Control method for suppressing power grid resonance through photovoltaic cluster impedance remodeling

Abstract

The invention discloses a control method for suppressing power grid resonance through photovoltaic cluster impedance remodeling, and belongs to the technical field of photovoltaic grid-connected control. According to the method, small time disturbance is introduced into an inverter sampling control layer, so that the sampling moments of different inverters are not synchronized any more in the statistical sense, and the time coherence among multiple machines is broken. The phase difference of the output impedance of each inverter forms a dispersion effect in the frequency domain, so that soft remodeling of cluster equivalent impedance is realized, the resonance peak value is effectively weakened, and the resonance energy is dispersed. The system further adopts a fractional delay shaping and sideband power feedback mechanism to carry out adaptive adjustment on a sampling disturbance range and control parameters so as to maintain stable margin and dynamic performance. Meanwhile, a safe backspacing mechanism is set, and standard control is automatically recovered under the abnormal working condition. According to the method, a physical damping device does not need to be added, energy loss is avoided, the method has the advantages of being simple in structure, high in compatibility, capable of achieving industrialization and the like, and the stability of the photovoltaic cluster grid-connected system can be remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of new energy power electronics and photovoltaic grid-connected control technology, specifically a control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters. Background Technology

[0002] In recent years, with the rapid development of new energy power generation, the installed capacity of photovoltaic grid-connected systems has continued to climb, and photovoltaic inverters have become one of the most widely distributed and numerous power electronic devices in the power system. Multiple photovoltaic inverters are connected to the grid in parallel through a common combiner bus to form a photovoltaic cluster grid-connected structure. This parallel operation mode can effectively improve power density and system redundancy, but it also introduces complex impedance coupling and stability issues.

[0003] In grid-connected photovoltaic (PV) systems, inverters typically employ a current-controlled topology, with an LCL filter on the output side to suppress high-frequency harmonics. The inverter controller tracks the grid voltage via a phase-locked loop (PLL), achieving coordinated control between the outer current loop and the inner voltage loop. Since the control parameters, modulation frequencies, and sampling timings of each inverter are essentially identical, the output impedance characteristics of different inverters in the system are similar, maintaining high coherence in the frequency domain. When the grid impedance matches the inverter output impedance within a specific frequency range, grid resonance can easily occur, manifesting as distortion of the grid connection point voltage and current waveforms, harmonic amplification, and even control oscillations. This phenomenon is particularly pronounced in weak grid environments, becoming one of the main constraints on the stable operation of grid-connected PV systems.

[0004] To address the resonance problem in photovoltaic systems, existing technologies mainly focus on two directions: one is to introduce damping by changing the physical structure; the other is to adjust the system impedance by optimizing the control strategy.

[0005] The first type of method is usually called physical damping or active damping. Its implementation includes connecting a resistor in series in the filter branch, using a passive damping network, or introducing a virtual resistor in the control loop to simulate damping characteristics. This type of method can reduce the system's resonant peak value to some extent, but it suffers from problems such as energy loss, increased heat generation, and decreased system efficiency. Furthermore, the damping parameters are fixed, making it difficult to adapt to complex operating conditions.

[0006] The second type of method belongs to controlled impedance compensation technology, which improves impedance characteristics by adding current feedback, virtual impedance, or damping control loops to the inverter controller. This method can dynamically adjust system damping within a certain range to avoid physical energy loss, but its performance depends on accurate modeling of system parameters. If the grid impedance or control delay changes, it can easily lead to compensation mismatch and trigger new oscillations. In addition, there are still sampling and control synchronization problems between multi-machine systems. The impedances of multiple inverters often maintain the same phase in the frequency domain, thus forming a group resonance effect.

[0007] In recent years, some studies have proposed using master-slave control, centralized coordination control, or communication synchronization mechanisms to improve the stability of multi-machine systems. However, these methods require additional communication networks and master-slave management structures, resulting in high system complexity and difficulty in guaranteeing real-time performance and reliability. When the system scales up or communication latency increases, the stability problem still cannot be fundamentally solved.

[0008] In summary, existing grid-connected photovoltaic clusters still have the following shortcomings in terms of resonance suppression: (1) It relies on physical damping or additional circuits, resulting in high energy loss and low efficiency; (2) The control compensation-based method is parameter-sensitive and highly dependent on the model; (3) The multi-machine system has high synchronization and is easy to form coherent resonance; (4) Communication-based coordination methods have complex structures and poor scalability.

[0009] Therefore, there is an urgent need for a resonance suppression method that can achieve dynamic impedance adjustment at the control layer without adding physical damping and is applicable to multi-machine systems, so as to achieve stable operation of photovoltaic clusters in complex power grid environments. Summary of the Invention

[0010] The purpose of this invention is to provide a control method for suppressing grid resonance by impedance reshaping in photovoltaic clusters. This method does not require the addition of physical damping devices, has no energy loss, and has the advantages of simple structure, strong compatibility, and industrial implementation, which can significantly improve the stability of photovoltaic cluster grid-connected systems.

[0011] The technical solution adopted in this invention is as follows: A control method for suppressing grid resonance by impedance reshaping in photovoltaic clusters is applicable to cluster systems comprising several grid-connected photovoltaic inverters. Each inverter includes a main controller, a sampling module, a PWM drive module, and grid-connected current and voltage measurement channels. The method is executed by an impedance reshaping control unit installed in the main controller of each inverter. The control unit includes a sampling-triggered random scheduling module, a fractional delay shaping module, a sideband power monitoring module, and a safety back-off module. The method includes the following steps: (1) Sampling random scheduling: In each PWM cycle, the sampling-triggered random scheduling module generates a set of sampling phase offset sequences {Δt(k)} based on the preset window length M and the maximum sampling offset Δt_max, satisfying the following constraints: ; where: Δt(k) represents the phase offset of the sampling trigger moment relative to the standard sampling moment in the k-th PWM period; M represents the number of PWM periods included in the statistical window; Δt_max represents the maximum allowable sampling phase offset; this sequence is generated by a low-pass filtered pseudo-random sequence, making the sampling phase change have the characteristics of zero mean, zero window sum, and bandwidth limitation, and adjusting the sampling moments of current and voltage in real time at the ADC trigger layer, so that the sampling time bases between inverters are statistically uncorrelated, reducing the group sampling coherence; (2) Measurement signal fractional delay shaping: The fractional delay shaping module calculates the fractional delay coefficient α(k) according to the current sampling offset Δt(k), and its definition is: ; where: α(k) represents the fractional delay coefficient of the current sampling period; Ts represents the time length of the PWM and sampling periods; the fractional delay shaping process is performed on the measurement signal x[n], and the calculation formula is: <{ ; where: x[n] represents the original measurement signal value (current or voltage) obtained in the current sampling period; x[n - 1] represents the measurement signal value in the previous sampling period; y[n] represents the signal output after fractional delay shaping; through the above processing, the linear influence of the sampling moment jitter on the phase characteristic of the output impedance is cancelled, and the output impedance of the inverter is stabilized; (3) Sideband power monitoring and parameter adaptive adjustment: The sideband power monitoring module calculates the sideband power index BP of the grid-connected point voltage or current in the resonance-sensitive frequency band in real time, and its definition is: ; where: P_band represents the effective power component of the signal in the 200 - 800 Hz frequency band; P_fund represents the effective power component at the fundamental frequency; BP represents the sideband power ratio, which is used to characterize the resonance energy level; when BP > BP_H (upper threshold), the controller automatically reduces Δt_max and reduces the bandwidth of the random sequence; when BP < BP_L (lower threshold), Δt_max and the bandwidth are gradually restored to the default parameters to achieve the adaptive suppression of sideband energy; (4) Safety fallback control: The safety fallback module continuously monitors the following parameters: sampling setup time t_setup; dead zone protection time t_dead; overcurrent status signal l_over; out-of-step and low voltage ride-through event flag S_fault. When it is detected that t_setup < t_setup_min, or there is a risk of dead zone crossing, or l_over or S_fault is triggered, the sampling random scheduling module and the fractional delay shaping module are immediately bypassed, and the system is restored to the conventional fixed-time sampling and control mode to ensure the system safety and grid connection stability.

[0012] The sampled phase bias sequence Δt(k) is output by a pseudo-random number generation function and processed by a finite impulse response (FIR) low-pass filter, with its cutoff frequency fc located in the range of 200 to 500 Hz.

[0013] Different inverters use different pseudo-random seeds and initial phase offsets during power-on initialization to ensure that the sampling time base of each inverter is statistically independent.

[0014] The fractional delay coefficient α(k) is filtered by a first-order IIR low-pass filter to limit its rate of change to no more than 500Hz, thus preventing high-frequency oscillations in the control system.

[0015] The sideband power BP is calculated by using a bandpass filter to extract the power components in the 200-800Hz frequency band, with BP_H and BP_L set to default thresholds of 0.08 and 0.04, respectively.

[0016] The safety rollback module includes a hardware limit detection submodule and an event detection submodule. The former is used to monitor whether t_setup, t_dead, and the ADC channel phase interval Δφ meet the minimum safety margin, while the latter is used to detect overcurrent, loss of synchronization, low voltage ride-through, or communication anomalies. If any condition is triggered, bypass will be executed.

[0017] The impedance reshaping control unit is deployed between the timer triggering layer and the ADC interrupt management layer of the inverter main controller. Sampling random scheduling and fractional delay shaping are both completed before the main control algorithm is executed.

[0018] This method is also applicable to multi-machine parallel systems containing energy storage converters, wind power inverters, or flexible DC converter units, and is used to improve their output impedance stability and resonance suppression performance.

[0019] The beneficial effects of this invention are as follows: Compared with existing technologies, the control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters proposed in this invention shows significant comprehensive advantages in terms of system stability, control efficiency, structural complexity, and industrial feasibility.

[0020] In existing grid-connected photovoltaic systems, multiple inverters typically employ a synchronous sampling and unified control strategy. This high synchronization results in the output signals of each inverter being completely synchronized. When the grid impedance and the inverter output impedance resonate at a certain frequency, the output disturbances of multiple inverters will superimpose, generating a significant resonance amplification effect in the system. Traditional vibration suppression methods mainly rely on adding physical damping elements to filters or introducing virtual damping in the control loop. While these methods can weaken resonance to some extent, they often lead to energy loss, increased heat load, and decreased system efficiency. Furthermore, coherence issues still exist between multi-inverter systems, making it impossible to fundamentally eliminate group resonance.

[0021] The innovation of this invention lies in its approach: instead of adjusting impedance by adding hardware or changing circuit parameters, it reconstructs the time and frequency domain characteristics through a control strategy. By introducing small and random time disturbances into the sampling control, the sampling times of different inverters are no longer statistically identical, thus disrupting the original coherence. Due to the different time distributions of the responses of each inverter, their output impedances exhibit subtle phase dispersion in the frequency domain. The equivalent impedance of the entire photovoltaic cluster changes from synchronous superposition to incoherent superposition, effectively weakening the system's resonance peak and distributing the resonance energy more evenly across a wider frequency band, significantly suppressing grid resonance phenomena.

[0022] Building upon this foundation, the present invention further incorporates a fractional delay shaping compensation module to correct the phase deviation introduced by sampling time disturbances, ensuring that the accuracy and stability of the control loop remain unaffected. Simultaneously, through sideband power monitoring and adaptive adjustment mechanisms, the system can automatically optimize the disturbance range and delay compensation parameters based on changes in the resonant state, enabling the impedance reshaping process to possess adaptive characteristics. This design achieves dynamic adjustment and real-time optimization of the control system, ensuring that the resonance suppression effect remains stable under various operating conditions.

[0023] Furthermore, this invention establishes a safe fallback mechanism. In the event of communication anomalies, overcurrent, or loss of synchronization, the system can automatically bypass and return to the standard control state, ensuring the safe operation of the entire photovoltaic cluster under any operating condition. This design makes the control strategy both innovative and balanced with industrial-grade reliability and engineering application operability.

[0024] In practical terms, the control method of this invention significantly improves the stability of photovoltaic clusters while maintaining the original system efficiency. Experimental results show that this method can effectively reduce harmonic energy at the grid connection point, expand the system's stability margin, and avoid energy loss and structural complexity issues caused by physical damping. Because this invention employs a software-based implementation, all control functions can be implemented on existing inverter control platforms via firmware upgrades, without changing the hardware topology or adding additional components. Therefore, it has extremely high compatibility and widespread application value.

[0025] In summary, this invention establishes a soft impedance reshaping mechanism based on a control strategy by introducing perturbations in the time domain and achieving impedance phase dispersion in the frequency domain. This mechanism does not rely on changes in physical topology; instead, it utilizes control algorithms to reshape the dynamic characteristics of the system at the signal level, enabling photovoltaic clusters to maintain high efficiency and stability under complex grid conditions. Compared with traditional methods, this invention achieves lossless, adjustable, easily implemented, and highly stable resonance suppression, providing a low-cost, high-reliability control solution for large-scale photovoltaic grid-connected systems, demonstrating outstanding technological innovation and application value. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the impedance reshaping control unit of the present invention; Figure 2 This is a schematic diagram of the signal interaction logic and data flow of the present invention; Figure 3 This is a flowchart illustrating the working process of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0028] See Figure 1 In systems with multiple photovoltaic inverters operating in parallel, each inverter typically uses a unified PWMIO cycle and synchronous sampling time base. This fully synchronous sampling method can easily lead to the coherent superposition of output impedances across multiple inverters under weak grid conditions, resulting in systemic resonance. Therefore, a control method for suppressing grid resonance through impedance reshaping in photovoltaic clusters is applicable to cluster systems comprising several grid-connected photovoltaic inverters. Each inverter includes a main controller, a sampling module, a PWM drive module, and grid-connected current and voltage measurement channels. This method is executed by an impedance reshaping control unit installed in the main controller of each inverter. The control unit includes a sampling trigger random scheduling module, a fractional delay shaping module, a sideband power monitoring module, and a safety backoff module. The impedance reshaping control unit is deployed between the timer trigger layer and the ADC interrupt management layer of the inverter main controller. Sampling random scheduling and fractional delay shaping are both completed before the main control algorithm is executed. By introducing controlled random disturbances at the sampling trigger time, the group coherence is broken. Simultaneously, in conjunction with delay shaping and power monitoring, dynamic detection and adaptive suppression of system resonance are achieved. This method includes the following steps: S1. First, the sampling-triggered random scheduling module generates a sampling phase offset sequence within each PWM cycle. Its value satisfies ; in, M represents the phase offset (in seconds) relative to the standard sampling time when the sampling is triggered in the k-th PVMI cycle, and M is the statistical window length (typically 32). This represents the maximum allowed offset (typically 1.5 μs). Sequence Generated by a pseudo-random function and then processed by a finite-bandwidth FIR low-pass filter, the filter coefficients... satisfy Cutoff frequency This results in a smooth, zero-mean random disturbance signal. In each PWM cycle, the controller... Adjusting the ADC sampling trigger timing ensures that the sampling time bases of each inverter are statistically uncorrelated. This effectively reduces the output impedance coherence peaks caused by multi-inverter sampling synchronization, thereby suppressing resonance at the group level.

[0029] In traditional photovoltaic (PV) cluster systems, the sampling and control updates of all inverters are typically triggered by the same PWM clock. This means that multiple inverters sample the grid voltage and current and calculate control variables at exactly the same time. Because their control laws, parameters, and sampling phases are completely synchronized, the dynamic response of the entire cluster under external disturbances is highly coherent. When the grid impedance... With inverter output impedance When a resonance relationship occurs in a certain frequency range, the superposition effect of current disturbances from multiple inverters will cause the resonance peak to be amplified sharply, resulting in a sharp resonance peak in the overall impedance curve of the system, and even triggering oscillation.

[0030] This invention introduces controlled random phase perturbation during the sampling triggering process. This random offset causes statistically independent, minute differences (on the order of ±1 μs) in the sampling times of each inverter. This random offset prevents the responses of each inverter from being completely in phase in the time domain, thus weakening the coherent superposition of current disturbances. In the frequency domain, this is equivalent to applying random phase diffusion to the originally coherently superimposed impedance spectrum, transforming the cluster's equivalent impedance from a "narrow-band peak type" to a "wideband smooth type," forming a chain effect of time-domain desynchronization, frequency-domain broadening, impedance smoothing, and resonance suppression. Therefore, the core mechanism of the sampling random scheduling module is to break synchronicity in the time domain and weaken resonance accumulation in the frequency domain. Experiments show that after adopting this method, the system resonance amplitude is reduced by 8-10 dB, and the phase margin is increased by approximately 20°.

[0031] S2. Secondly, considering that random sampling offset may cause signal phase drift, this invention sets up a fractional delay shaping module in the signal preprocessing layer to counteract the phase effect of sampling offset. The fractional delay coefficient is calculated based on the sampling offset. ; And perform linear interpolation shaping on the original sampled signal: ; in, The sampling period is typically 50 μs. This is the signal for the current sampling period. This is the signal from the previous cycle. Output is integer. This algorithm is equivalent to a delay operator. The first-order approximation can eliminate phase errors caused by sampling jitter across the frequency range. Since it involves only two multiplications and one addition, the calculation time is less than 0.1 μs, having no impact on the real-time performance of the main control. After shaping, the phase of the sampled signals from each inverter remains stable, and the system output impedance curve is continuous and smooth.

[0032] Because the sampling time is randomly offset, there is a slight phase difference between the voltage and current signals in different inverters. Without compensation, this difference introduces additional delay in the current or voltage loop control, affecting system accuracy. The mechanism of the fractional delay shaping module is to perform linear interpolation approximation compensation on the sampled signal by calculating the sampling offset Δt(k) in real time, that is, to reconstruct an approximate value of the signal at the "ideal sampling time" in the digital domain. Its mathematical equivalent operation is a continuous-time delay operator. First-order expansion: ; This cancels out the additional phase shift introduced by sampling phase drift in the frequency response. As a result, although the sampling trigger times of each inverter are different, their control input signals are re-aligned in phase, ensuring that the closed-loop output impedance of each inverter remains consistent. Therefore, random disturbances disrupt coherence, while delay shaping restores phase consistency; the two balance each other, achieving the design goal of "desynchronization without instability".

[0033] S3. Subsequently, the sideband power monitoring and adaptive adjustment module evaluates the system resonant energy in real time. This module uses the grid-connected current... As input, the fundamental and resonant frequency band components are extracted using low-pass and band-pass filters, respectively: ; in, This is a low-pass filter operator with a cutoff frequency of 50Hz. This is a bandpass filter operator for the 200–800 Hz range. The filtering results are used to calculate the sideband power ratio. ; Where RMS() represents the root mean square function, and BP is the dimensionless sideband power ratio. If BP > (Upper limit threshold, typical value 0.08) indicates enhanced system resonance; the controller automatically reduces the amplitude of random disturbances and decreases the filter bandwidth. ; If BP < (Lower threshold, typical value 0.04) and lasts for 5 seconds, then the system restores default parameters: ; Through this adaptive closed loop, the system automatically converges the disturbance when resonance is enhanced, and maintains randomness in the steady state, thus ensuring a balance between vibration suppression and stability.

[0034] When the system resonance intensifies, distinct sideband frequency components appear on both sides of the fundamental frequency in the current spectrum, with energy concentrated in the 200–800 Hz band. Therefore, sideband power can be directly used as a measure of the resonance degree. The sideband power monitoring module uses a bandpass filter to extract the signal in this frequency band in real time and calculates the sideband power. and fundamental power The comparison forms a ratio: ; When BP increases, it indicates that the resonant energy is enhanced; when BP decreases, the system tends to stabilize.

[0035] Based on this metric, the present invention constructs an adaptive adjustment mechanism to adjust the random disturbance parameters. With filter cutoff frequency The damping effect changes dynamically with the back pressure (BP). When BP is high, the system automatically reduces the amplitude of random disturbances and slows down the rate of phase change to enhance the damping effect; when BP is low, the disturbance amplitude is moderately increased to maintain the dispersion of the cluster impedance. This energy-driven closed-loop regulation enables the system to self-sense the resonance trend and autonomously converge to the stable operating point without relying on an external model. Therefore, the entire method realizes a negative feedback vibration suppression mechanism between "resonance energy and disturbance intensity" in the energy domain. Experiments show that this strategy can reduce the grid-connected current sideband power by 8.5 dB and increase the resonance threshold by 25%.

[0036] S4. Finally, to ensure the safe operation of the control system under abnormal conditions, this invention incorporates a safety fallback mechanism to monitor critical operating states in real time. Monitoring parameters include: ADC sampling setup time. PWM dead time protection time Overcurrent signal Loss of step or low voltage ride-through indicator and communication timeout signal When any anomaly is detected, such as <1μs, sampling triggers intrusion into dead zone, or When =1, the system immediately bypasses the impedance reshaping control unit and forces it to be set. =0, =0, restoring the inverter to standard timed sampling mode. Abnormal states are recorded in the controller log. When all monitored parameters return to normal for 5 consecutive seconds, the system automatically de-circuits and resumes normal operation. The entire rollback and recovery process is completed within one sampling cycle, with a switching time of less than 50μs, and causes no disturbance to the output current.

[0037] In photovoltaic grid-connected systems, events such as voltage dips, communication anomalies, or sampling setup failures can lead to control loop interruptions. If the impedance reshaping algorithm continues to execute under these conditions, random disturbances and time-delay shaping may introduce uncertainties, jeopardizing system stability. To address this, this invention establishes a safety backoff mechanism that monitors the system's operating status in real time: when any abnormal signal is detected (such as overcurrent, sampling synchronization failure, or communication timeout), the impedance reshaping control unit is immediately bypassed, restoring the sampling mode to standard synchronous sampling (Δt=0, α=0). This process takes only one control cycle (approximately 50μs), ensuring the system instantly returns to a normal stable control state. Once all abnormal signals return to normal and remain so for 5 seconds, the system automatically unblocks and smoothly recovers to the impedance reshaping mode. This mechanism ensures that the controller has a safe backoff path in all situations, achieving a self-healing characteristic of "immediate fault bypass and automatic recovery," guaranteeing reliability in industrial applications.

[0038] See Figure 2 The signal interaction logic and data flow are as follows: The grid voltage and current signals are sampled by sensors and input to the shaping module via the main controller; The random scheduling module generates a sampling offset Δti, triggers ADC sampling, and transmits the offset data to the shaping module. After performing fractional delay compensation on the sampled signal, the shaping module outputs a correction signal to the main control loop; The main control loop performs current and voltage dual-loop calculations and outputs PWM control signals to the inverter power drive unit. The grid-connected current output by the inverter is fed back to the sideband power monitoring module; After the sideband power monitoring module extracts the resonance information, it sends it to the adaptive adjustment module. The adaptive module updates the control parameters and feeds them back to the random scheduling module and the shaping module; The entire process is monitored in real time by the safety rollback module, with bypass control available when necessary. The communication synchronization module maintains parameter consistency and data exchange among multiple machines.

[0039] This signal chain forms a closed-loop control path from "disturbance-measurement-adjustment-protection", enabling real-time, dynamic, and self-healing impedance reshaping control.

[0040] See Figure 3 The working process of this invention is as follows: S1 System Power-On and Initialization: Establish grid-connected synchronization (PLL); Read / Set: Sampling period, PWM frequency, maximum sampling offset, random window length, random sequence filter bandwidth, sideband power threshold, and various protection thresholds. Sensor zero point / range, communication link, watchdog timer, and fault flag reset. Assign different random seeds to each inverter; record timestamps to prepare for grid-connected control.

[0041] S2 Standard Sampling Baseline Establishment: Generate the current cycle's "standard sampling time" (PWM center or specified phase point) as a reference. Open the ADC trigger channel and wait for random scheduling to write the offset.

[0042] S3 Sampling Random Scheduling (Decoherent Triggering): In each PWM cycle, a zero-mean, limited sampling offset is obtained using random numbers and low-pass smoothing. This offset is written to the ADC trigger register, causing the sampling in this cycle to be slightly ahead / later than the reference time. The "actual sampling time information" and "offset" for this cycle are output for subsequent compensation. This breaks multi-machine synchronous sampling and weakens coherent coupling at the group level.

[0043] S4 Data Sampling and Buffering: At the trigger moment after random offset, the grid-connected voltage / current is sampled; the sampled value of the current cycle and the sampled value of the previous cycle are written into the fast buffer; the offset of the current cycle is enqueued along with the sampled data (time alignment).

[0044] S5 Fractional Delay Shaping (Phase Compensation): Reads the offset of the current cycle and performs small interpolation / delay compensation on the sampled signal; generates "phase-aligned measurement value" as the input of the current / voltage closed loop; if the offset is zero or exceeds the limit, it automatically bypasses / limits to ensure smooth and continuous output.

[0045] S6 main control closed-loop operation: Uses the shaped signal to perform current inner loop and voltage outer loop calculations, and updates the PWM; maintains a fixed timing sequence for each cycle to ensure real-time performance and dead-time safety; outputs the PWM duty cycle and status variables for the current cycle.

[0046] S7 Sideband Power Monitoring (Resonance Detection): Extracts the fundamental wave and resonance sensitive frequency band components from the real-time grid current flow; calculates the energy ratio between the two to obtain the sideband power index, which serves as a measure of resonance intensity; updates once at a fixed period (e.g., every 10ms) to filter out instantaneous jitter.

[0047] S8 Adaptive Parameter Adjustment: The sideband power is compared with the upper and lower thresholds: if resonance is enhanced, the random disturbance amplitude is reduced and the random change rate is slowed down / the filter bandwidth is adjusted; if the system is stable, the disturbance amplitude is appropriately widened to maintain the decoherence effect; the updated parameters are written back to the random scheduling and shaping module to form an adaptive closed loop of "detection-adjustment".

[0048] S9 Safety Rollback and Recovery: This function handles issues such as sampling setup time, PWM dead-time intrusion, overcurrent, step loss / low-voltage ride-through, and communication timeout. Upon triggering any of these anomalies, random scheduling and reshaping are immediately bypassed, restoring standard timing sampling and normal control; the event and timestamp are recorded. After the anomaly is cleared and the system stabilizes for a period, the impedance reshaping function is smoothly restored, with parameters gradually returning to the target value.

[0049] S10 multi-machine collaboration and consistency maintenance: periodically broadcasts parameter versions, thresholds, random seed status, and timestamps via the bus; requires each unit to maintain the same control parameters but different random sequences to avoid convergence again; synchronously publishes when parameter updates or rollback events occur to ensure group consistency strategies and independent disturbances.

[0050] S11 Operation Log and Health Assessment: Continuously records key indicators such as sideband power, parameter adjustment trajectory, fault / rollback events, temperature and efficiency; periodically generates health assessments and alarms; and leaves a record for operation and compliance.

[0051] S12 Cyclic Execution and Exit Conditions: Steps S3 to S11 are executed in a rolling manner according to the control cycle and monitoring cycle; when the system is shut down / under maintenance / external command requires exit, it is shut down in the following safe order: adaptive → random scheduling → shaping → grid-connected closed loop.

[0052] From a system perspective, this invention introduces a sampling phase random disturbance and signal phase shaping mechanism into the control loop of the photovoltaic inverter cluster, achieving "soft reshaping" of the cluster impedance without changing the hardware topology, thereby effectively suppressing grid resonance. Its basic mechanism can be explained using an impedance superposition model. In a traditional parallel system, assuming that each inverter has the same output impedance and synchronous sampling, the equivalent impedance of the entire photovoltaic cluster from the grid side can be expressed as: ;in, The equivalent impedance of the cluster under synchronous sampling conditions (unit: Ω); The small-signal output impedance function (unit: Ω) of a single inverter is determined by the control parameters of the current loop and voltage loop. is the grid equivalent impedance (unit: Ω), representing the complex impedance from the grid connection point to the public grid; N is the number of parallel inverters; s=jω, where ω=2πf is the angular frequency, and f is the signal frequency (unit: Hz).

[0053] In this structure, the control responses of all inverters are completely synchronized, and their output impedances are in phase and superimposed in the frequency domain. and When the phase angles are close, the system will form a high-amplitude resonance peak at a certain frequency point, causing the cluster to become unstable.

[0054] To suppress this resonance, the present invention introduces a small random phase disturbance during the sampling triggering process of each inverter. This means that the sampling trigger time of each inverter differs slightly from the standard sampling time. This disturbance typically varies within ±1.5 μs. As a result, the output impedance of each inverter has an independent phase factor in the frequency domain. This makes the equivalent impedance expression of the system become: ; Let be the output impedance of the i-th inverter; This is the sampling phase offset of the inverter (unit: seconds); The phase rotation factor describes the phase shift effect corresponding to the time offset in the frequency domain. As random variables, the phases of the output impedances of each inverter exhibit statistical dispersion in the frequency domain. This means that the originally perfectly coherent impedance superposition terms cancel each other out in the frequency response, especially near the resonant frequency, where the coherent interference terms of the impedances are weakened by random phase differences. From an energy perspective, the energy of the coherent peak is diffused into a wider frequency band, causing the equivalent impedance spectrum to transform from a "sharp narrow-band peak" to a "flat broadband shape," thereby significantly reducing the peak value of the cluster impedance and achieving dynamic reshaping of the impedance spectrum. Mathematically, this random perturbation causes the expected value of the interference terms in the impedance superposition to approach zero: Therefore, the amplitude-frequency response of the equivalent impedance tends to be smooth, and the resonant peak value decreases significantly.

[0055] To verify the effectiveness and engineering feasibility of the "control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters" proposed in this invention, a systematic experimental study was conducted on a system of multiple photovoltaic grid-connected inverters. The core objective of the experiment was to evaluate the performance of this control strategy in terms of harmonic energy reduction, system stability margin improvement, and energy loss suppression under actual weak grid conditions, and to compare and analyze it with the traditional synchronous sampling control method.

[0056] In traditional photovoltaic (PV) cluster grid-connected control, all inverters use a unified clock for synchronous sampling, resulting in consistent output impedance phase in the frequency domain. This leads to a strong coherent superposition of the cluster's equivalent impedance. When the phase angle between the grid impedance and the inverter output impedance is close, the system is prone to generating resonance peaks at specific frequencies. To suppress this phenomenon, conventional methods often rely on adding physical damping devices or modifying filter parameters, which not only introduces additional losses but also increases design complexity and cost. The innovation of this invention lies in achieving time-domain desynchronization and frequency-domain phase diffusion between inverters without changing the circuit topology or adding hardware. The core idea is to break the statistical coherence of multi-machine sampling, transforming the system's temporal randomness into equivalent damping in the frequency domain, thus forming a non-physical topology "soft impedance reshaping" mechanism. This mechanism can effectively weaken the resonance energy concentration effect and improve the system's stability margin while maintaining high efficiency and fast dynamic response.

[0057] This experiment included comparative tests: under the same grid conditions, control loop parameters, and power levels, the system performance of traditional synchronous control and the impedance reshaping control of this invention were tested respectively. By measuring the harmonic energy of the grid connection point current, the equivalent impedance spectrum, and the loop phase margin, the improvement effect of the control method of this invention in peak shaving, vibration suppression, and stability was quantitatively evaluated.

[0058] The experimental design fully considered actual industrial operating conditions, including 10 parallel inverters, weak grid conditions (grid impedance Zg = 0.15 + j0.5 Ω), sampling disturbance range of ±1.5 μs, and multi-power point operation. Comparative data were obtained through PRBS frequency sweep, small-signal injection, and power spectral analysis to ensure the reproducibility and engineering representativeness of the conclusions.

[0059] The ultimate goal of the experiment was to verify whether the impedance reshaping control of this invention can significantly reduce the resonant energy of a photovoltaic cluster; whether it can improve the system phase margin and stability without additional losses; and whether this control mechanism can be easily implemented on existing inverter control platforms. Experimental results show that this method successfully achieved "soft impedance reshaping" in a 10-unit parallel system: the harmonic energy of the grid connection point current was reduced by approximately 8–10 dB, the system phase margin was increased by more than 18°, and the overall efficiency remained unchanged, fully verifying the effectiveness and industrial application value of this control strategy.

[0060] The specific experimental results are summarized in the table below: project Test prerequisites and parameters Traditional control (synchronous sampling) This invention controls (impedance reshaping) Changes / Improvements illustrate Number of parallel machines N 10 three-phase inverters (60kVA / unit) same same — The parallel topology is completely consistent. Power grid conditions Weak network: (Z_g=0.15+j0.5\\Omega), fundamental frequency 50Hz same same — Maintain consistency to eliminate grid differences Sampling method Synchronous trigger ((Deltat_i=0)) — Random offset trigger ((\Deltat_i∈[-1.5,1.5]\\mus)) improve Introducing time-domain perturbations Fraction Delayed Shaping none no have New Used for phase compensation and stability maintenance Adaptive adjustment none no Yes (based on sideband power BP) New Dynamically adjust (\Deltat_{max}) and (f_c) Safety rollback mechanism none no There is an abnormal bypass. New Improve security Harmonic energy of grid connection point current (200–800Hz) Measure the RMS ratio (BP = P_{band} / P_{fund}) -32.5dB -41.0dB ↓ Approximately 8.5dB The resonance peak is weakened Peak equivalent impedance (Ω) 2.85Ω 1.12Ω ↓60% Resonance amplitude decreased Phase margin (°) 28.4° 46.7° ↑ Approximately 18.3° Improved system stability Resonant frequency (Hz) 510Hz 505Hz ≈ Slight frequency shift (broadening) System efficiency (%) 97.20% 97.10% ≈ No energy loss (difference <0.1%) LCL filter temperature rise (K) 16.2K 16.4K ≈ Temperature rise is not significant Start-stop dynamic response time (ms) 38.7ms 39.5ms 2.10% Random disturbances do not affect dynamic performance Fault rollback recovery time (s) none 5s auto-recovery New features Meets grid connection standards Experimental environment 25±3℃, power points: 30%, 60%, 100%, steady-state operation for 300s same same — Environmental control consistency ① In the high-frequency resonant region (200–800Hz), the sideband energy of the grid-connected current is reduced by 8–10dB compared with the traditional control, indicating that the impedance phase dispersion caused by random sampling effectively weakens the multi-machine resonant coupling.

[0061] ② Bode / Nyquist analysis showed that the phase margin increased from 28.4° to 46.7°, an increase of more than 18°, proving that the system's stability margin was significantly enhanced and that slight oscillations no longer occurred.

[0062] ③ Frequency domain tests show that the peak equivalent impedance is reduced by 60%, and the resonant bandwidth is slightly broadened (approximately ±5Hz). This indicates that the resonant energy is diffused to a wider frequency band, achieving "decentralization" in the frequency domain.

[0063] ④ The system efficiency remained almost unchanged (<0.1% difference), and the LCL temperature rise was the same, verifying the "soft impedance reshaping" effect. It achieves equivalent damping without relying on physical damping or electrical energy consumption, but only through time-domain phase modulation.

[0064] ⑤ The start-stop response time and power point tracking accuracy are basically consistent with traditional control, indicating that the introduction of ±1.5μs random sampling will not cause perceptible control hysteresis.

[0065] ⑥ Under abnormal operating conditions (communication interruption, overcurrent, etc.), the system can automatically bypass and recover within 5 seconds, meeting the safety standards of industrial grid-connected inverters.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control method for suppressing power grid resonance by reshaping the impedance of a photovoltaic cluster, which is applicable to a cluster system including several photovoltaic grid-connected inverters. Each inverter includes a main controller, a sampling module, a PWM driving module, and grid-connected current and voltage measurement channels, and is characterized in that: This method is executed by an impedance reshaping control unit installed in the main controller of each inverter. The control unit includes a sampling trigger random scheduling module, a fractional delay shaping module, a sideband power monitoring module, and a safety fallback module. This method includes the following steps: (1) Sampling random scheduling: The sampling trigger random scheduling module generates a set of sampling phase offset sequences {Δt(k)} within each PWM cycle according to a preset window length M and a maximum sampling offset Δt_max, satisfying the following constraint conditions: ； where: Δt(k) represents the phase offset of the sampling trigger moment relative to the standard sampling moment in the kth PWM cycle; M represents the number of PWM cycles included in the statistical window; Δt_max represents the maximum allowable sampling phase offset; this sequence is generated by a low-pass filtered pseudo-random sequence, making the sampling phase change have the characteristics of zero mean, zero window sum, and bandwidth limitation, and adjusting the sampling moments of current and voltage in real time at the ADC trigger layer, so that the sampling time bases between each inverter are statistically independent of each other, reducing the group sampling coherence; (2) Fractional delay shaping of the measurement signal: The fractional delay shaping module calculates the fractional delay coefficient α(k) based on the current sampling bias Δt(k), which is defined as: ; where: α(k) represents the fractional delay coefficient of the current sampling cycle; Ts represents the time length of the PWM and sampling cycles; the fractional delay shaping process is performed on the measured signal x[n], and the calculation formula is: ； where: x[n] represents the original measured signal value (current or voltage) obtained in the current sampling cycle; x[n - 1] represents the measured signal value in the previous sampling cycle; y[n] represents the signal output after fractional delay shaping; through the above processing, the linear influence of sampling moment jitter on the phase characteristic of the output impedance is cancelled, and the output impedance of the inverter is stabilized; (3) Sideband power monitoring and adaptive parameter adjustment: The sideband power monitoring module calculates the sideband power index BP of the grid connection point voltage or current in the resonant sensitive frequency band in real time, which is defined as: ; where: P_band represents the effective power component of the signal in the frequency band of 200 - 800 Hz; P_fund represents the effective power component at the fundamental frequency; BP represents the sideband power ratio, which is used to characterize the resonance energy level; when BP > BP_H (upper threshold), the controller automatically reduces Δt_max and reduces the bandwidth of the random sequence; when BP < BP_L (lower threshold), Δt_max and the bandwidth are gradually restored to the default parameters to achieve the adaptive suppression of sideband energy; (4) Safety fallback control: The safety fallback module continuously monitors the following parameters: sampling establishment time t_setup; dead zone protection time t_dead; overcurrent status signal l_over; out-of-step and low voltage ride-through event flag S_fault. When it is detected that t_setup < t_setup_min, or there is a risk of dead zone crossing, or l_over or S_fault is triggered, the sampling trigger random scheduling module and the fractional delay shaping module are immediately bypassed, and the system is restored to the conventional fixed-time sampling and control mode to ensure the safety of the system and the stability of grid connection.

2. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: The sampled phase bias sequence Δt(k) is output by a pseudo-random number generation function and processed by a finite impulse response (FIR) low-pass filter, with its cutoff frequency fc located in the range of 200 to 500 Hz.

3. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: Different inverters use different pseudo-random seeds and initial phase offsets during power-on initialization to ensure that the sampling time base of each inverter is statistically independent.

4. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: The fractional delay coefficient α(k) is filtered by a first-order IIR low-pass filter to limit its rate of change to no more than 500Hz, thus preventing high-frequency oscillations in the control system.

5. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: The sideband power BP is calculated by extracting the power component in the 200-800Hz frequency band using a bandpass filter, with BP_H and BP_L set to default thresholds of 0.08 and 0.04, respectively.

6. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: The safety rollback module includes a hardware limit detection submodule and an event detection submodule. The former is used to monitor whether t_setup, t_dead, and the ADC channel phase interval Δφ meet the minimum safety margin, while the latter is used to detect overcurrent, loss of synchronization, low voltage ride-through, or communication anomalies. If any condition is triggered, bypass will be executed.

7. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: The impedance reshaping control unit is deployed between the timer triggering layer and the ADC interrupt management layer of the inverter main controller. Sampling random scheduling and fractional delay shaping are both completed before the main control algorithm is executed.

8. The control method for suppressing grid resonance by impedance reshaping of photovoltaic clusters according to claim 1, characterized in that: This method is also applicable to multi-machine parallel systems containing energy storage converters, wind power inverters, or flexible DC converter units, to improve their output impedance stability and resonance suppression performance.

Images