A method and system for monitoring a ground resistance of a photovoltaic grid-connected inverter

Abstract

This invention discloses a method and system for monitoring the grounding resistance of a photovoltaic grid-connected inverter, belonging to the field of electrical safety monitoring technology. The excitation calibration module uses spectrum analysis to screen low-interference dual-frequency points and complete amplitude and phase calibration; the impedance calculation module constructs a dual-terminal injection loop on the DC side and PE side, and fits to obtain an impedance matching set; the preprocessing module uses SOGI-PLL phase-locked loop and adaptive notch filtering to purify the sampling signal; the impedance solution module outputs the impedance solution result based on the PE line resistance-constrained overdetermined equation set and capacitive iterative refinement; the RCD coordination module uses Mahalanobis distance to identify test components and dynamically adjusts the threshold to avoid false tripping; and the detection and early warning module achieves remote monitoring and alarming through reliability assessment and graded early warning. In summary, this invention simultaneously solves two major technical defects: measurement distortion under grid-connected conditions and RCD malfunction, improving the accuracy of grounding and insulation resistance measurements, effectively preventing inverters from malfunctioning due to test signals, and ensuring the power generation efficiency and operational stability of the power plant.

Description

Technical Field

[0001] This invention relates to the field of electrical safety monitoring technology, specifically to a method and system for monitoring the grounding resistance of a photovoltaic grid-connected inverter. Background Technology

[0002] Currently, monitoring of grounding resistance and insulation resistance in photovoltaic grid-connected inverters mostly employs independent external monitoring devices or a simple DC injection method within the inverter. Under grid-connected operation, the monitoring signal signal-to-noise ratio drops significantly due to the combined effects of multiple interference factors, including stray currents from the power grid, switching harmonics from inverter power devices, parallel current sharing of multiple PE lines, and parasitic capacitance coupling from the DC side to ground. This severely degrades measurement accuracy and distorts the data, failing to accurately reflect the dynamic resistance changes of the local grounding electrode and thus failing to meet the power plant's maintenance requirements for precise assessment of grounding status. Furthermore, in existing live-line measurement schemes, the test excitation signal is superimposed on the stray voltage of the ground, which can easily trigger malfunctions in the inverter's residual current protection device, causing the inverter to disconnect from the grid and shut down unexpectedly during the testing process, directly resulting in power generation loss. At the same time, frequent tripping interferes with the stable operation of the power plant, leading to secondary faults induced by measurement actions.

[0003] The aforementioned defects are particularly prominent in actual grid-connected operation scenarios: On the one hand, the power frequency harmonics and switching frequency harmonics of the power grid will form a complex stray current distribution on the PE line, which will cause spectral aliasing with the test excitation signal, resulting in the inability to effectively separate the real grounding impedance component during impedance calculation. The final output grounding resistance and insulation resistance values ​​can deviate by tens or even hundreds of ohms, completely losing their reference value; On the other hand, the residual current protection device has extremely high sensitivity to zero-sequence current detection. A test excitation current of milliampere level can be identified as a leakage fault and trigger the grid disconnection protection, which makes the monitoring function and the safety protection function directly conflict. Operation and maintenance personnel have to make trade-offs between real-time monitoring or its detection accuracy and the continuity of power generation, which seriously restricts the practical application and promotion of online monitoring of the grounding and insulation status of photovoltaic power plants. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for monitoring the grounding resistance of a photovoltaic grid-connected inverter, thus solving the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Firstly, a monitoring system for the grounding resistance of a photovoltaic grid-connected inverter includes:

[0007] The excitation calibration module is used to acquire the grid connection point voltage sampling signal, identify the interference distribution of power frequency and switching harmonics through fast Fourier spectrum analysis, and obtain the excitation frequency point and amplitude-phase calibration set;

[0008] The impedance calculation module is used to construct a two-terminal injection loop and, in conjunction with the excitation frequency, obtain the impedance matching set through injection loop impedance matching calibration and shunt ratio parameter determination.

[0009] The preprocessing module is used to track the grid phase using a second-order generalized integrator phase-locked loop and perform orthogonal demodulation cancellation to obtain a preprocessed sampled dataset.

[0010] The impedance calculation module is used to obtain the impedance calculation result set based on the preprocessed sampling dataset and amplitude-phase calibration set, through the coarse solution of the double-ended injection impedance constrained by the known PE line resistance, and after iterative compensation of the capacitive component.

[0011] The RCD collaborative module is used to recover abnormal operating conditions based on the preprocessed sampled dataset, through the identification and pre-identification of the test components of Mahalanobis distance, so as to obtain the RCD collaborative control result set.

[0012] The detection and early warning module is used to perform abnormal data removal based on the RCD collaborative control result set, impedance calculation result set, and preprocessed sampling dataset, and present the results to the cloud platform and user terminal through hierarchical early warning, so as to realize real-time monitoring of photovoltaic grid-connected inverters.

[0013] Secondly, a method for monitoring the grounding resistance of a photovoltaic grid-connected inverter includes the following steps:

[0014] Collect grid connection point voltage sampling signals, identify the interference distribution of power frequency and switching harmonics through fast Fourier spectrum analysis, and obtain excitation frequency points and amplitude-phase calibration sets;

[0015] A dual-ended injection loop is constructed, and by combining the excitation frequency, the impedance matching set is obtained through injection loop impedance matching calibration and shunt ratio parameter measurement.

[0016] A second-order generalized integrator phase-locked loop is used to track the grid phase and perform orthogonal demodulation cancellation to obtain a preprocessed sampled dataset;

[0017] Based on the preprocessed sampling dataset and amplitude-phase calibration set, the impedance calculation result set is obtained by coarsely solving the double-ended injection impedance with known PE line resistance constraints and iteratively compensating for the capacitive component.

[0018] Based on the preprocessed sampled dataset, abnormal operating conditions are recovered by identifying and pre-identifying the test components of Mahalanobis distance, so as to obtain the RCD collaborative control result set;

[0019] Based on the RCD collaborative control result set, impedance calculation result set, and preprocessed sampling dataset, abnormal data is removed and presented to the cloud platform and user terminal through hierarchical early warning, so as to realize real-time monitoring of photovoltaic grid-connected inverters.

[0020] The above-described solution of the present invention has at least the following beneficial effects:

[0021] The excitation calibration module screens low-interference dual-frequency points and, in conjunction with dual-ended injection loops and impedance matching set calibration, suppresses multiple interferences such as power frequency harmonics, switching harmonics, and PE parallel shunt at the source. The preprocessing module uses SOGI-PLL power frequency cancellation and three-level adaptive switching harmonic notch filtering to further purify the sampling signal. The impedance calculation module separates resistive grounding and capacitive coupling components through PE line resistance constrained overdetermined equations and capacitive iterative refinement, ultimately reflecting the dynamic resistance change of the grounding electrode in a true manner, effectively solving the technical problems of measurement accuracy degradation and data distortion under grid-connected conditions.

[0022] Through the four-dimensional feature Mahalanobis distance recognition of the RCD collaborative module, test excitation and actual leakage current can be accurately distinguished. With the pre-identification dual threshold judgment and dynamic threshold adaptive raising, the protection threshold is temporarily raised by 1.5 times the test component during the measurement, but not exceeding three times the upper limit, which ensures that there is no false tripping and leaves a safety margin. When there is actual leakage current, the 100μs change rate fast recovery mechanism immediately switches back to protection and triggers tripping. This greatly reduces the probability of residual current protection false tripping caused by insulation detection process, while not weakening the inverter's detection and response capability to actual leakage current faults on site. It effectively avoids shutdown due to false triggering of monitoring test signals, reduces power generation loss of the power station, and ensures the continuous and stable grid-connected operation and real-time monitoring of the photovoltaic power station. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the structure of a monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to the present invention;

[0024] Figure 2 This is a flowchart illustrating a method for monitoring the grounding resistance of a photovoltaic grid-connected inverter according to the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example 1

[0027] Please refer to Figure 1 This invention provides a monitoring system for the grounding resistance of a photovoltaic grid-connected inverter, comprising:

[0028] The excitation calibration module is used to acquire the grid connection point voltage sampling signal, identify the interference distribution of power frequency and switching harmonics through fast Fourier spectrum analysis, and obtain the excitation frequency point and amplitude-phase calibration set;

[0029] The impedance calculation module is used to construct a two-terminal injection loop and, in conjunction with the excitation frequency, obtain the impedance matching set through injection loop impedance matching calibration and shunt ratio parameter determination.

[0030] The preprocessing module is used to track the grid phase using a second-order generalized integrator phase-locked loop and perform orthogonal demodulation cancellation to obtain a preprocessed sampled dataset.

[0031] The impedance calculation module is used to obtain the impedance calculation result set based on the preprocessed sampling dataset and amplitude-phase calibration set, through the coarse solution of the double-ended injection impedance constrained by the known PE line resistance, and after iterative compensation of the capacitive component.

[0032] The RCD collaborative module is used to recover abnormal operating conditions based on the preprocessed sampled dataset, through the identification and pre-identification of the test components of Mahalanobis distance, so as to obtain the RCD collaborative control result set.

[0033] The detection and early warning module is used to perform abnormal data removal based on the RCD collaborative control result set, impedance calculation result set, and preprocessed sampling dataset, and present the results to the cloud platform and user terminal through hierarchical early warning, so as to realize real-time monitoring of photovoltaic grid-connected inverters.

[0034] In this embodiment, the system adopts a six-module collaborative architecture to form a complete closed loop from excitation generation, hardware calibration, signal preprocessing, impedance calculation, protection coordination to hierarchical early warning. The excitation calibration module uses spectrum analysis to screen low-interference frequency points, such as avoiding power frequency harmonics and switching harmonics in the 20Hz to 200Hz range, to generate dual-frequency orthogonal test signals and complete amplitude and phase calibration, thereby improving the measurement signal-to-noise ratio from the source.

[0035] The impedance calculation module constructs a dual-ended injection loop on the DC side and the PE side. For example, it fits the inherent impedance and mutual coupling impedance through 25 sets of standard load combinations to form an impedance matching set and eliminate hardware bias. The preprocessing module uses SOGI-PLL phase-locked loop and three-stage adaptive notch filtering to attenuate the switching frequency and its second and third harmonics by more than 30dB, outputting a clean preprocessed dataset.

[0036] The impedance calculation module constructs an overdetermined system of equations based on the known constraint of the PE line resistance. For example, it uses the Levenberg-Marquardt algorithm to solve the equations and refines them through capacitive iteration to achieve precise decoupling of grounding and insulation impedance. The RCD coordination module identifies the test component and dynamic threshold rise through the Mahalanobis distance. For example, it temporarily raises the protection threshold by 1.5 times the test component to avoid false tripping.

[0037] The detection and early warning module achieves reliable monitoring and tiered alerts through a five-dimensional credibility score and a three-level warning system. For example, if the credibility score is below 60, it will automatically interpolate and complete the score.

[0038] In a preferred embodiment of the present invention, the excitation calibration module includes:

[0039] The frequency interference extraction unit is used to perform Fourier transform on the grid connection point voltage sampling signal and extract all interference frequency points and amplitudes of power frequency harmonics and switching harmonics through spectrum analysis;

[0040] During use, the phase voltage signal at the inverter's grid connection point is acquired, with a sampling rate set to 20kHz and a continuous acquisition duration of 200ms. A 4096-point Fast Fourier Transform is performed on the acquired time-domain voltage sequence to obtain the amplitude and phase spectra within the frequency range of 0Hz to 10kHz. The frequency points and amplitude values ​​corresponding to the 50Hz fundamental frequency and its second to tenth harmonics are marked in the amplitude spectrum. Simultaneously, the frequency points of the fundamental switching frequency generated by the inverter's SPWM modulation and its upper and lower sideband harmonics are located. All marked interference frequency points and their corresponding amplitude and phase information are summarized to form an interference frequency point set.

[0041] The frequency extraction unit generates candidate frequency points based on the low-frequency signal injection method. For each candidate frequency point, it calculates the minimum frequency interval between it and each interference frequency point, and selects candidate frequency points whose intervals with all interference frequency points are greater than a preset interval threshold, generating a candidate frequency point set. From the candidate frequency point set, it selects the two candidate frequency points with the lowest background noise amplitude and whose frequency difference between the two candidate frequency points is not less than a preset frequency difference threshold, which are then used as excitation frequency points. The excitation frequency point serves as the test signal frequency injected into the PE circuit, avoiding grid-connected harmonic interference to accurately calculate grounding and insulation impedance. By using two different excitation frequency points, an overdetermined equation system can be constructed to separate the resistive grounding electrode impedance and the capacitive DC-to-ground coupling component, eliminating capacitive shunt distortion caused by single-frequency point measurement.

[0042] The background noise amplitude is the amplitude of the spectrum position corresponding to the original candidate frequency point being traversed. Specifically, when traversing each original candidate frequency point generated with a step size of 1Hz within 20Hz to 200Hz, the system jumps to the spectrum coordinates corresponding to that frequency point and reads the inherent background noise amplitude under that coordinate when there is no test excitation. This is used to measure the degree of spurious noise pollution at that frequency position.

[0043] In practice, within the low-frequency range of 20Hz to 200Hz, all candidate frequency points are traversed in 1Hz increments. For each candidate frequency point, the minimum frequency interval between it and all frequency points in the interference frequency point set, as well as the background noise amplitude at that frequency point, are calculated. A frequency interval threshold is set, for example, 10Hz, and candidate frequency points whose intervals with all interference frequency points are greater than this threshold are selected to form a candidate frequency point set. In the candidate frequency point set, the two frequency points with the lowest background noise amplitude are selected as excitation frequency points, and the frequency difference between the two excitation frequency points is not less than 20Hz. This avoids the overlap of the spectra of the two sets of orthogonal test signals, ensuring that the signal components can be completely separated and do not interfere with each other during dual-frequency demodulation.

[0044] The minimum frequency interval is the frequency difference between a single candidate frequency point and the nearest one among all interference frequencies, which ensures that the spectrum of the test signal and the interference spectrum are separated and do not overlap.

[0045] The waveform unit is used to generate a test waveform data table that can be loaded into a signal generator based on the excitation frequency through orthogonal encoding and discretization sampling.

[0046] The angular frequency is calculated based on two excitation frequencies, and two sinusoidal signals are designed: and The two paths have a 90-degree phase difference; the synthesized signal is a linear superposition of the two paths. .set up Confirm that the synthesized peak value does not exceed the 12V safety limit. Take twice the period of the least common multiple of the two frequency points as the measurement period length, based on the sampling rate. The synthesized signal is discretized to generate a test signal waveform data table. This data table is then written to the signal generator buffer and configured for external GPIO rising edge triggering mode. The table has two columns: the first column is the discrete sampling time index, and the second column is the synthesized signal at the corresponding time. Discrete amplitude is used to store digital waveform samples, load them into the signal generator buffer, and output complete dual-frequency orthogonal test excitation according to the timing sequence after triggering;

[0047] in , There are two excitation frequencies; , Given the angular frequencies of the two signals, convert the frequencies into sinusoidal oscillation angular velocities for time-domain waveform solving; For time domain time variables; , The amplitudes of the two sinusoidal signals are both set to 5V to control the amplitude of the test excitation voltage and meet the safety extra-low voltage specifications.

[0048] The first single-frequency sinusoidal test excitation time domain function is used to generate the first frequency point continuous alternating test voltage as the first set of independent measurement excitations, which is used to collect the voltage and current response of the PE circuit at this frequency point, and provide the first set of frequency domain observation data for subsequent impedance calculation; It is a standard sinusoidal oscillation term, which enables periodic alternating voltage output;

[0049] This is the time-domain function of the second single-frequency sinusoidal test excitation. It is the circumference constant. This achieves 90° phase orthogonality between the two signals, facilitating orthogonal demodulation and separation of components at various frequencies. A fixed quadrature offset is introduced for the phase term of the second signal;

[0050] The dual-frequency synthesized total test signal is superimposed with two orthogonal frequency point signals, which can be used to simultaneously acquire two sets of frequency point responses with a single injection. For discrete sampling rates, This means taking the larger of the two frequency points and multiplying it by a coefficient of 64 to ensure at least 64 sampling points per signal cycle, thus improving demodulation accuracy.

[0051] The calibration unit, based on test waveform data and multi-value standard resistance boxes, uses multi-point synchronous sampling and frequency domain component extraction, combined with a least-squares linear fitting algorithm, to obtain an amplitude and phase calibration set that varies with resistance value and stores it in non-volatile memory. This set contains voltage and current channel calibration parameters and phase offsets, used to perform resistance-related corrections on the amplitude and phase of the sampled and demodulated voltage and current, eliminating hardware measurement bias.

[0052] In use, connect the output of the signal generator to a standard resistance box with an accuracy class of 0.01, and set the resistance of the standard resistance box to 10Ω. Drive the signal generator to output the above-mentioned synthesized test signal, and use a high-precision data acquisition card to synchronously acquire the voltage signal and loop current signal across the standard resistance box.

[0053] Fourier transforms were performed on the acquired voltage and current waveforms, and the corresponding amplitude and phase parameters were extracted at the two excitation frequencies. The theoretical amplitude and phase at each frequency were calculated based on the nominal resistance of the standard resistance box, and the gain coefficients of the voltage and current channels, as well as the phase offset, were obtained at the two excitation frequencies. , ;

[0054] The resistance values ​​of the standard resistance box were sequentially changed to 1Ω, 5Ω, 20Ω, 50Ω, and 100Ω. The above measurement and calculation process was repeated to obtain the calibration coefficients corresponding to each resistance value. The least squares method was used to linearly fit the relationship between the calibration coefficients and the resistance value to obtain the voltage channel calibration formula: ;

[0055] Current channel calibration formula: ;

[0056] Calibration parameters , , , , , , Write to non-volatile memory to form an amplitude and phase calibration set.

[0057] in This represents the actual voltage amplitude after calibration. This is the zero-order voltage calibration constant; This is the gain coefficient related to voltage resistance. The actual load resistance of the circuit is the equivalent resistance of the signal generator output circuit. During calibration, it is the resistance value set by the standard resistance box. During actual measurement, it is the composite load resistance composed of photovoltaic insulation and grounding electrode. During calibration, the nominal resistance value of the resistance box is directly read.

[0058] The raw voltage value directly acquired by the ADC is the raw sampled voltage directly output by the analog-to-digital converter without error compensation, and is superimposed with circuit gain drift and zero-point deviation; it synchronously acquires the waveform across the resistor through the acquisition card, and extracts the raw voltage amplitude at the corresponding test frequency point through Fourier transform. This is the zero-point offset compensation amount for the voltage channel;

[0059] This is the actual current amplitude after calibration; This is the zero-order calibration constant for the current. This is the gain coefficient related to the current resistance value; This is the compensation amount for the zero-point offset of the current channel; Refers to the parameters corresponding to the current path; This refers to the parameters corresponding to the voltage channel;

[0060] , To ensure accurate voltage and current values ​​after eliminating hardware sampling errors, which are then used for subsequent impedance and grounding resistance calculations, and to avoid measurement distortion caused by hardware deviations;

[0061] The signal generator is an excitation generation unit built into the inverter that can output controllable low-frequency orthogonal dual-frequency test waveforms. It is used to obtain an amplitude and phase calibration set by multi-point calibration with the help of standard resistors, and to eliminate the inherent deviations of amplitude and phase in the sampling hardware.

[0062] The resistance value index is set by sequentially switching the discrete standard resistance values ​​of the timed standard resistance box.

[0063] The zero-order calibration constant is a fixed amplitude deviation constant inherent in the sampling channel when there is no load. It is obtained by solving the least-squares linear fit through multiple sets of standard resistance point measurement data.

[0064] The zero-point offset compensation is a fixed DC offset correction value caused by the zero-point drift of the sampling hardware. It is obtained by linear fitting of multi-resistance calibration data and used as a constant term in the calibration formula.

[0065] In this embodiment, a four-unit cascaded architecture is adopted to form a complete excitation link from interference identification, frequency point selection, waveform generation to amplitude and phase calibration, thereby solving the measurement distortion problem under grid-connected conditions from the source. The frequency interference extraction unit uses 4096-point FFT to scan the entire 0~10kHz spectrum, for example, simultaneously marking the second to tenth harmonics of the power frequency and the SPWM switching frequency sideband, to achieve accurate positioning of interference across the entire frequency band and avoid missing any potential interference sources.

[0066] The frequency extraction unit traverses the 20-200Hz range in 1Hz steps. For example, by using a 10Hz interval threshold and a 20Hz frequency difference constraint for dual screening, it selects the dual-frequency point with the lowest background noise from more than a hundred candidate points. This ensures complete separation from the interference spectrum and avoids overlap between the two frequency points. The dual-frequency point design can construct an overdetermined set of equations to separate resistive grounding and capacitive coupling components, eliminating capacitive distortion in single-frequency measurements, which is something that conventional single-frequency solutions cannot achieve.

[0067] The waveform unit uses two orthogonal sine synthesis with a 90° phase difference. For example, the amplitude is set to 5V and the synthesis peak is controlled within the 12V safety limit. With a 64x oversampling rate, two sets of independent frequency domain data can be obtained simultaneously with a single injection.

[0068] The calibration unit uses least-squares fitting of six sets of standard resistance points to generate a linear calibration formula that varies with resistance, rather than a fixed gain coefficient, thus compensating for hardware gain drift under different loads. The entire amplitude and phase calibration set is stored in non-volatile memory and can be recalled upon power-on without the need for repeated calibration.

[0069] In a preferred embodiment of the present invention, the impedance measurement module includes:

[0070] The loop construction unit is used to construct a dual-ended injection loop on the DC side and PE side within the photovoltaic inverter. Each loop is connected in series with a DC blocking capacitor, a current limiting resistor, an optocoupler relay, and a sampling resistor. They share a signal source via an isolation transformer to obtain a hardware configuration table for impedance calculation. This step provides a physical measurement channel for subsequent acquisition of raw voltage and current sampling data and calculation of grounding and insulation impedance by building a hardware path that can synchronously inject dual-frequency test signals into the DC and PE ground wires.

[0071] The hardware configuration table records the model, resistance value, capacitance value, withstand voltage, voltage division ratio, and series-parallel topology correspondence of each branch's DC blocking capacitor, current limiting resistor, sampling resistor, isolation transformer, and voltage divider network, providing all inherent hardware parameters for subsequent impedance compensation calculations and sampling data calibration.

[0072] In the dual-ended injection circuit, the first injection terminal is connected between the positive terminal of the DC bus and the ground reference to apply test excitation and collect the insulation impedance of the photovoltaic DC side to ground; the second injection terminal is connected between the PE terminal and the ground reference to apply test excitation and collect the impedance of the local grounding electrode and the parallel shunt impedance of multiple PE circuits.

[0073] The purpose of using a unified and shared ground reference is to ensure that the two injected excitations have the same zero potential point, eliminate the potential reference deviation of the two-end measurements, and then solve the equations to decouple the three types of parameters: insulation resistance, grounding resistance, and DC-to-ground stray capacitance.

[0074] In practice, a dual-ended injection circuit is constructed inside the inverter. Each injection circuit is connected in series with a DC blocking capacitor, a current limiting resistor, an optocoupler relay, and a current sampling resistor. The capacitance value of the DC blocking capacitor is calculated using the following formula. Ensure that the capacitive reactance at the lowest test frequency does not exceed one-tenth of the current-limiting resistor value. The current-limiting resistor value is calculated using the following formula. The maximum injected current The current is set to 10mA. The current sampling resistor is a 0.1% accuracy metal film resistor, determined by the following formula. Two injection loops are equipped with isolation transformers with a 1:1 turns ratio and a 3kV withstand voltage on the signal source side. The primary windings of the transformers are connected in parallel and share the same signal generator output. A voltage sampling divider network is set at the output of each injection loop, with the divider ratio matched to the ADC input voltage range. The voltage and current sampling signals are amplified by an instrumentation amplifier and then sent to a multi-channel ADC; the parameter values ​​and connection relationships of each component are recorded to form a hardware configuration table.

[0075] in This is the nominal capacitance value of the DC blocking capacitor, used to block the high voltage DC component of the photovoltaic DC bus, allowing only the AC test excitation signal to pass through, and preventing the DC high voltage from damaging the downstream signal acquisition circuit.

[0076] The lowest excitation frequency with the smaller value among the two sets of excitation frequencies; The resistance value of the current-limiting resistor in the loop is used to limit the maximum injected test current in the loop from not exceeding the safety threshold, so as to prevent excessive excitation current from damaging the sampling element and triggering the inverter leakage protection.

[0077] The peak voltage of the orthogonal synthesized test signal is the maximum instantaneous voltage after the superposition of two orthogonal test sinusoidal signals. For the maximum injection current, e.g., 10mA;

[0078] The resistance value of the current sampling resistor is used to calculate the real-time test current of the circuit by collecting the voltage drop across the resistor, thus providing the original current sampling signal for impedance calculation. This is the full-scale input voltage of the back-end ADC chip, which is the maximum linear input voltage that the ADC input pin can withstand. It can be obtained by directly reading the nominal electrical parameters in the ADC device datasheet; ADC stands for Analog-to-Digital Converter.

[0079] The numerator in the calculation formula refers to the design coefficient, constraining the DC blocking capacitor. The capacitive reactance shall not exceed 1 / 10 of the current-limiting resistor;

[0080] The bridging topology unit is used to construct a hardware topology for the test signal bridging branch based on two excitation frequencies by leading out a bridging branch from the PE line on the primary side of the residual current transformer and connecting frequency selection and control elements in series. This topology is a parallel branch circuit structure that is connected across the PE lines on both sides of the residual current transformer and integrates filters, switches, and sampling resistors. It is used to shunt the test excitation current to ground to avoid the test current triggering the leakage current protection device to malfunction, and at the same time collects the shunt data for algorithm compensation.

[0081] The PE line leading out of the primary side of the residual current transformer is the whole machine protective ground line PE passing through the primary side of the residual current transformer (RCT). An additional parallel branch line is led out between the incoming and outgoing lines of the transformer, and this branch line is connected across the PE lines on both sides of the transformer.

[0082] The jumper branch is a branch line connected across the PE lines at both ends of the RCT. It is used to shunt the dual-frequency test excitation signal, allowing the test current to flow through the jumper branch, which greatly reduces the test current flowing into the primary side of the RCT, preventing the monitoring excitation from triggering the leakage protection to trip falsely. At the same time, the shunt amplitude is collected for subsequent algorithm compensation.

[0083] In practice, a bridging branch is led out from the PE line on the primary side of the residual current transformer, and the other end of the branch is connected to the reference potential at the midpoint of the DC bus. The bridging branch is connected in series with two stages of double-T notch filters adapted to two test frequencies, a voltage follower for interstage impedance isolation, a MOS controllable switch controlled by the MCU timing, and a bypass current sampling resistor.

[0084] The series frequency selection and control components consist of frequency selection devices that filter out power frequency interference, and supporting components that control on / off switching and current acquisition, which are sequentially assembled in the branch. The control components are MOS and controllable switches, which are controlled by MCU to switch the branch on / off, and only conduct the bypass during measurement.

[0085] The shunt unit is used to calculate the impedance values ​​of the shunt branch and the main PE branch at two excitation frequency points based on the hardware topology of the shunt branch of the test signal. The impedance value of the main PE branch is used as the numerator, and the sum of the impedance values ​​of the main PE branch and the shunt branch is used as the denominator. The two are divided to obtain the shunt amplitude of the test current flowing into the shunt branch at that frequency point. This amplitude represents the ratio of the current amplitude shunted by the shunt branch to the total current amplitude of the PE circuit under the same excitation signal. When demodulating the sampled signal, this amplitude ratio is used to compensate the current amplitude collected by the ADC, eliminate the amplitude attenuation error caused by the bypass shunt, and ensure the accuracy of impedance calculation.

[0086] The impedance value of the bridging branch is the total equivalent AC impedance of the entire bridging branch, including the double-T notch filter circuit, voltage follower, and sampling resistor, at each excitation frequency. It is obtained by adding the equivalent impedance of the two-stage double-T network, the output impedance of the voltage follower, and the resistance value of the current sampling resistor at the corresponding frequency according to the circuit series impedance superposition rule, to obtain the total bypass impedance at the corresponding frequency.

[0087] The impedance value of the main PE branch is the AC impedance of the original PE protective ground wire passing through the residual current transformer at the excitation frequency. The calculation method is as follows: collect the resistance of the PE conductor itself and the inductive reactance of the distributed inductance at the excitation frequency, and combine them to obtain the AC impedance of the main PE branch.

[0088] Both sets of impedances are complex impedances, containing resistive and reactive components, used to accurately calculate the shunt amplitude and phase shift. The phase shift is the phase difference between the bypass current and the original injected excitation after the excitation signal flows through the bypass filter network and cable impedance. It is used to compensate for the phase difference during algorithm demodulation, correcting errors in solving the imaginary part of the impedance and capacitive stray parameters, and avoiding decoupling distortion from grounding or insulation impedance. It is obtained by: introducing a standard sinusoidal excitation at the corresponding frequency into the bridging branch, simultaneously acquiring the original injected excitation waveform and the bridging branch current waveform, and comparing the phase difference between the two waveforms at the same moment; this difference is the phase shift at the corresponding frequency.

[0089] The impedance fitting unit is used to obtain the inherent impedance and mutual coupling impedance based on the apparent impedance measurement and least-squares fitting of multiple standard load combinations under a double-ended injection circuit and a bridging branch, forming an impedance matching set. This step is used to collect multiple sets of standard load impedance data to fit the inherent coupling parameters of the hardware, eliminating measurement errors caused by the hardware's own impedance.

[0090] In practice, the double-ended injection circuit is connected to a standard test platform, which is equipped with a standard grounding resistance box and a standard insulation resistance box. The first injection end is connected to the high-voltage end of the standard insulation resistance, the low-voltage end of the standard insulation resistance is connected to the upper end of the standard grounding resistance, and the lower end of the standard grounding resistance is connected to the earth reference.

[0091] The second injection terminal is directly connected to the upper end of the standard grounding resistor. The standard grounding resistance is set to 10Ω and the standard insulation resistance to 1MΩ. Test signals are then output from both injection terminals, and the voltage and current at each terminal are collected to calculate the apparent impedance. , The standard grounding resistance was sequentially changed to 1Ω, 5Ω, 20Ω, 50Ω, and 100Ω, and the standard insulation resistance was sequentially changed to 100kΩ, 500kΩ, 2MΩ, and 5MΩ, forming 25 sets of test conditions. Under each set of conditions, the apparent impedance of the two injection terminals was measured. Based on the multiple sets of measurement data, the inherent impedance of each injection loop was fitted using the least squares method. And the mutual coupling impedance between the two injection terminals. The fitting formula is: The inherent impedance parameters, mutual coupling impedance parameters, and bypass shunt ratio template are written together into the calibration parameter area to form an impedance matching set.

[0092] in , The apparent impedances of the DC-side injection terminal and the apparent impedance of the PE-side injection terminal are the complex ratio of the effective value of the loop voltage to the effective value of the current at a specific frequency point, which is used to solve for the insulation to ground and the equivalent impedance of grounding.

[0093] , The excitation voltage and loop current at the DC injection terminal; , The excitation voltage and loop current at the PE injection terminal; The inherent complex impedance of the circuit is the resistance and reactance of the hardware itself injected into the circuit. It is used to deduct the inherent losses of the hardware and restore the true load impedance.

[0094] This is the inherent resistance component. The imaginary unit is the mathematical symbol used in complex impedances to distinguish between the resistive and reactive components, and it represents the phase difference between voltage and current. This is the inherent reactance component, which is the reactance formed by the inductance and capacitance of the circuit hardware itself, and does not include the reactance of the load being measured.

[0095] The total apparent impedance of the entire machine is the total impedance converted from the measured voltage and current of the entire equipment. It is used to fit the impedance of the separated hardware and the load under test. The external load impedance is the equivalent composite impedance of the photovoltaic DC insulation and grounding electrode to be measured, which is the target measurement quantity of this scheme.

[0096] The mutual coupling impedance is obtained by collecting two sets of measured apparent impedances and two inherent impedances under multiple different standard loads, eliminating the load term by solving the simultaneous equations, and obtaining the complex value of the mutual coupling impedance between the channels by the difference; it is the complex impedance corresponding to the electromagnetic crosstalk between the two injected lines, and is used to offset the measurement deviation caused by the mutual interference between the channels during the calculation.

[0097] In this embodiment, a four-unit collaborative architecture is adopted, forming a complete hardware calibration link from loop construction, bypass topology, current shunting calculation to impedance fitting, providing hardware parameter support for accurate backend calculation. The loop construction unit innovatively adopts a dual-end injection design on both the DC side and the PE side. For example, the two paths share the same signal source and a unified ground reference through a 3kV isolation transformer, which can simultaneously acquire two sets of independent observation data for insulation and grounding. This provides a physical basis for decoupling the three types of parameters in the subsequent simultaneous overdetermined equation system, which cannot be achieved by conventional single-end injection schemes.

[0098] The DC blocking capacitor is designed with a 1 / 10 capacitive reactance constraint. For example, at a current of 10mA, it ensures low-loss passage of the test signal while completely blocking the high voltage of the DC bus, thus balancing safety and accuracy.

[0099] The bridging topology unit constructs a frequency-selective bypass between the PE lines on both sides of the residual current transformer. For example, two-stage double-T notch filters adapted to two test frequencies are connected in series, allowing only the test excitation current to flow through the bypass, while the actual power frequency leakage current still passes through the transformer for detection, thus achieving functional decoupling of testing and protection.

[0100] The shunt unit employs precise complex impedance calculations, simultaneously outputting both shunt amplitude and phase offset parameters, rather than just amplitude ratio, thus fully compensating for vector errors introduced by bypassing. The impedance fitting unit performs least-squares fitting using 25 standard load combinations (5 grounding levels × 5 insulation levels), simultaneously solving for both inherent impedance and mutual coupling impedance parameters to form an impedance matching set, which can simultaneously eliminate both hardware impedance and crosstalk errors from both paths.

[0101] In a preferred embodiment of the present invention, the preprocessing module includes:

[0102] The sampling unit is used to obtain the original sampling set according to the hardware configuration table by synchronous triggering and automatic DMA transfer when the ground impedance detection is started.

[0103] In practice, upon measurement startup, the MCU sends a 10ms high-level trigger pulse to the signal generator and simultaneously initiates multi-channel ADC synchronous sampling, resulting in seven sampling channels: first injection terminal voltage, first injection terminal current, second injection terminal voltage, second injection terminal current, PE line zero-sequence current, DC bus positive terminal to ground voltage, and grid A-phase voltage. Each channel is driven by the same sampling clock for synchronous sampling and holding, with the sampling rate consistent with the signal generator. Sampled data is automatically transferred to a circular buffer via DMA. Each channel has 4096 points, with half-full and full-full interrupts. During interrupts, data is copied to the processing buffer. The total sampling time is two complete measurement cycles.

[0104] During sampling, the MCU synchronously controls the timing of the injection circuit electronic switches and bypass switches to ensure that the switch states stabilize in advance. After sampling, the signal output is turned off and all switches are disconnected. The complete sampling data from the seven channels is arranged by time to form the original sampling set. This set carries all the synchronously acquired original waveforms and serves as the basic input data for power frequency cancellation and harmonic filtering algorithms.

[0105] MCU stands for Microcontroller Unit, which is used to uniformly control all timing logic of signal generators, ADC sampling, various electronic switches, filtering operations, and data buffering.

[0106] The voltage at the first injection terminal is the real-time voltage at the excitation output terminal of the DC bus positive terminal to ground.

[0107] The first injection terminal current is the test excitation current flowing through the DC side injection circuit;

[0108] The voltage at the second injection terminal is the real-time voltage at the PE protection line's ground excitation output terminal;

[0109] The second injection terminal current is the test excitation current flowing through the injection circuit on the PE side;

[0110] The zero-sequence current of the PE line is the leakage stray current generated by the three-phase imbalance on the PE line; the PE line is the protective grounding wire.

[0111] The voltage of the positive terminal of the DC bus to ground is the potential of the positive terminal of the photovoltaic DC bus relative to the earth reference.

[0112] The A-phase voltage of the power grid is the power frequency voltage of the A-phase to ground in the AC power grid, which is used for phase-locking to cancel power frequency interference; where A is the A-phase in the three-phase AC power.

[0113] The power frequency cancellation unit is used to track the power grid phase through a second-order generalized integrator phase-locked loop and perform orthogonal demodulation cancellation based on the original sampling set, to obtain the power frequency desampled dataset and the power grid phase sequence.

[0114] In practice, the A-phase voltage sequence of the power grid is extracted from the original sampled data. A second-order generalized integrator phase-locked loop is used to track the grid phase, and the SOGI transfer function is... ;

[0115] in , ;

[0116] SOGI outputs are orthogonal and The components are obtained by the Park transform. and , The input frequency correction value of the PI regulator is integrated to obtain the real-time grid phase. Constructing two power frequency reference signals based on the power grid phase: , ;

[0117] For the remaining six channels, the data is multiplied point-by-point with the two reference signals to obtain in-phase and quadrature product sequences. A moving average filter is then applied to the product sequences within a one-power-frequency cycle to obtain the in-phase DC component. and orthogonal DC components Calculate the amplitude of power frequency interference. With phase The power frequency interference waveform is reconstructed and subtracted point by point from the original data. The processed data forms a power frequency de-sampled dataset, which is a seven-channel sampled waveform after power frequency interference from the power grid has been removed; this dataset is used for the next step of filtering out switching frequency harmonics.

[0118] in This is the complex frequency domain transfer function of the SOGI phase-locked loop, used to filter out harmonics, and outputs orthogonal power frequency components for phase locking;

[0119] SOGI is the damping adjustment coefficient; It is the power frequency angular frequency; For the Laplace complex frequency operator; This is the damping term in the molecule of the transfer function; The second characteristic denominator of the transfer function; , SOGI outputs quadrature voltage components;

[0120] and This represents the DC component of the orthogonal axis after the Park transform;

[0121] , The power grid frequency sine and cosine reference sequences are used as demodulation references to separate the signal from the power frequency interference. , To generate sine and cosine values ​​from the real-time power grid phase;

[0122] The amplitude of the power frequency interference is quantified, and it is used to reconstruct the interference waveform to cancel it out. It records the phase of the power frequency interference to ensure that the canceled waveform is synchronized with the interference. The arctangent function is used to solve for the interference phase.

[0123] The power grid phase sequence is a real-time power grid phase value recorded point by point, which is used to compensate for the residual phase deviation at the power frequency during subsequent impedance calculation;

[0124] The switching frequency detection unit is used to extract grid voltage channel data from the de-power frequency sampling dataset, and search for the frequency point with the maximum amplitude near the nominal switching frequency through Fourier transform to determine the actual switching frequency;

[0125] The harmonic filtering unit is used to obtain a preprocessed sampling dataset based on the actual switching frequency, through a three-stage adaptive IIR notch filter and a fourth-order Butterworth low-pass filter, combined with attenuation verification and adjustment, resulting in power frequency cancellation and harmonic suppression. This dataset consists of seven-channel synchronous waveform data after power frequency interference cancellation and multi-stage notch filtering of switching frequency harmonics, along with a synchronous power grid phase sequence. It does not contain the two main types of interference: power frequency and switching harmonics. As the pure input data for the back-end impedance calculation module, it is combined with the impedance matching set to deduct inherent hardware impedance and channel mutual coupling interference, accurately solving for DC-to-ground insulation impedance and PE grounding equivalent impedance.

[0126] Specifically, the grid voltage channel data is extracted from the power frequency sampling data, and the maximum amplitude frequency point is searched near the nominal switching frequency using Fourier transform to determine the actual switching frequency. A three-stage adaptive IIR notch filter is designed based on the actual switching frequency, targeting... , , The transfer function of each notch filter stage is:

[0127] in , According to real-time detection The parameters at each level are dynamically updated with an update period of 10ms. The data from the six signal channels are sequentially passed through a three-stage notch filter, followed by a fourth-order Butterworth low-pass filter, with the cutoff frequency set to... Twice that of the previous filter. After filtering, the switch harmonic attenuation was verified to be greater than 30dB; if insufficient, the r value was adjusted to increase the notch depth. The data from the seven channels after power frequency cancellation and harmonic notch filtering, together with the power grid phase sequence, formed a preprocessed sampling dataset.

[0128] in This is the transfer function of a discrete-domain IIR notch filter, used to filter out the switching frequency and its second and third harmonics, and suppress high-frequency interference introduced by the switching of power devices. The normalized angular frequency of the notch center; To transform the complex frequency domain operator and characterize the discrete sampled signal; This is the filter pole shrinkage coefficient, which controls the notch bandwidth and attenuation depth;

[0129] Its numerator is a zero-point term, used to cancel the target harmonic frequency; the denominator is a pole term, which determines the filter convergence characteristics. This is the ADC synchronization sampling frequency;

[0130] In this formula, the zero point of the numerator is aligned with the frequency of the harmonic to be filtered out to achieve amplitude suppression; the pole of the denominator is determined by... The control ensures system stability and adjusts the notch depth. This formula adaptively filters out switching harmonics from the pre-processed sampled data, reducing harmonic errors in impedance calculation.

[0131] In this embodiment, a complete signal preprocessing link is formed from synchronous sampling, power frequency cancellation, switching frequency detection to harmonic filtering, providing clean input data for back-end impedance calculation.

[0132] The sampling unit achieves synchronous acquisition across seven channels, for example, 4096 points per channel. The DMA automatically transfers the data to the circular buffer and coordinates with the MCU to synchronously control the timing of the injection switch and bypass switch, ensuring strict alignment between sampling and excitation and avoiding calculation errors caused by phase deviation.

[0133] The power frequency cancellation unit innovatively adopts SOGI-PLL phase-locked loop to track the power grid phase. For example, it reconstructs the power frequency interference waveform through quadrature demodulation and moving average filtering and cancels it point by point. Compared with the traditional fixed notch filter scheme, it can adapt to power grid frequency fluctuations and improve the power frequency suppression depth by more than 20dB. At the same time, it outputs a point-by-point power grid phase sequence, which can be used for subsequent impedance calculation to compensate for the residual phase deviation of the power frequency. This is something that conventional preprocessing schemes do not have.

[0134] The switching frequency detection unit searches for the maximum amplitude near the nominal switching frequency using Fourier transform, thus locking onto the actual switching frequency in real time instead of relying on a fixed nominal value, adapting to slight frequency drift under different operating conditions. The harmonic filtering unit is designed with a three-level adaptive IIR notch filter based on the measured switching frequency, targeting the fundamental, second, and third harmonics respectively. Combined with a fourth-order Butterworth low-pass filter, the switching harmonic attenuation reaches over 30dB.

[0135] By dynamically adjusting the notch depth by the r value, the filtering effect and signal fidelity are balanced. The signal-to-noise ratio of the final preprocessed dataset is improved by an order of magnitude, which significantly improves the accuracy of the back-end impedance calculation.

[0136] In a preferred embodiment of the present invention, the impedance calculation module includes:

[0137] The demodulation unit is used to obtain the voltage and current amplitude and phase parameters of four channels at two frequencies by using dual-frequency orthogonal demodulation and resistance-related calibration correction based on the preprocessed sampled dataset and amplitude and phase calibration set, thus forming an orthogonal demodulation result set.

[0138] The voltage and current amplitude and phase parameters of the four channels at two frequency points refer to the accurate voltage amplitude, voltage phase, current amplitude, and current phase vector data obtained after quadrature demodulation and hardware amplitude and phase calibration at two frequency points for each of the four channels.

[0139] In practice, four channels of data are extracted from the preprocessed sampled data: the discrete sampled voltage waveform at the first injection terminal, the discrete sampled current waveform at the first injection terminal, the discrete sampled voltage waveform at the second injection terminal, and the discrete sampled current waveform at the second injection terminal. For the two excitation frequencies, orthogonal reference signal sequences are constructed respectively. ; ;

[0140] in Each channel data is multiplied point by point with the two reference signals to obtain four product sequences;

[0141] A moving average filter is applied to the product sequence, with a window spanning two test periods, to obtain the in-phase DC component and quadrature DC component at each frequency point for each channel. Taking the first injection terminal voltage as an example, the amplitude is: ;

[0142] Phase: ;

[0143] The current is calculated similarly. For the initial calculation, the estimated grounding resistance is taken as 10Ω, and calibration correction is performed using the amplitude and phase calibration set. The voltage calibration formula is: ;

[0144] Similarly, the phase is calibrated as follows: After calibration, the amplitude and phase of four channels and two frequency points are obtained, forming a quadrature demodulation result set.

[0145] in, , For the first The discrete sine and cosine reference sequences of each excitation frequency point, i.e., the quadrature demodulation reference waveforms, are used to separate the effective signal at the excitation frequency point and filter out irrelevant noise.

[0146] For the first Road test excitation frequency; The discrete index of the sampling point; The sampling period;

[0147] For the first injection point at frequency The original voltage amplitude obtained by demodulation is the original voltage amplitude at a single frequency point, which characterizes the magnitude of the loop voltage response at that frequency point;

[0148] It is the square of the in-phase component of the voltage; The square of the voltage quadrature component; The first injection terminal frequency point The original voltage phase characterizes whether the voltage lags or leads the reference signal in phase.

[0149] Given the arctangent function, the phase is determined from the orthogonal and in-phase components;

[0150] The calibrated voltage amplitude is the accurate voltage amplitude after compensating for hardware errors, and is used for calculating the actual load impedance. The voltage calibration constant term coefficients are used as hardware zero-input offset compensation parameters. This is the voltage primary calibration coefficient, used to correct amplitude measurement errors by following the estimated grounding resistance;

[0151] Estimated initial value for grounding resistance; The measured original amplitude obtained from demodulation; This is the voltage calibration offset constant;

[0152] To calibrate and correct the voltage phase, it is to ensure the accurate voltage phase after compensating for hardware phase shift, thus guaranteeing the accurate solution of complex impedance; The measured original phase obtained from demodulation; For frequency point The corresponding phase compensation offset;

[0153] The four channels acquire four signals: the first injection terminal voltage, the first injection terminal current, the second injection terminal voltage, and the second injection terminal current.

[0154] The quadrature demodulation result set includes the calibrated voltage amplitude for each channel and each frequency point. Voltage phase The calibrated current amplitude and current phase are used to match the original components in phase or quadrature, the estimated grounding resistance, and the calibration coefficients for each group.

[0155] The coarse solution unit is used to obtain the coarse impedance value by least squares solution of the overdetermined equation system under the constraint of known PE line resistance, based on the orthogonal demodulation result set and impedance matching set. This includes the photovoltaic insulation resistance to ground, the parasitic capacitance of the photovoltaic DC bus to ground, the equivalent resistance of the local grounding electrode, and the parallel grounding resistance on the grid side. It serves as the initial value for the iterative compensation and calibration feedback loop of the capacitive component, and is used to iteratively correct the error caused by parasitic capacitive shunting to obtain a high-precision impedance fine solution.

[0156] In practice, the impedance matching set is invoked to perform inherent impedance compensation on the quadrature demodulation results. Taking the first injection terminal as an example, the compensated load impedance is: ;

[0157] Similarly, calculate Establish a circuit model with parameters including: insulation resistance, DC capacitance to ground, PE line resistance (known from factory calibration and not considered an unknown), local grounding electrode resistance, and grid parallel grounding resistance (approximately purely resistive). According to Kirchhoff's current law, the load impedances at the two injection terminals satisfy the following at each frequency:

[0158] ;

[0159] ;

[0160] Where the insulation resistance is: ;

[0161] There are four complex equations (two frequency points and two injection terminals), which is equivalent to eight real equations. There are four real unknowns to be solved, and the system of equations is overdetermined. Construct the objective function:

[0162] ;

[0163] The Levenberg-Marquardt algorithm is used to find the minimum value. The initial value for the iteration is set to... , , , The convergence condition is that the iteration exceeds 100 times. After convergence, the output coarse solution values ​​are: coarse solution value of photovoltaic insulation resistance to ground, coarse solution value of photovoltaic DC bus to ground parasitic capacitance, coarse solution value of equipment grounding electrode resistance, and coarse solution value of grid-side parallel grounding resistance.

[0164] in For the first The load complex impedance after deducting the inherent hardware impedance at the first injection terminal of the frequency point is used as the observation value of the circuit equation to establish the correlation between the measured response and the grounding insulation parameters.

[0165] , For the first Voltage phasor and current phasor after calibration at the first injection terminal of the frequency point;

[0166] The first injection channel has an inherent complex impedance. For the first The load complex impedance after deducting the inherent hardware impedance at the second injection terminal of the frequency point;

[0167] For the first The equivalent complex impedance of the photovoltaic-to-ground insulation branch at a frequency point characterizes the leakage current characteristics of the photovoltaic panel to ground and distinguishes between resistive insulation leakage and capacitive displacement current.

[0168] Fixed resistor for PE protection line (factory calibration known); The equivalent resistance of the local grounding electrode (an unknown to be solved); Let be the equivalent resistance of the power grid connected to ground in parallel (an unknown to be solved). The photovoltaic insulation resistance to ground is an unknown variable to be solved. The photovoltaic DC parasitic capacitance to ground (unknown to be solved);

[0169] The objective function for the total residual of the least squares fitting is the sum of squares of the differences between the measured impedance and the theoretical impedance for all frequencies and all channels. A set of coarse solutions for the optimal grounding and insulation unknown parameters is obtained by minimizing F.

[0170] The channel number at the injection end; The measured load impedance at the k-th frequency point of the m-th channel; The theoretical load impedance is calculated from unknown parameters, that is, the theoretical impedance function with the four major grounding insulation parameters as independent variables;

[0171] This is a parallel operation symbol for circuits;

[0172] The iterative unit is used to obtain the impedance solution result set by joint iterative optimization based on the coarse impedance solution value through capacitive component shunting and calibration parameter feedback.

[0173] Iteration units include:

[0174] The updated sub-unit is used to take the coarse impedance solution as the initial value for iteration. Through capacitive reactance calculation and shunt ratio analysis, the purely resistive current component is separated and the coarse impedance solution is updated.

[0175] In practice, the coarse solution value is used as the initial value for iteration, and capacitive component refinement is performed. In each iteration, the capacitive reactance of the DC-to-ground capacitance at the two test frequency points is first calculated: The ratio of capacitive current to resistive current in the insulated branch is obtained. ;in For the first Capacitive reactance corresponding to the parasitic capacitance of photovoltaic cells to ground at each test frequency point; For the first The angular frequencies corresponding to each test frequency point; This refers to the capacitive current flowing through the capacitance to ground in the insulated branch; It is the resistive leakage current flowing through the insulation resistance in the insulated branch;

[0176] Based on the shunt ratio, the purely resistive current component flowing through the grounding electrode is separated from the total current, and the grounding resistance value is recalculated: ;in This refers to the local grounding electrode resistance after this round of iterations and updates; This represents the real part of the complex vector of the second injection terminal voltage; It represents the real part of the complex vector of the purely resistive current in the grounding branch;

[0177] Similarly, update the insulation resistance value. The DC-to-ground capacitance value is updated by the ratio of the capacitive currents at the two frequency points. .

[0178] The convergence determination subunit is used to obtain the impedance solution set by back-feeding back the calibration parameters and re-demodulating the solution based on the updated equivalent resistance of the local ground electrode, combined with the relative deviation convergence determination.

[0179] In practice, after obtaining the updated equivalent resistance of the local grounding electrode, this is used as the new estimated grounding resistance and fed back to the calibration stage. The voltage and current calibration coefficients are recalculated, and orthogonal demodulation and calibration correction are re-performed on the original sampled data. The load impedance is then updated, and the least squares solution is redone. The relative deviation between this round and the previous round is calculated. ;in The relative deviation is calculated based on multiple parameters and used to determine whether the iteration has converged. , , , The new local grounding electrode resistance, photovoltaic-to-ground insulation resistance, photovoltaic-to-ground parasitic capacitance, and voltage calibration coefficient are calculated for this iteration.

[0180] like If the threshold is exceeded, the iteration continues; otherwise, the iteration terminates. After the iteration terminates, the final refined value, along with the convergence flag and the objective function residual, is output, forming the impedance solution result set.

[0181] In this embodiment, a three-level architecture of demodulation, coarse solution, and iterative processing is adopted. From frequency domain component extraction and solving of overdetermined equations to capacitive iterative refinement, a complete solution chain is formed, achieving high-precision decoupling of grounding and insulation impedance. The demodulation unit adopts dual-frequency orthogonal demodulation and resistance-related calibration. For example, it calls the amplitude and phase calibration set to dynamically adjust the gain coefficient according to the estimated resistance, instead of using a fixed gain, which can compensate for hardware gain drift under different loads.

[0182] The coarse solution unit innovatively introduces the known constraint of PE line resistance. For example, it constructs a set of 8 real equations with 2 frequency points and 2 injection terminals to solve an overdetermined system of equations with 4 unknowns. Combined with the Levenberg-Marquardt algorithm to minimize the residuals, the anti-interference ability is significantly enhanced and the noise robustness is doubled compared with conventional underdetermined or just-determined systems of equations. At the same time, it solves four types of parameters at once: insulation resistance, capacitance to ground, grounding resistance, and parallel resistance of the power grid, without the need for additional measurements.

[0183] The iterative unit adopts a joint iterative mechanism of capacitive current shunt separation and calibration feedback. For example, in each round, the pure resistive current is separated first to update the impedance parameters, and then the new resistance value is fed back to the calibration link to demodulate and solve again. The convergence is determined by the relative deviation of the four parameters, which can effectively eliminate the capacitive shunt distortion of DC to ground parasitic capacitance.

[0184] In a preferred embodiment of the present invention, the RCD collaboration module includes:

[0185] The feature unit is used to construct a four-dimensional feature vector for identifying zero-sequence current test components based on the preprocessed sampled dataset and the shunt amplitude through orthogonal demodulation extraction and theoretical value calculation.

[0186] The component identification unit is used to identify the test excitation component in the zero-sequence current and output an identification flag based on the four-dimensional feature vector and the factory calibration template parameters, through Mahalanobis distance calculation and threshold determination.

[0187] In practice, zero-sequence current channel data is extracted from the preprocessed sampled data. Using the same quadrature demodulation method as the previous step, the amplitude and phase of the zero-sequence current at the two test frequencies are extracted. The shunt amplitude, along with the test current and phase at the second injection terminal, are used to calculate the theoretical value of the test component in the zero-sequence current. ; Construct a four-dimensional feature vector: ;

[0188] Template parameters were obtained from factory calibration. In the final testing phase, twenty measurements were performed under a 10Ω standard grounding resistance. The eigenvector was calculated for each measurement, and the mean vector and covariance matrix were statistically obtained. The positive definiteness of the covariance matrix was verified by Cholesky decomposition. The Mahalanobis distance between the measured eigenvector and the template was calculated. The Mahalanobis distance is used to measure the degree of matching between the measured features and the factory template; where This is the four-dimensional mean vector of the factory calibration template. The transpose of the eigenvalue deviation vector. To calibrate the four-dimensional covariance matrix at the factory, It is the inverse of the covariance matrix;

[0189] If Mahalanobis distance If the distance is less than the preset distance threshold, it is determined to be a test excitation component, and the test identification flag is set, indicating that the excessive component in the current zero-sequence current is determined to be a test excitation signal injected by the equipment itself, and not a real leakage fault on site; the amplitude estimation values ​​of the two frequency test components are output synchronously for subsequent dynamic adjustment of the leakage protection threshold; otherwise, it is determined to be a real leakage, and the flag is cleared and reset; it means that a real leakage has been detected, and the protection threshold cannot be raised to prevent leakage tripping.

[0190] in The theoretical amplitude of the zero-sequence current test component at the k-th frequency point; The measured current amplitude at the k-th frequency point of the second injection terminal; The bypass shunting coefficient at the k-th frequency point; The theoretical phase of the zero-sequence test component at the k-th frequency point; The measured phase of the current at the k-th frequency point of the second injection terminal; This represents the inherent phase offset of the bypass channel at the k-th frequency. It is a four-dimensional feature vector; , The theoretical amplitude of the zero-sequence current test component at different frequencies; , The measured amplitudes of the second injection terminal current at different frequencies; , The measured phase of the zero-sequence current at different frequencies; , The measured phase of the second injection terminal current at different frequencies;

[0191] The threshold raising unit is used to adaptively raise the residual current protection threshold based on the test excitation component in the zero-sequence current and the original protection threshold, through pre-identification of dual threshold determination and dynamic threshold adjustment of the measurement window, and to form a protection threshold adjustment record.

[0192] Specific execution steps: Pre-identification is initiated 10ms before the test injection, and zero-sequence current background noise is collected for 5ms. Pre-identification uses a dual-threshold judgment: the first level calculates the effective noise value within the 5ms window. ;

[0193] The noise is then compared to 0.1 times the original protection threshold. If the noise exceeds the threshold, it is determined that the noise is too high, and the measurement is canceled. The second stage performs a Fourier transform on the noise to extract the 50Hz power frequency component. Calculate the proportion of power frequency: ;

[0194] like If the percentage exceeds the preset threshold, a potential power frequency leakage current hazard is identified, and the pre-identification fails. After both stages pass, formal measurement and threshold adjustment are initiated. The maximum test component is calculated based on the test excitation component in the zero-sequence current. ;

[0195] Set threshold adjustment amount Dynamic threshold: ;

[0196] The upper limit is three times the original threshold. A threshold register of the RCD module is written 20ms before the test injection. During the injection, the amplitude of the test component is continuously monitored; if it exceeds the estimate by 120%, the threshold is updated in real time. After the test, there is a 40ms delay to allow the signal to completely attenuate before restoring the original threshold. Timing is precisely controlled by a hardware timer. The original threshold, dynamic threshold, and adjustment duration are recorded to form a protection threshold adjustment record.

[0197] in The effective value of zero-sequence background noise current within a 5ms acquisition window; This represents the total number of sampling points within a 5ms noise acquisition window. The instantaneous value of the discrete sampling of the zero-sequence current at point n; The proportion of the 50Hz power frequency leakage current component in the total effective value of the noise. The amplitude of the 50Hz power frequency zero-sequence current separated from the noise; The effective value of the zero-sequence background noise current collected in the 5ms pre-identification window is used to determine whether the ambient noise is too high before measurement. The sum of the zero-sequence test component amplitudes at the two test frequencies represents the maximum interference current injected into the test signal. , Estimate the amplitude of the zero-sequence channel test components at different frequency points; To protect the threshold rise compensation amount, used to offset test injection current interference; The residual current tripping threshold after dynamic elevation during the measurement process is the adaptively raised residual current protection threshold. The original residual current protection trip threshold is fixed at the factory when the equipment is manufactured;

[0198] The leakage current handling unit is used to monitor the effective value and rate of change of the residual current in real time within the measurement window based on the adaptively raised residual current protection threshold and real-time residual current monitoring. If the rate of change exceeds the preset change threshold, it is determined to be a real leakage current fast rising edge, and a fast recovery is immediately executed, including shutting down the signal generator, disconnecting the injection switch and bypass switch, switching back to the original threshold and triggering the protection trip; so as to realize the rapid identification of real leakage current anomalies and system recovery actions.

[0199] The calibration unit calculates the average value of the zero-sequence current after the test as a new baseline and compares it with the old baseline. If the deviation rate exceeds a preset deviation threshold, the baseline is updated and written to the RCD offset calibration register. The calibrated baseline affects the zero-sequence current processing of the next measurement, forming a closed loop. It outputs protection status flags, baseline calibration results, and fast recovery trigger flags to obtain the RCD collaborative control result set. This set is used to output the leakage current protection collaborative control status of the entire insulation test process to the upper-level master controller, and is also used for fault log storage, maintenance alarms, and insulation diagnostic logic linkage.

[0200] In practice, a sliding window is used to calculate the effective value, with the window being one power frequency cycle and the step being one sampling point.

[0201] If three consecutive windows exceed the dynamic threshold, protection is triggered. For the first The effective value of the residual current corresponding to each sliding window; This represents the total number of sampling points within a single power frequency cycle. This is the instantaneous sampling signal of the zero-sequence current; This is the index number of the current sliding window; This is the offset of the sampling points within the window, with a value ranging from 0 to... ;

[0202] In this embodiment, a complete protection coordination link is formed from test component identification, dynamic threshold adjustment, rapid leakage current recovery to baseline closed-loop calibration, which solves the problem of RCD false tripping caused by test injection.

[0203] The feature and component identification unit uses four-dimensional feature vectors and Mahalanobis distance matching. For example, it constructs a four-dimensional feature of the amplitude ratio of two frequency points plus the phase difference of two frequency points. Combined with the mean and covariance template of the 20 factory calibration statistics, the test component is determined by Mahalanobis distance, which further improves the identification accuracy. Compared with the simple amplitude comparison scheme, it can effectively distinguish the test signal from the real leakage current and avoid misjudgment.

[0204] The threshold raising unit adopts a pre-identification dual threshold and dynamic adjustment mechanism. For example, it first eliminates leakage hazards by dual judgment of the effective value of noise and the proportion of power frequency, and then raises the protection threshold by 1.5 times the test component and not more than three times the upper limit. This ensures that there is no false tripping during the test and leaves sufficient safety margin.

[0205] The leakage current handling unit employs rapid rate-of-change determination; for example, a 100A / s rate of change triggers a 100μs rapid recovery. The signal shutdown, switch disconnection, threshold switching back, and protection triggering are executed synchronously, resulting in a protection response speed an order of magnitude faster than conventional solutions in real leakage situations. The calibration unit utilizes a baseline adaptive closed-loop system; for example, it automatically updates the baseline when the deviation rate exceeds 5%, and automatically compensates for drift during long-term operation, ensuring the protection threshold remains consistently accurate.

[0206] In a preferred embodiment of the present invention, the detection and early warning module includes:

[0207] The confidence scoring unit is used to obtain a comprehensive confidence score based on the impedance solution result set, preprocessed sampled data and RCD collaborative control result set through a weighted average algorithm.

[0208] The completion unit is used to obtain a sequence of valid measurement values ​​by interpolation based on the credibility assessment results and historical database. This sequence is a set of time-series measurement data obtained after credibility screening and interpolation completion, including grounding resistance, insulation resistance, and photovoltaic capacitance to ground. It is divided into two categories: original effective values ​​and completed effective values.

[0209] The graded early warning unit is used to monitor the comparison results of each parameter within the effective measurement value sequence with the corresponding early warning threshold in real time, take the highest trigger level as the current early warning level, and upload it to the cloud platform and user terminal to trigger audible and visual alarms and push notifications according to the early warning level.

[0210] In terms of signal-to-noise ratio, the ratio of the signal amplitude at two test frequency points to the average noise amplitude within 5 Hz around the frequency points is calculated, and the average of the two frequency points is input into the scoring function.

[0211] Dual-frequency consistency dimension: Compare the grounding resistance values ​​calculated independently at the two frequency points and calculate the relative deviation;

[0212] Impedance phase rationality dimension: The grounding electrode impedance should be purely resistive, and the absolute value of the phase should be calculated;

[0213] Capacitor rationality dimension: Call the final calculated value of the photovoltaic equivalent capacitance to ground, determine whether it is within the reasonable range of 100nF to 10μF, and calculate the degree of deviation from the typical median value;

[0214] Each dimension is mapped to a score from 0 to 100 using a piecewise linear scoring function; a weighted average algorithm is then used to obtain a comprehensive credibility score; if the score is lower than a preset credibility threshold, it is considered abnormal data. In case of anomalies, valid records before and after the current moment are retrieved from the historical database (stores valid records from the last 30 days). If valid records exist both before and after the current moment, linear interpolation is used to obtain a sequence of valid measurements.

[0215] Based on valid measurements and reliability results, a three-level early warning judgment is executed. Three-level thresholds are set for grounding resistance and insulation resistance, and the highest trigger level is taken as the current early warning level.

[0216] Valid data from the most recent 30 days is extracted from the historical database. A single exponential smoothing method is used to predict the trend for the next 7 days. The estimated number of days to reach each warning threshold is calculated based on the trend slope. ,in To predict the time required to reach the alarm threshold; The alarm threshold corresponding to the grounding resistance level; For current effective measurement of grounding resistance; The slope of the grounding resistance change is fitted from historical data; if the slope is less than or equal to 0, the grounding resistance has no upward or deteriorating trend, and no trend warning is calculated.

[0217] A negative slope will not trigger a trend warning; a positive slope with an expected number of days less than 30 will trigger a trend warning. A measurement result data structure is generated, including current grounding resistance, insulation resistance, DC-to-ground capacitance, confidence score, warning level, trend parameters, expected alarm days, and timestamp. Results are displayed locally on a screen (numerical values, level indication, and a seven-day trend chart) and uploaded to the cloud platform and user terminals. Level 2 and above trigger buzzers and indicator lights; Level 3 and above push alarm messages. When the user enables linkage protection, a power limiting or grid disconnect command is sent to the inverter control module. All measurement data is written to the historical database, covering the oldest record exceeding 30 days.

[0218] In this embodiment, a complete output chain is formed from data quality assessment and anomaly completion to hierarchical alarms, ensuring reliable measurement results and timely and effective early warnings.

[0219] The confidence scoring unit innovatively constructs a five-dimensional scoring system, which combines five dimensions such as signal-to-noise ratio, dual-frequency consistency, impedance phase rationality, capacitance rationality, and protection status abnormality to calculate the comprehensive confidence level. Compared with a single threshold judgment scheme, it can comprehensively evaluate the quality of measurement data.

[0220] The data completion unit employs a reliability screening and historical data completion mechanism. For example, a reliability score below 60 is considered abnormal, and data is completed through linear interpolation of valid records before and after the data completion point or extrapolation of 7 days of data, rather than directly discarding abnormal data, ensuring the continuity of monitoring data. Three consecutive abnormalities automatically trigger system self-checks, enabling proactive fault detection. The graded early warning unit combines three-level thresholds with trend prediction. For example, three early warning thresholds are set for grounding and insulation, combined with a single exponential smoothing method to predict the trend over the next 7 days and calculate the expected number of alarm days, allowing for early detection of deterioration trends rather than just post-event alarms. It also supports local display, cloud upload, audible and visual alarms, and multi-channel graded response for linked protection. Level 2 and above trigger audible and visual alarms, and level 3 and above push alarms, which can be linked to inverter power limiting or grid disconnection. The entire solution ensures data reliability and achieves proactive early warning, significantly improving the efficiency and safety of power plant operation and maintenance.

[0221] Example 2

[0222] Please refer to Figure 2 Specifically: A method for monitoring the grounding resistance of a photovoltaic grid-connected inverter, comprising the following steps:

[0223] S100: Collects grid-connected voltage sampling signals, identifies the interference distribution of power frequency and switching harmonics through fast Fourier spectrum analysis, and obtains excitation frequency points and amplitude-phase calibration sets;

[0224] S200: Construct a dual-ended injection loop and, in conjunction with the excitation frequency, obtain the impedance matching set through injection loop impedance matching calibration and shunt ratio parameter measurement;

[0225] S300: A second-order generalized integrator phase-locked loop is used to track the grid phase and perform quadrature demodulation cancellation to obtain a preprocessed sampled dataset;

[0226] S400: Based on the preprocessed sampling dataset and amplitude-phase calibration set, the impedance calculation result set is obtained by coarsely solving the double-ended injection impedance with known PE line resistance constraints and iteratively compensating for the capacitive component.

[0227] S500: Based on the preprocessed sampled dataset, abnormal operating conditions are recovered through the identification and pre-identification of the test components of Mahalanobis distance, so as to obtain the RCD collaborative control result set;

[0228] S600: Based on the RCD collaborative control result set, impedance calculation result set, and preprocessed sampling dataset, it performs abnormal data removal and presents the results to the cloud platform and user terminal through hierarchical early warning, so as to realize real-time monitoring of photovoltaic grid-connected inverters.

[0229] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A monitoring system for the grounding resistance of a photovoltaic grid-connected inverter, characterized in that, The system includes: The excitation calibration module is used to acquire the grid connection point voltage sampling signal, identify the interference distribution of power frequency and switching harmonics through fast Fourier spectrum analysis, and obtain the excitation frequency point and amplitude-phase calibration set; The impedance calculation module is used to construct a two-terminal injection loop and, in conjunction with the excitation frequency, obtain the impedance matching set through injection loop impedance matching calibration and shunt ratio parameter determination. The preprocessing module is used to track the grid phase using a second-order generalized integrator phase-locked loop and perform orthogonal demodulation cancellation to obtain a preprocessed sampled dataset. The impedance calculation module is used to obtain the impedance calculation result set based on the preprocessed sampling dataset and amplitude-phase calibration set, through the coarse solution of the double-ended injection impedance constrained by the known PE line resistance, and after iterative compensation of the capacitive component. The RCD collaborative module is used to recover abnormal operating conditions based on the preprocessed sampled dataset, through the identification and pre-identification of the test components of Mahalanobis distance, so as to obtain the RCD collaborative control result set. The detection and early warning module is used to perform abnormal data removal based on the RCD collaborative control result set, impedance calculation result set, and preprocessed sampling dataset, and present the results to the cloud platform and user terminal through hierarchical early warning, so as to realize real-time monitoring of photovoltaic grid-connected inverters.

2. The monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 1, characterized in that, The excitation calibration module includes: The frequency interference extraction unit is used to perform Fourier transform on the grid connection point voltage sampling signal and extract all interference frequency points and amplitudes of power frequency harmonics and switching harmonics through spectrum analysis; The frequency extraction unit is used to generate candidate frequency points based on the low-frequency signal injection method, and for each candidate frequency point, calculate the minimum frequency interval between it and each interference frequency point, and filter out candidate frequency points whose interval with all interference frequency points is greater than a preset interval threshold, generating a candidate frequency point set. From the candidate frequency point set, select the two candidate frequency points with the lowest background noise amplitude and whose frequency difference between the two candidate frequency points is not less than a preset frequency difference threshold, and then use them as excitation frequency points. The waveform unit is used to generate a test waveform data table that can be loaded into a signal generator based on the excitation frequency through orthogonal encoding and discretization sampling. The calibration unit is used to obtain an amplitude and phase calibration set that varies with resistance value based on the test waveform data table and a multi-resistance standard resistance box, through multi-point synchronous sampling and frequency domain component extraction, combined with the least squares linear fitting algorithm.

3. The monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 2, characterized in that, Impedance calculation module, including: The loop construction unit is used to construct a DC-side and PE-side dual-ended injection loop in the photovoltaic inverter. Each loop is connected in series with a DC blocking capacitor, a current limiting resistor, an optocoupler relay and a sampling resistor. They share a signal source through an isolation transformer to obtain a hardware configuration table for impedance calculation. In the dual-ended injection circuit, the first injection terminal is connected between the positive terminal of the DC bus and the ground reference, and the second injection terminal is connected between the PE terminal and the ground reference. A bridging topology unit is used to construct a hardware topology for a test signal bridging branch based on two excitation frequencies by leading out a bridging branch from the PE line on the primary side of the residual current transformer and connecting frequency selection and control elements in series. The shunt unit is used to calculate the impedance values ​​of the bridging branch and the main PE branch at two excitation frequency points based on the hardware topology of the bridging branch of the test signal. The impedance value of the main PE branch is used as the numerator, and the sum of the impedance values ​​of the main PE branch and the bridging branch is used as the denominator to obtain the shunt amplitude. Impedance fitting unit is used to obtain the intrinsic impedance and mutual coupling impedance based on the apparent impedance measurement and least squares fitting of multiple standard load combinations, based on the double-ended injection circuit and the bridging branch, to form an impedance matching set.

4. The monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 3, characterized in that, The preprocessing module includes: The sampling unit is used to obtain the original sampling set according to the hardware configuration table by synchronous triggering and automatic DMA transfer when the ground impedance detection is started. The power frequency cancellation unit is used to track the power grid phase through a second-order generalized integrator phase-locked loop and perform orthogonal demodulation cancellation based on the original sampling set, to obtain the power frequency desampled dataset and the power grid phase sequence. The switching frequency detection unit is used to extract grid voltage channel data from the de-power frequency sampling dataset, and search for the frequency point with the maximum amplitude near the nominal switching frequency through Fourier transform to determine the actual switching frequency; The harmonic filtering unit is used to obtain a preprocessed sampling dataset based on the actual switching frequency, through a three-stage adaptive IIR notch filter and a fourth-order Butterworth low-pass filter, combined with attenuation verification and adjustment.

5. A monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 4, characterized in that, Impedance calculation module, including: The demodulation unit is used to obtain the voltage and current amplitude and phase parameters of four channels at two frequencies by using dual-frequency orthogonal demodulation and resistance-related calibration correction based on the preprocessed sampled dataset and amplitude and phase calibration set, thus forming an orthogonal demodulation result set. The coarse solution unit is used to obtain the coarse impedance values ​​by least squares solution of the overdetermined equations under the constraint of known PE line resistance, based on the orthogonal demodulation result set and impedance matching set. These values ​​include the photovoltaic insulation resistance to ground, the parasitic capacitance of the photovoltaic DC bus to ground, the equivalent resistance of the local grounding electrode, and the parallel grounding resistance on the grid side. The iterative unit is used to obtain the impedance solution result set by joint iterative optimization based on the coarse impedance solution value through capacitive component shunting and calibration parameter feedback.

6. A monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 5, characterized in that, Iteration units include: The updated sub-unit is used to take the coarse impedance solution as the initial value for iteration. Through capacitive reactance calculation and shunt ratio analysis, the purely resistive current component is separated and the coarse impedance solution is updated. The convergence determination subunit is used to obtain the impedance solution set by back-feeding back the calibration parameters and re-demodulating the solution based on the updated local grounding electrode equivalent resistance, combined with the relative deviation convergence determination.

7. A monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 6, characterized in that, The RCD collaboration module includes: The feature unit is used to construct a four-dimensional feature vector for identifying zero-sequence current test components based on the preprocessed sampled dataset and the shunt amplitude through orthogonal demodulation extraction and theoretical value calculation. The component identification unit is used to identify the test excitation component in the zero-sequence current and output an identification flag based on the four-dimensional feature vector and the factory calibration template parameters, through Mahalanobis distance calculation and threshold determination. The threshold raising unit is used to adaptively raise the residual current protection threshold based on the test excitation component in the zero-sequence current and the original protection threshold, through pre-identification of dual threshold determination and dynamic threshold adjustment of the measurement window, and to form a protection threshold adjustment record.

8. A monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 7, characterized in that, The RCD collaboration module also includes: The leakage current handling unit is used to monitor the effective value and rate of change of the residual current in real time within the measurement window based on the adaptively raised residual current protection threshold and real-time residual current monitoring. If the rate of change exceeds the preset change threshold, it is determined to be a real leakage current fast rising edge, and fast recovery is immediately performed, including shutting down the signal generator, disconnecting the injection switch and bypass switch, switching back to the original threshold and triggering the protection trip. The calibration unit is used to calculate the average value as a new baseline based on the zero-sequence current after the test, and compare the deviation rate with the old baseline. If the deviation rate exceeds the preset deviation threshold, the baseline is updated and written to the RCD offset calibration register to obtain the RCD cooperative control result set.

9. A monitoring system for the grounding resistance of a photovoltaic grid-connected inverter according to claim 8, characterized in that, The detection and early warning module includes: The confidence scoring unit is used to obtain a comprehensive confidence score based on the impedance solution result set, preprocessed sampled data and RCD collaborative control result set through a weighted average algorithm. The completion unit is used to obtain a valid measurement value sequence by interpolation based on the credibility assessment results and historical database; The graded early warning unit is used to monitor the comparison results of each parameter within the effective measurement value sequence with the corresponding early warning threshold in real time, take the highest trigger level as the current early warning level, and upload it to the cloud platform and user terminal to trigger audible and visual alarms and push notifications according to the early warning level.

10. A method for monitoring the grounding resistance of a photovoltaic grid-connected inverter, used to implement the monitoring system for the grounding resistance of a photovoltaic grid-connected inverter as described in any one of claims 1 to 9, characterized in that, Includes the following steps: Collect grid connection point voltage sampling signals, identify the interference distribution of power frequency and switching harmonics through fast Fourier spectrum analysis, and obtain excitation frequency points and amplitude-phase calibration sets; A dual-ended injection loop is constructed, and by combining the excitation frequency, the impedance matching set is obtained through injection loop impedance matching calibration and shunt ratio parameter measurement. A second-order generalized integrator phase-locked loop is used to track the grid phase and perform orthogonal demodulation cancellation to obtain a preprocessed sampled dataset; Based on the preprocessed sampling dataset and amplitude-phase calibration set, the impedance calculation result set is obtained by coarsely solving the double-ended injection impedance with known PE line resistance constraints and iteratively compensating for the capacitive component. Based on the preprocessed sampled dataset, abnormal operating conditions are recovered by identifying and pre-identifying the test components of Mahalanobis distance, so as to obtain the RCD collaborative control result set; Based on the RCD collaborative control result set, impedance calculation result set, and preprocessed sampling dataset, abnormal data is removed and presented to the cloud platform and user terminal through hierarchical early warning, so as to realize real-time monitoring of photovoltaic grid-connected inverters.

Images