midpoint potential is zero.

[0009] Preferably, the virtual vector synthesis module constructs a virtual vector, specifically as follows: For the six median vectors in the retained vector set, they are paired according to the principle of spatial proximity and opposite midpoint current directions. The paired median vectors are then combined into a virtual median vector using the following formula: in and For two paired vectors, within a preset allowable deviation range, they satisfy... ; Virtual vector in one control cycle The average midpoint current within is: This makes the net injection of the midpoint potential by the mid-vector zero, fundamentally eliminating the low-frequency oscillation of the midpoint potential caused by the mid-vector.

[0010] Preferably, the midpoint balance adjustment module calculates the duty cycle allocation coefficient. Specifically: Midpoint potential deviation As input, the original duty cycle allocation coefficient is output via the PI controller. : in For proportional gain, For integral gain, the reference value of 0.5 corresponds to the symmetrical distribution when the midpoint potential is balanced; right Apply amplitude limiting constraints: in , To prevent narrow pulses caused by an excessively small duty cycle allocation factor; when Limited to or At this point, the integrator stops accumulating and performs anti-integral saturation processing.

[0011] Preferably, the operating condition sensing module determines the current operating condition as follows: According to the feature vector The following conditions shall be used to determine the outcome: when and When the condition is determined to be normal operating condition, all virtual vectors in the candidate virtual vector set can be selected, and the PI proportional gain of the midpoint balance adjustment module is set to... ; when and When the condition is determined to be a midpoint imbalance, the PI proportional gain of the midpoint balance adjustment module is set to... At the same time, the duty cycle allocation coefficient limit range is narrowed to ; when When the fault crossing condition is determined, the continuous transition control module is activated.

[0012] Preferably, the control of the continuous transition control module during the voltage drop process at the grid connection point is specifically as follows: Detected , the rotor current limiting amplitude is linearly tightened from the rated current according to a slope to , and no step switching is adopted during the tightening process; meanwhile, the proportional gain of the PI controller in the neutral point balance adjustment module is switched to , and the amplitude limiting range of the duty cycle distribution coefficient is reduced to ; during the tightening process of the rotor current limiting amplitude, if is detected, the tightening is suspended, and the current is maintained unchanged, and the tightening is resumed after .

[0013] Preferably, the control of the continuous transition control module during the recovery process of the grid-connected point voltage is specifically: after is detected, the active power recovery rate is adaptively determined by the following formula: wherein is the rated active power, is the recovery time constant required by the grid connection standard, and the value range is ; when occurs during the recovery process, the active power recovery rate is automatically reduced to , ensuring that the neutral point potential deviation does not exceed the limit; when the active power is restored to and lasts for more than , the PI proportional gain of the neutral point balance adjustment module is restored from according to a slope to , completing the continuous transition from the fault ride-through operating condition to the normal operating condition.

[0014] Preferably, the selection of the optimal virtual vector by the modulation execution module is specifically: according to the rotor current reference value and the current measured rotor current value , the current error is calculated; in the candidate virtual vector set, the predicted current value after each virtual vector acts for one control cycle is calculated one by one : wherein, is leakage inductance, For equivalent resistance, The rotor back electromotive force; Choose the current tracking cost function The smallest virtual vector is output as the optimal virtual vector.

[0015] Preferably, the system further includes a narrow pulse processing module connected to the modulation execution module, specifically: When the actual action time of the P-type or N-type component of the virtual small vector Less than the preset narrow pulse threshold At that time, the duration of action of this component is forcibly set to... At the same time, the duration of action of the remaining components within the same cycle is reduced proportionally, so that one control cycle... The sum of the durations of all components within the function is still equal to ; The narrow pulse threshold The value is determined based on the dead time and minimum on time of the power switching devices in the three-level NPC converter, and is taken as 1.2 to 1.5 times the dead time.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention employs a virtual vector synthesis strategy to synthesize virtual small vectors by merging P-type / N-type redundant small vectors and virtual medium vectors by pairing adjacent medium vectors. This ensures that the net injection of all candidate virtual vectors into the midpoint potential is zero under the default allocation coefficients. Common-mode voltage suppression is achieved by pre-eliminating high common-mode vectors, and midpoint potential regulation is achieved by adjusting the duty cycle allocation coefficients of the virtual small vectors. The two control objectives act on different degrees of freedom, structurally decoupling the relationship and avoiding the dilemma of "sacrificing midpoint regulation capability to suppress common-mode" or "introducing high common-mode to balance the midpoint" in traditional schemes. By employing a linear slope to tighten current limiting during the voltage dip phase, using an adaptive active power recovery rate based on the midpoint potential during the voltage recovery phase, and using a ramp transition for the PI gain during operating condition switching, there are no discrete mode switching points throughout the entire process. This means that the rate of change of the control variable is bounded and will not undergo abrupt changes at any given moment, thereby avoiding problems such as secondary dips in the midpoint potential, electromagnetic torque impacts, and mechanical vibrations in the transmission chain, and providing protection for mechanical components such as bearings and gearboxes in the doubly-fed generator unit. All control components are based on deterministic mathematical operations: the 27 basic vectors are exhaustively enumerable, the common-mode voltage calculation formula is explicitly verifiable, the PI regulator parameters have clear physical meanings and constraints, the operating condition judgment thresholds are precisely measurable, and the virtual vector synthesis process is fully analytical. There are no black-box steps requiring manual parameter tuning, nor does it rely on data-driven model training. Any engineer can directly reproduce this solution based on its publicly available information and implement it in a real controller. The system identifies the current system state through a condition-aware module. Under normal operating conditions, a baseline PI gain and a wide limiting range are used to achieve a faster midpoint adjustment speed. Under midpoint imbalance conditions, the gain is increased and the limiting is narrowed to enhance the adjustment capability. Under fault ride-through conditions, the gain is further increased and the limiting is significantly narrowed to prioritize safety. Simultaneously, the pause tightening mechanism and adaptive recovery rate in continuous transient control enable the system to proactively derate under extreme conditions to maintain basic stability, rather than forcibly pursuing performance and causing runaway, thus significantly improving overall robustness. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a flowchart illustrating the screening and synthesis process of virtual vectors in the present invention. Figure 3 This is a connection diagram of the system modules of the present invention; Figure 4 This is the control quantity change curve for the fault-crossing continuous transition process of the present invention; Figure 5 This is a flowchart of the fault-crossing continuous transition control of the present invention. Detailed Implementation

[0018] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0019] Reference Figure 1 and 3 As shown, a three-level electrical drive chain control system for a doubly fed wind turbine includes five core modules: a condition sensing module 101, a virtual vector synthesis module 102, a midpoint balance adjustment module 103, a modulation execution module 104, and a continuous transition control module 105, as well as an optional narrow pulse processing module 106.

[0020] The connection relationships of each module are as follows: The input of the operating condition sensing module 101 is the real-time sampling signal of the three-level NPC converter, including the voltage of the upper bus capacitor. Lower bus capacitor voltage DC bus voltage Rotor three-phase current Grid connection point voltage and rotor position angle The output of the operating condition sensing module 101 is connected to the operating condition signal input of the virtual vector synthesis module 102, the operating condition signal input of the midpoint balance adjustment module 103, and the enable terminal of the continuous transition control module 105.

[0021] The output of the virtual vector synthesis module 102 is connected to the candidate vector input of the modulation execution module 104, providing the modulation execution module with the currently available set of candidate virtual vectors. The input of the midpoint balance adjustment module 103 is the midpoint potential deviation. The output terminal is connected to the duty cycle control input terminal of the modulation execution module 104, providing the duty cycle allocation coefficient to the modulation execution module. The output of the modulation execution module 104 is connected to the drive circuit of the three-level NPC converter to generate a PWM drive signal.

[0022] The input terminal of the continuous transition control module 105 is connected to the operating condition output of the operating condition sensing module 101, and the output terminal is connected to the gain control terminal of the midpoint balance adjustment module 103 and the current limiting control terminal of the modulation execution module 104, respectively. The narrow pulse processing module 106 is connected in series between the output terminal of the modulation execution module 104 and the converter drive circuit, and is used to correct extremely short pulses.

[0023] The overall workflow of the system is as follows: the operating condition sensing module 101 collects operating parameters in real time and calculates the midpoint potential deviation. and grid connection point voltage per unit value The current operating condition is determined. The virtual vector synthesis module 102 generates a candidate virtual vector set based on the candidate vector rules corresponding to the current operating condition. The midpoint balance adjustment module 103 calculates the duty cycle allocation coefficient based on the real-time midpoint potential deviation and the PI parameters corresponding to the current operating condition. The modulation execution module 104 selects the optimal vector from the candidate virtual vector set, and combines it with... The coefficient generates a PWM signal, which drives the converter after narrow pulse processing. When a fault ride-through condition is detected, the continuous transition control module 105 intervenes, adjusting the current limit and active power recovery rate according to the preset gradual change rules to ensure a smooth and shock-free transition process.

[0024] In this embodiment, each phase arm of the three-level NPC converter has three switching states: P, O, and N, corresponding to the upper transistor being on, the midpoint transistor being on, and the lower transistor being on, respectively. The three-phase combination generates... There are three basic space voltage vectors. The voltages of the three-phase output terminals corresponding to each vector relative to the DC side midpoint N are respectively... , , The common-mode voltage is defined as the average of the three values: The common-mode voltage was calculated for each of the 27 vectors, and the results can be categorized as follows: There are six large vectors: PNN, NPN, NNP, NPP, PPN, and PNP. Taking PPN as an example, phase A is connected to the positive terminal (...). B phase connected to the positive electrode ( C phase connected to negative electrode ( ), common mode voltage After verification, the absolute values ​​of the common-mode voltages of the six large vectors were all... The core characteristic of large vectors is the absence of redundant paired vectors, meaning there is no other switching state that can produce the same line voltage but with opposite common-mode polarity, thus making it impossible to cancel out common-mode components through synthesis.

[0025] There are 12 small vectors in total, divided into 6 pairs of redundant P-type / N-type small vectors: POO and ONN, OPO and NON, OOP and NNO, PPO and OON, POP and ONO, and OPP and NOO. Each pair of redundant vectors generates the same line voltage, but the midpoint current direction is opposite, and the common-mode voltage polarity is also opposite. The common-mode voltage of the P-type small vector is... The corresponding N-type redundant small vector common-mode voltage is Since the two have equal amplitudes and opposite polarities, they have the conditions to achieve common-mode cancellation through equal time allocation.

[0026] There are six mid-vectors: PON, OPN, NPO, NOP, ONP, and PNO. The common-mode voltage of a mid-vector is zero, making it the only type of non-zero vector with zero common-mode voltage. Two spatially adjacent mid-vectors (such as PON and PNO) have equal magnitude and opposite directions of midpoint current, and can be paired to synthesize a virtual mid-vector, achieving zero net injection of midpoint potential.

[0027] There are three zero vectors: PPP, NNN, and OOO. The common-mode voltages of PPP and NNN are respectively... and The amplitude is the largest and there are no redundant pairs; all three phases of OOO are connected to the neutral point, and the common-mode voltage is 0.

[0028] The vector selection principle of this invention is as follows: retain vectors that have redundant pairing or pair cancellation conditions and can achieve low common-mode voltage through synthesis, and discard vectors that cannot cancel common-mode voltage through synthesis. Accordingly, zero vectors PPP and NNN are selected because they have the largest common-mode voltage amplitude ( And because there were no redundant pairs, they were removed; although the common-mode voltage amplitude of the 6 large vectors was... (Comparable to small vectors), but there are no redundant paired vectors, so the common mode cannot be further reduced through synthesis and is therefore also removed; the remaining 19 basic vectors (including 6 pairs of redundant P-type / N-type small vectors, 6 medium vectors, and zero vector OOO) are retained to form a set of retained vectors, which serve as the basic material for subsequent virtual vector synthesis.

[0029] The selection principle for retaining vectors is to retain as many redundant degrees of freedom as possible while meeting the common-mode voltage suppression requirements, so that there is sufficient adjustment margin for subsequent midpoint potential regulation. Therefore, this scheme prioritizes retaining small vectors with redundant pairs (i.e., P-type / N-type small vector pairs), because these vectors are the main actuators for midpoint potential regulation.

[0030] In this embodiment, the method for synthesizing virtual small vectors is as follows: like Figure 2 As shown, after common-mode voltage calculation and redundancy judgment, the 27 basic space voltage vectors are divided into three categories: large vectors and zero vectors PPP / NNN are eliminated because they have no redundant pairs; 6 pairs of redundant small vectors (6 each of P-type and N-type) are all retained because they have the freedom to adjust the midpoint; and 6 medium vectors and zero vector OOO are retained because their common-mode voltage is zero. This results in a retained set of 19 basic vectors, which are further synthesized into virtual small vectors and virtual medium vectors, constructing a candidate virtual vector set which is then output to the modulation execution module.

[0031] Among these, the virtual small vector is the core method for decoupling the midpoint potential from the common-mode voltage in this scheme. Its basic idea is: within one control cycle... Inside, a pair of redundant P-type / N-type small vectors are applied sequentially. By adjusting the ratio of their application time, the average effect of their synthesis is spatially equivalent to a new virtual vector, while a continuously adjustable net injection amount can be generated at the midpoint current.

[0032] The POO / ONN redundant vector pair is used as an example for explanation. The P-type small vector POO corresponds to the switching state where phase A is connected to the positive terminal, phase B to the neutral point, and phase C to the neutral point. At this time, the neutral point current... That is, the current flowing out of the midpoint is equal to the current of phase A. The corresponding switching state for an N-type small vector ONN is phase A connected to the midpoint, phase B connected to the negative terminal, and phase C connected to the negative terminal. At this time, the midpoint current... That is, the current flowing into the midpoint is equal to the current of phase A.

[0033] It can be seen that for the same phase current When POO is applied, the midpoint current is (Flow out of the midpoint, making) reduce, (increase), the midpoint current is also when ONN is active. (Inflow to the midpoint, making) rise, (Lower). Therefore, and The direction is opposite to the change in midpoint potential, that is... The physical meaning of this relationship is that redundant P-type / N-type small vectors have equal magnitudes and opposite directions of effect on the midpoint potential, and naturally possess complementary characteristics.

[0034] In a control cycle Inside, let the duration of action of the small P-type vector be... The action time of the N-type small vector is ,in Let be the duty cycle allocation coefficient. Then the synthesized virtual small vector is: The spatial position of the composite vector lies between the P-type and N-type small vectors. Because their line voltages are identical, the amplitude of the composite vector is the same as that of the original small vectors, differing only in the distribution of their action time. The average midpoint current is: when hour, The virtual small vector has no net injection of midpoint potential. This is the default operating point, where the virtual small vector is only responsible for generating the required voltage vector and does not participate in midpoint regulation.

[0035] when hour, This generates a positive midpoint current, causing the upper capacitor to discharge and the lower capacitor to charge, thus lowering the midpoint potential. When hour, This generates a negative midpoint current, charging the upper capacitor and discharging the lower capacitor, causing the midpoint potential to rise. Through continuous adjustment... The value of can generate a continuously adjustable midpoint current to offset the midpoint potential shift caused by the load current.

[0036] Since the common-mode voltage of a virtual small vector depends only on the common-mode voltages of its P-type and N-type components multiplied by the time ratio, and the common-mode voltages of the P-type and N-type small vectors are equal (the same redundancy pair), therefore regardless of... The value of the virtual small vector's common-mode voltage remains unchanged. This means that the midpoint potential adjustment (adjustment) The common-mode voltage level and the common-mode voltage level (which depends on the selection threshold of the retained vector set) are two completely decoupled control degrees of freedom, which is the core advantage of this scheme.

[0037] In this embodiment, the middle vector is the three-level space vector with an amplitude of The vector located at the midpoint of each side of the hexagon. In traditional modulation schemes, the midpoint vector is an important component of the synthesized reference voltage. However, when the midpoint vector is applied, it generates a non-zero midpoint current, and this midpoint current changes with the phase of the load current. This is one of the main reasons for the low-frequency oscillation of the midpoint potential (the frequency is three times the output frequency).

[0038] This scheme adopts a virtual mid-vector strategy, which pairs two mid-vectors that are spatially adjacent and acts proportionally, so that the midpoint currents of the two cancel each other out.

[0039] like Figure 2 The virtual vector synthesis stage shown has six vectors arranged in spatial angular order as follows: PON, OPN, NPO, NOP, ONP, and PNO, which are evenly distributed at 60° intervals.

[0040] The midpoint currents of two adjacent midpoint vectors are equal in magnitude and opposite in direction. Taking spatially adjacent PON and PNO pairs as an example: In the PON state, phase A is connected to the positive terminal, phase B to the midpoint, and phase C to the negative terminal; the midpoint current... (Phase B current flows out from the midpoint). In PNO state, phase A is connected positive, phase B is connected negative, and phase C is connected to the midpoint. The midpoint current... (C-phase current flows into the midpoint). In a symmetrical three-phase system, when the reference voltage vector is located near the angle between the midpoint vectors, ,therefore .

[0041] More precisely, for any pair of adjacent midpoint vectors, if each acts for half a cycle within a control cycle, then the average midpoint current is: Under the condition of three-phase symmetrical steady-state operation and a power factor close to 1, the midpoint currents of adjacent mid-vectors are approximately equal in magnitude and opposite in direction, i.e. ,therefore The net injection of the virtual vector to the midpoint potential is close to zero.

[0042] The formula for the composition of vectors in a virtual system is: The synthesized virtual center vector lies spatially on the angle bisector of the two original center vectors, and its amplitude is close to that of the original center vectors. Since the common-mode voltage of both center vectors is zero (the sum of the three-phase voltages of the center vectors is zero), the common-mode voltage of the synthesized virtual center vector is also zero, and no additional common-mode component is introduced.

[0043] By employing a virtual mid-vector strategy, the third-harmonic oscillation component of the midpoint potential caused by the mid-vector is significantly suppressed, leaving only residual low-frequency components caused by factors such as load current asymmetry, thus reducing the workload of the midpoint balance adjustment module.

[0044] This embodiment describes the PI control implementation method of the midpoint balance adjustment module.

[0045] The midpoint balance adjustment module is the core of the closed-loop control of this system. Its function is to automatically adjust the duty cycle distribution coefficient of the virtual small vector based on the real-time detected midpoint potential deviation. This stabilizes the midpoint potential at the target value (i.e., the voltages of the two capacitors are equal). ).

[0046] The entire control loop adopts a classic single-loop negative feedback structure: the reference value is set to... The error signal is obtained by subtracting the actual midpoint potential from the sampled feedback, and then sent to the PI controller to calculate the original duty cycle. After limiting and anti-integral saturation processing, it is output to the virtual small vector synthesis module. The operating condition sensing module dynamically adjusts the PI gain and limiting range according to the current operating state to achieve adaptive control under multiple operating conditions.

[0047] The input to the PI controller is the midpoint potential deviation. The output is the original duty cycle allocation coefficient. : The physical meanings of each term in the formula are as follows: Base term 0.5: when At this point, the distribution coefficient is 0.5, the P-type and N-type small vectors each act for half a cycle, and the net current at the midpoint is zero. This is the system's equilibrium operating point. (Proportional term) Proportional control adjusts the distribution coefficient in real time based on the magnitude and direction of the current deviation, producing a regulating effect proportional to the deviation. Proportional gain. The larger the value, the faster the adjustment speed, but excessively large values ​​may cause oscillations. Integral term. Integral control aims to eliminate steady-state deviation. As long as a deviation exists, the integral term will continue to accumulate until the deviation reaches zero. Integral gain. The larger the value, the faster the steady-state deviation is eliminated, but an excessively large value may lead to overshoot.

[0048] PI regulator output Limiting is necessary for two reasons: First, the physical range of the duty cycle allocation coefficient is [0,1], and values ​​outside this range have no physical meaning. Second, it's to prevent narrow pulses, where the duration of a component is too short, causing the power switching device to turn off before it's fully turned on, increasing switching losses and electromagnetic interference. Therefore, the limiting range is set to... Typical value , .

[0049] The formula for limiting the amplitude is: To prevent integral saturation, the integrator stops accumulating when the output is limited. Specifically, in each control cycle, the sum of the proportional and integral terms is calculated first. Then determine if it is being bandwidth limited. If the current integral term is positive, then the integral term is frozen and positive deviations are no longer accumulated; if If the current integral term is negative, then the integral term is frozen, and negative deviations are no longer accumulated. This is the anti-integral saturation processing.

[0050] proportional gain The benchmark value The desired midpoint potential deviation can be estimated using the following formula: exist Eliminated within one control cycle, the maximum midpoint current that the virtual small vector can provide is: The capacitance is ,but ,in 1020 control cycles are used. Personnel in the relevant field can make minor adjustments based on this, including the integral gain. Desirable The integral time constant Take 1050 control cycles.

[0051] This embodiment describes the operating mode and parameter adaptive mechanism of the working condition sensing module.

[0052] The function of the operating condition sensing module is to automatically identify the type of operating condition based on the current operating status of the system and adjust the control parameters of other modules so that the system can achieve the optimal performance-safety trade-off under different operating conditions.

[0053] This scheme defines three operating conditions: normal operating condition, midpoint imbalance condition, and fault ride-through condition.

[0054] The criteria for determining normal operating conditions are as follows: and Under this operating condition, the grid voltage is normal, the midpoint potential deviation is within the allowable range, and the system operates with optimal performance as the target. At this time, the PI proportional gain is taken as the reference value. The duty cycle limit range is [0.2, 0.8], allowing for a larger adjustment range to achieve a faster response speed.

[0055] The criteria for determining the midpoint imbalance condition are: and Under this condition, the grid voltage is normal, but the midpoint potential deviation exceeds the normal range, requiring enhanced midpoint regulation capability. At this time, the PI proportional gain is increased to... This accelerates the adjustment speed; simultaneously, the duty cycle limit range is narrowed to [0.3, 0.7] to avoid overshooting due to excessive adjustment under large deviations. The combination of gain enhancement and limit narrowing increases the adjustment strength while limiting the maximum amplitude of a single adjustment, preventing oscillations.

[0056] The criteria for determining fault ride-through conditions are as follows: Under this operating condition, the grid voltage drops, and the system's primary objective is to ensure safe ride-through, not to pursue performance. At this point, the PI proportional gain is further increased to... Because the inrush current is large during a fault, the midpoint potential is easily disturbed, requiring stronger regulation capability; at the same time, the duty cycle limiting range is significantly narrowed to [0.35, 0.65], that is... The maximum deviation does not exceed 0.15, ensuring that the midpoint adjustment of the virtual small vector does not excessively consume the adjustment margin, and leaving sufficient safety margin to cope with unpredictable disturbances.

[0057] The operating condition determination uses a hysteresis comparison method to prevent frequent switching near the critical point. Specifically, the threshold for entering the midpoint imbalance condition is... The threshold for exiting the midpoint imbalance condition and returning to the normal condition is: There is a hysteresis band of 0.01 between the two. The entry threshold for the fault ride-through condition is... The exit threshold is It also retains hysteresis.

[0058] The parameter switching between the three operating conditions is not abrupt, but rather a ramp transition. The rate of change of the PI gain is limited to... Within, that is, from Upgraded to It takes at least 0.2 seconds. This gradual transition method avoids the control quantity jump caused by abrupt parameter changes, and is a manifestation of the "continuous transition" concept of this solution in the switching of normal operating conditions.

[0059] In this embodiment, the control strategy of the continuous transition control module during the voltage drop phase is as follows: When the grid connection point voltage drops and the system enters the fault ride-through condition, the most direct risk is an excessive rotor-side inrush current, which may cause overcurrent damage to power devices or DC bus overvoltage. Traditional solutions typically use direct limiting to suddenly restrict the current reference value to a lower level. This step-like limiting will cause a sudden change in control quantity at the moment of switching, triggering a series of secondary problems.

[0060] like Figure 4 and 5 As shown, the entire fault ride-through process is divided into three stages: current limiting adaptive tightening (drop transition), fault ride-through steady-state operation, and active power adaptive recovery (recovery transition). The three stages are smoothly connected to form a complete continuous transition control path.

[0061] This scheme adopts a linear slope tightening strategy, that is, the current limit value starts from the rated value and decreases linearly at a fixed slope until the target limit value is reached.

[0062] The specific implementation method is as follows: Detected The time is recorded as At this time, the rotor current limiting value .from start, Decrease according to the following formula: Among them, slope This means the current is tightened by 5 times the rated current per second. If the rated current is 1000A, then it tightens by 5000A per second. The target value for tightening the limit is... That is, when Down to The descent stops at this point, and the value is maintained until the fault ends. Based on the aforementioned slope, the duration of the tightening process is... That is, 100 milliseconds.

[0063] The 100-millisecond tightening time is fast enough to meet the response speed requirements of fault ride-through, yet slow enough to avoid the shock caused by sudden changes in control input. Compared to step-limiting, the rate of change of current changes from infinity to a finite value, significantly reducing electromagnetic torque shock and midpoint potential disturbance.

[0064] During the tightening process, a midpoint potential pause mechanism is implemented: if a pause is detected... If so, the tightening will be suspended, and the current situation will be maintained. Unchanged, until Then, tightening continues. The purpose of this mechanism is that when the midpoint potential has deviated significantly, it indicates that the system's current regulation capability is nearing its limit. If the current limit is tightened further, it may exacerbate the midpoint imbalance or even trigger protection actions. By pausing the tightening, time is given for the midpoint regulation module to pull the potential back to the normal range before the next tightening step can continue, ensuring that the entire process proceeds within a safety margin.

[0065] Simultaneously, during the drop phase, the PI proportional gain of the midpoint balance adjustment module switches to... The duty cycle limiting range is narrowed to [0.35, 0.65]. The gain increase enhances the response speed of the midpoint adjustment, while the narrowing of the limiting restricts the amplitude of a single adjustment. Together, these two factors ensure that the midpoint remains stable during faults.

[0066] In this embodiment, once the grid connection point voltage returns to normal, the system needs to smoothly transition from fault ride-through condition back to normal operation condition. The core issue during the recovery phase is controlling the active power recovery rate—too fast a recovery will cause drastic fluctuations in the DC bus and neutral point potentials, while too slow a recovery will not meet the grid connection standard's requirements for recovery time.

[0067] This scheme adopts an adaptive active power recovery rate strategy based on the midpoint potential. The recovery rate is not a fixed value but dynamically adjusted according to the degree of deviation of the current midpoint potential. A smaller midpoint potential deviation indicates a more stable system state, allowing for a faster recovery rate; a larger deviation indicates a heavier system regulation burden, requiring a slower recovery rate.

[0068] The formula for calculating the recovery rate is: The meanings of the two terms in the formula are as follows: the first term It is the basic recovery rate required by the grid connection standard, that is, in The active power will be restored from 0 to the rated value within a certain time. The typical range of values ​​is The specific value is determined by the applicable grid connection standard. The second term is the recovery rate of the midpoint potential constraint. When When the second term equals the first term; when the midpoint deviation is less than When the second term is less than the first term, the recovery rate is limited by the midpoint potential margin.

[0069] The min function selects the smaller of the two values ​​as the actual recovery rate, when the midpoint potential deviation is large (approaching or exceeding) The recovery rate automatically decreases to prioritize midpoint stability; when the midpoint potential deviation is small (much smaller than 100%), the recovery rate decreases automatically. The recovery rate is close to the standard requirement to meet the grid connection time requirement.

[0070] The specific steps of the recovery process are as follows: First, when When the grid voltage is restored, active power recovery is initiated. The initial active power is at the level during fault ride-through (usually zero or a very small value). The second step is to calculate the current recovery rate using the formula described above for each control cycle. Then update the active power reference value. : The third step, Gradually increase, while monitoring the midpoint potential deviation in real time. If If the value increases, the recovery rate will automatically decrease; if... If the rate decreases, the recovery rate will automatically increase, forming an implicit closed-loop regulation. Fourth step, when Restore to and Continue to exceed When the system is determined to have basically returned to normal, the PI proportional gain is then increased from... By slope (That is, a decrease of 2 times the reference gain per second) linearly reduced to Meanwhile, the duty cycle limit range is from Gradually relax to [0.2, 0.8]; Fifth step, when And the PI gain recovered to Afterwards, the system fully returns to normal operating conditions, and the continuous transition control module exits operation.

[0071] The entire recovery process involves no abrupt parameter switching, and the rate of change of all control variables is bounded, ensuring the smoothness of the transition process and the stability of the system.

[0072] In this embodiment, the virtual vector selection method of the modulation execution module is as follows: the function of the modulation execution module is to select the optimal virtual vector from the candidate virtual vector set to achieve the best rotor current tracking effect. This scheme adopts the idea of ​​finite control set model predictive control (FCS-MPC), which evaluates the tracking effect of each candidate vector one by one through the predictive model and selects the best one.

[0073] The prediction model is based on the discretized form of the rotor voltage equation. The voltage equation on the rotor side is: in For rotor voltage vector, The rotor current vector, Equivalent resistance For leakage inductance, This is the rotor back electromotive force.

[0074] Discretizing the above equation and using the forward Euler method, we obtain the... Predicted current value at time: in For the first candidate virtual vector set A virtual vector, For the first The measured current value at time [time]. For the first The estimated value of the back electromotive force at time t. The back electromotive force can be calculated from the stator voltage, stator current, and mutual inductance, or it can be estimated using an observer.

[0075] For each vector in the candidate virtual vector set The predicted current value is calculated using the formula described above. Then, each predicted value is compared with the current reference value. The square of the error between: Choose to The smallest virtual vector is taken as the optimal output vector for this control cycle.

[0076] The composition of the candidate virtual vector set depends on the spatial angular position of the reference voltage vector. The virtual vector synthesis module selects virtual vectors from the angular interval where the reference vector is located and its adjacent intervals to enter the candidate set, thereby reducing the number of candidate vectors and the computational load. Taking one angular interval as an example, the candidate virtual vectors include: small virtual vectors within that interval. (Synthesized from corresponding P-type / N-type redundant pairs), virtual mid-vectors on adjacent boundaries (Synthesized by proportional pairing of two adjacent median vectors), virtual small vectors in the next adjacent interval (Synthesized from another set of redundant pairs), and zero vector OOO, for a total of 4 candidate virtual vectors, with relatively low computational cost.

[0077] For finer control, the number of candidate vectors in each angle interval can be increased, but this will also increase the computational load. In practical engineering, 4 to 6 candidate vectors are usually sufficient to achieve good current tracking performance.

[0078] After the optimal vector is selected, the duty cycle allocation coefficient output by the midpoint balance adjustment module is used. The virtual vector is decomposed into the corresponding basic vector sequence and the action time, and then a three-phase PWM drive signal is generated and output to the drive circuit of the converter.

[0079] In this embodiment, the narrow pulse processing module is implemented as follows: In the actual operation of a three-level NPC converter, power switching devices (such as IGBTs) have limitations on minimum on-time and dead-time. If the duration of a certain switching state is too short (usually less than 1-2 microseconds), the device may not be able to fully turn on or off, leading to increased switching losses, aggravated electromagnetic interference, and even damage to the device.

[0080] The function of the narrow pulse processing module is to detect and correct excessively short pulses, ensuring that the duration of all switching states is not less than the minimum allowable value of the device.

[0081] The specific implementation method is as follows: In each control cycle, after the modulation execution module determines the duration of all basic vectors within that cycle, the narrow pulse processing module checks the duration of each basic vector one by one. If the duration of action of a certain vector If so, it is considered a narrow pulse and needs correction.

[0082] The correction method is to change the duration of the vector's action from... Forced to increase This means supplementing to the minimum pulse width. The supplementary time needs to be subtracted from the action time of other vectors within the same cycle to ensure the total duration of the entire control cycle. Unchanged. The deduction ratio is allocated according to the proportion of the original action time of each of the other vectors: in, This represents all vectors other than the narrow pulse vector. In this way, the duration of each non-narrow pulse vector is reduced proportionally, maintaining their relative proportions and thus minimizing their impact on the modulation effect.

[0083] If multiple narrow pulses occur simultaneously within a period, they are processed one by one: first, the narrowest one is processed, corrected, and the duration of each vector is recalculated. Then, it is checked whether there are any more narrow pulses, and so on, until the duration of all vectors is no less than [a certain value]. .

[0084] Narrow pulse threshold The value needs to be determined based on the characteristics of the power device. Generally speaking, It should be greater than the sum of the dead time and the minimum on-time. In engineering practice, it is typically taken as 1.2 to 1.5 times the dead time. For example, if the dead time is 2 microseconds, then... The time interval can be 2.4 to 3.0 microseconds.

[0085] Narrow pulse processing introduces a certain voltage error because the actual applied vector sequence is not entirely consistent with the ideal sequence. However, due to... Relative to control cycle It is very small (usually) The introduced error is within an acceptable range. In contrast, the risk of device damage caused by narrow pulses is much more serious, therefore narrow pulse processing is an essential protective measure.

[0086] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A three-level electrical drive chain control system for a doubly-fed wind turbine generator, characterized by, include: The operating condition sensing module is used to collect the operating parameters of the three-level NPC converter of the doubly-fed wind turbine in real time and construct feature vectors. in, The midpoint potential deviation is the upper bus capacitance. Voltage and lower bus capacitance voltage difference DC bus voltage, For rotor electrical frequency, For modulation ratio, The voltage per unit value at the grid connection point is used to determine the current operating condition based on the feature vector. The virtual vector synthesis module, connected to the operating condition sensing module, is used to remove the common-mode voltage amplitude from the 27 basic space voltage vectors of the three-level NPC converter. The vectors are used to obtain a set of retained vectors. Then, based on the set of retained vectors, virtual small vectors and virtual mid-vectors with zero net injection of midpoint potential are constructed to form a candidate virtual vector set. The midpoint balance adjustment module, connected to the virtual vector synthesis module, is used to adjust the midpoint potential based on the real-time midpoint potential deviation. Calculate the duty cycle allocation coefficients for P-type and N-type mini-vectors in the virtual mini-vectors. This ensures that the mean current of the virtual small vector is zero within the control cycle, while feedback correction counteracts the disturbance of the load current to the midpoint potential. The modulation execution module, connected to the virtual vector synthesis module and the midpoint balance adjustment module, is used to select the optimal virtual vector from the candidate virtual vector set and allocate it based on the duty cycle coefficient. Generate the PWM drive signal for the three-level NPC converter; The continuous transition control module, connected to the operating condition sensing module and the modulation execution module, is used to adaptively adjust the rotor current limiting value and active power recovery rate during the voltage dip and recovery process at the grid connection point, so that the midpoint potential deviation is maintained at a constant value during the transition process. Within.

2. The three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 1, characterized in that: The virtual vector synthesis module removes common-mode voltage amplitude. The vector, specifically: For the 27 basic space voltage vectors, the common-mode voltage of each vector is calculated using the following formula: in , , These are the voltages of the three-phase output terminals A, B, and C corresponding to the vector, relative to the midpoint of the DC side. Eliminate satisfied The six large vectors PPN, PNP, PNN, NPP, NPN, and NNP, and the common-mode voltage amplitude equal to The zero vectors PPP and NNN are retained, and the remaining 19 basic vectors constitute the reserved vector set, which includes 6 pairs of redundant P-type / N-type small vectors (POO and ONN, OPO and NON, OOP and NNO, PPO and OON, POP and ONO, OPP and NOO), 6 medium vectors (PON, OPN, NPO, NOP, ONP, PNO), and 1 zero vector OOO.

3. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 2, characterized in that: The virtual vector synthesis module constructs virtual mini-vectors, specifically as follows: For each pair of redundant P-type / N-type small vectors in the retained vector set The virtual small vector is synthesized according to the following formula: in The duty cycle allocation coefficient for the P-type small vector is output by the midpoint balance adjustment module; The direction of the midpoint current corresponding to the P-type small vector is: The direction of the midpoint current corresponding to the N-type small vector is and The virtual small vector in one control cycle The average midpoint current within is: when hour, The net injection of the virtual small vector to the midpoint potential is zero.

4. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 2, characterized in that: The virtual vector synthesis module constructs a virtual vector, specifically as follows: For the six median vectors in the retained vector set, they are paired according to the principle of spatial proximity and opposite midpoint current directions. The paired median vectors are then combined into a virtual median vector using the following formula: in and For two paired vectors, within a preset allowable deviation range, they satisfy... ; Virtual vector in one control cycle The average midpoint current within is: This makes the net injection of the midpoint potential by the mid-vector zero, fundamentally eliminating the low-frequency oscillation of the midpoint potential caused by the mid-vector.

5. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 3, characterized in that: The midpoint balance adjustment module calculates the duty cycle allocation coefficient. Specifically: Midpoint potential deviation As input, the original duty cycle allocation coefficient is output via the PI controller. : in For proportional gain, For integral gain, the reference value of 0.5 corresponds to the symmetrical distribution when the midpoint potential is balanced; right Apply amplitude limiting constraints: in , To prevent narrow pulses caused by an excessively small duty cycle allocation factor; when Limited to or At this point, the integrator stops accumulating and performs anti-integral saturation processing.

6. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 1, characterized in that: The operating condition sensing module determines the current operating condition as follows: According to the feature vector The following conditions shall be used to determine the outcome: when and When the condition is determined to be normal operating condition, all virtual vectors in the candidate virtual vector set can be selected, and the PI proportional gain of the midpoint balance adjustment module is set to... ; when and When the condition is determined to be a midpoint imbalance, the PI proportional gain of the midpoint balance adjustment module is set to... At the same time, the duty cycle allocation coefficient limit range is narrowed to ; when When the fault crossing condition is determined, the continuous transition control module is activated.

7. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 6, characterized in that: The control of the continuous transition control module during the voltage drop process at the grid connection point is specifically as follows: Detected At that time, rotor current limiting value From rated current By slope linear tightening to No step switching is used during the tightening process; At the same time, the PI proportional gain of the midpoint balance adjustment module is switched to The duty cycle allocation coefficient limit range is reduced to ; If the rotor current limiting value is tightened during the process, if it is detected If so, the tightening will be suspended, and the current situation will be maintained. Unchanged, until Then it continued to tighten.

8. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 7, characterized in that: The control of the continuous transition control module during the grid connection point voltage recovery process is specifically as follows: Detected Subsequently, the active power recovery rate is adaptively determined by the following formula: in Rated active power, The recovery time constant is the value required by the grid connection standard, and its range is [value range missing]. ; During the recovery process At that time, the active power recovery rate automatically decreases to Ensure that the midpoint potential deviation does not exceed the limit; When active power recovers to and Continue to exceed At that time, the PI proportional gain of the midpoint balance adjustment module is from By slope Restore to It completes the continuous transition from fault-crossing conditions to normal operating conditions.

9. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 1, characterized in that: The modulation execution module selects the optimal virtual vector, specifically as follows: According to the rotor current reference value and the current measured value of rotor current Calculate the current error ; In the candidate virtual vector set, each virtual vector is calculated one by one. Acting on a control cycle Predicted current value : in, For leakage inductance, For equivalent resistance, This is the rotor back electromotive force; Choose the current tracking cost function The smallest virtual vector is output as the optimal virtual vector.

10. A three-level electrical drive chain control system for a doubly-fed wind turbine generator according to claim 3, characterized in that: The system also includes a narrow pulse processing module, connected to the modulation execution module, specifically: When the actual action time of the P-type or N-type component of the virtual small vector Less than the preset narrow pulse threshold At that time, the duration of action of this component is forcibly set to... At the same time, the duration of action of the remaining components within the same cycle is reduced proportionally, so that one control cycle... The sum of the durations of all components within the function is still equal to ; The narrow pulse threshold The value is determined based on the dead time and minimum on time of the power switching devices in the three-level NPC converter, and is taken as 1.2 to 1.5 times the dead time.

Images