Inverter fault detection method and inverter fault detection system

Abstract

The application provides an inverter fault detection method and an inverter fault detection system. The method comprises the following steps: obtaining an inverted first ground impedance according to first sampling data of an inverter, and determining a first fault detection result of the inverter according to the first ground impedance and a first fault threshold; in the case that the first fault detection result represents that no fault of the inverter is detected, identifying an operating condition of the inverter based on the first sampling data; the operating condition represents an electrical state and a disturbance degree of the inverter; predicting a ground insulation state reference value of the inverter based on the first sampling data; updating a second fault threshold according to the ground insulation state reference value and the operating condition, and determining a second fault detection result of the inverter according to the first sampling data, the ground insulation state reference value and the second fault threshold. The method provided by the application improves the reliability of inverter fault detection.

Description

Technical Field

[0001] This application relates to the field of inverter insulation testing technology, and in particular to an inverter fault detection method and inverter fault detection system. Background Technology

[0002] During the long-term grid-connected operation of photovoltaic power plants, the insulation performance of the photovoltaic module encapsulation layer, DC cable sheath, and internal insulation medium of the inverter will deteriorate due to factors such as ultraviolet radiation, diurnal temperature difference, changes in ambient humidity, and material fatigue and aging, resulting in a decrease in the ground insulation impedance of the inverter circuit.

[0003] In inverter fault detection, ground fault detection methods such as bridge measurement and leakage current detection are commonly used. These methods rely on shutdown detection or fixed threshold judgment during operation to detect the insulation status of the DC side. Therefore, the fault detection methods in these technologies are difficult to adapt to parameter drift caused by bus voltage fluctuations, switching actions, and environmental changes, resulting in low reliability of fault detection. Summary of the Invention

[0004] This application provides an inverter fault detection method and an inverter fault detection system to solve the problem of low reliability in fault detection.

[0005] In a first aspect, this application provides an inverter fault detection method, the method comprising:

[0006] The first ground impedance is obtained by inverting the first sampling data of the inverter, and the first fault detection result of the inverter is determined based on the first ground impedance and the first fault threshold.

[0007] If the first fault detection result indicates that no fault is detected in the inverter, the operating condition of the inverter is identified based on the first sampling data; the operating condition represents the electrical state and disturbance level of the inverter.

[0008] Based on the first sampled data, a reference value for the inverter's insulation state to ground is predicted; the reference value for the insulation state to ground includes: a reference value for the equivalent inductance voltage on the secondary side of the potential-induced decay PID circuit, a reference value for the leakage current, a reference value for the impedance to ground, and a prediction uncertainty;

[0009] The second fault threshold is updated based on the ground insulation status reference value and the operating condition, and the second fault detection result of the inverter is determined based on the first sampling data, the ground insulation status reference value and the second fault threshold.

[0010] Optionally, before updating the second fault threshold based on the ground insulation state reference value and the operating condition, the method further includes:

[0011] Based on the prediction uncertainty, determine whether to apply an coded diagnostic pulse to the PID circuit;

[0012] When it is determined that the coded diagnostic pulse is applied to the PID circuit, the PID circuit is controlled to apply the coded diagnostic pulse; wherein the coded diagnostic pulse satisfies at least one of amplitude constraint, duration constraint, total energy constraint, and positive and negative volt-second balance constraint.

[0013] The second sampling data of the inverter is collected to update the reference value of the insulation status to ground.

[0014] Optionally, the first sampled data includes DC bus voltage, inverter output power, and leakage current; the control of the PID circuit to apply the coded diagnostic pulse includes:

[0015] When the inverter meets the safety constraints, the PID circuit is controlled to apply the coded diagnostic pulse.

[0016] The safety constraints include at least one of the following: inverter protection status flag constraints, DC bus voltage constraints, inverter output power constraints, leakage current constraints, and PID circuit repair function constraints.

[0017] Optionally, the second fault threshold includes a ground impedance threshold; before updating the second fault threshold based on the ground insulation state reference value and the operating conditions, the method further includes:

[0018] Based on the operating conditions, determine whether to update the second fault threshold;

[0019] The step of updating the second fault threshold based on the ground insulation state reference value and the operating conditions includes:

[0020] If it is determined that the second fault threshold needs to be updated, the second fault threshold is updated based on the ground insulation status reference value and the operating conditions.

[0021] Optionally, the first sampling data includes the sampled value of the equivalent inductance voltage on the secondary side of the PID circuit and the sampled value of the leakage current; updating the second fault threshold according to the ground insulation state reference value and the operating conditions includes:

[0022] The inverted second ground impedance is obtained based on the first sampled data;

[0023] An anomaly score is obtained based on at least one of the following: the difference between the second impedance to ground and the reference value of the impedance to ground, the difference between the sampled value of the secondary equivalent inductance voltage and the reference value of the secondary equivalent inductance voltage, and the difference between the sampled value of the leakage current and the reference value of the leakage current. The anomaly score is then added to the anomaly score set.

[0024] Based on the operating conditions, obtain the conditional quantiles from the abnormal score set that characterize the upper limit of the statistical tolerance of abnormal scores when the insulation is normal under the operating conditions.

[0025] The threshold correction amount is obtained based on the conditional quantiles and the prediction uncertainty;

[0026] Update the second fault threshold according to the threshold correction amount;

[0027] The step of determining the second fault detection result of the inverter based on the first sampled data, the ground insulation state reference value, and the second fault threshold includes:

[0028] The second fault detection result of the inverter is determined based on at least one of the following: the difference between the second impedance to ground and the reference value of the impedance to ground; the difference between the sampled value of the secondary equivalent inductance voltage and the reference value of the secondary equivalent inductance voltage; the difference between the sampled value of the leakage current and the reference value of the leakage current; the changing trend of the reference value of the impedance to ground; and the second fault threshold.

[0029] Optionally, the second fault threshold includes a ground impedance threshold; the second fault detection result of the inverter is determined based on the difference between the second ground impedance and the ground impedance reference value, and the changing trend of the ground impedance reference value, including:

[0030] If the difference between the second impedance to ground and the reference value of the impedance to ground is greater than or equal to the impedance to ground threshold, then the second fault detection result is determined to be a fault detection result characterizing that the inverter has an impedance to ground fault.

[0031] If the difference between the ground impedance and the ground impedance reference value is less than the ground impedance threshold, and the ground impedance reference value shows a continuous downward trend, then the second fault detection result is determined to be a fault detection result that characterizes the inverter having an insulation degradation trend.

[0032] Optionally, the method further includes:

[0033] The inverted third ground impedance is obtained based on the first sampled data;

[0034] Based on the error signal between the first sampled data and the reference value of the insulation state to ground, and the second fault threshold, a fast time scale cumulative criterion, a medium time scale cumulative criterion, and a slow time scale cumulative criterion are constructed.

[0035] For each time scale, a recursive accumulation operation with a forgetting coefficient is performed on the positive deviation component of the error signal that exceeds the second fault threshold to obtain the recursive accumulation result for the corresponding time scale; wherein, the recursive accumulation result of the fast time scale is used to identify sudden grounding faults; the recursive accumulation result of the medium time scale is used to identify continuous anomalies; and the recursive accumulation result of the slow time scale is used for long-term insulation degradation analysis.

[0036] The level of insulation degradation trend is determined based on the recursive cumulative results of the fast time scale, the recursive cumulative results of the medium time scale, and the recursive cumulative results of the slow time scale corresponding to the decreasing slope and decreasing acceleration of the third ground impedance.

[0037] Optionally, the first sampling data includes high-frequency local feature data and low-frequency trend feature data; the high-frequency local feature data includes the sampled value of the equivalent inductance voltage on the secondary side of the PID circuit and the sampled value of the leakage current; the low-frequency trend feature data includes the sampled value of the DC bus voltage, the sampled value of the positive bus to ground voltage, the sampled value of the negative bus to ground voltage, the inverter output power, the pulse width modulation switching state, and the sampled value of environmental parameters.

[0038] The step of predicting the reference value of the inverter's insulation state to ground based on the first sampled data includes:

[0039] High-frequency local features of the high-frequency local feature data are extracted using the first branch of the dual-branch time-series prediction model.

[0040] The low-frequency trend features of the low-frequency trend feature data are extracted by the second branch of the dual-branch time series prediction model.

[0041] The high-frequency local features and the low-frequency trend features are merged in the feature fusion layer of the dual-branch time series prediction model to obtain fused features;

[0042] Based on the fusion characteristics, the following are output by the multiple output heads of the dual-branch timing prediction model: the reference value of the secondary side equivalent inductance voltage, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty.

[0043] Optionally, the method further includes:

[0044] Based on the first sampling data, the ground insulation state reference value, and the second sampling data, a multi-feature diagnostic result is obtained; the multi-feature diagnostic result includes at least one of the following: the secondary side equivalent inductance voltage difference, the degree of voltage imbalance between the positive and negative busbars and ground, the leakage current difference, the trend of the ground impedance reference value, and the pulse response of the coded diagnostic pulse;

[0045] Based on the multi-feature diagnostic results, the probability of occurrence of each fault category is adjusted; the fault category includes at least one of the following: internal grounding fault, external DC circuit grounding fault, AC side and filter branch abnormality, parasitic capacitance change, bus transient disturbance, and sensor abnormality.

[0046] Secondly, this application provides an inverter fault detection system, which includes: a sampling module, an artificial intelligence recognition module, and a control module;

[0047] The sampling module is used to collect the first sampling data of the inverter;

[0048] The control module is used to obtain the inverted first ground impedance based on the first sampled data of the inverter, and to determine the first fault detection result of the inverter based on the first ground impedance and the first fault threshold.

[0049] The artificial intelligence recognition module is used to identify the operating condition of the inverter based on the first sampled data when the first fault detection result indicates that no fault is detected in the inverter; the operating condition indicates the electrical state and disturbance level of the inverter; and predicts the reference value of the inverter's insulation state to ground based on the first sampled data; the reference value of the insulation state to ground includes: the reference value of the equivalent inductance voltage on the secondary side of the potential-induced decay PID circuit, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty;

[0050] The control module is used to update the second fault threshold according to the ground insulation state reference value and the operating condition, and to determine the second fault detection result of the inverter according to the first sampling data, the ground insulation state reference value and the second fault threshold.

[0051] The inverter fault detection method and system provided in this application obtain the inverted first ground impedance based on the first sampled data of the inverter, and determine the first fault detection result of the inverter based on the first ground impedance and a first fault threshold. If the first fault detection result indicates that no fault is detected in the inverter, the operating condition of the inverter is identified based on the first sampled data. Based on the first sampled data, a reference value for the inverter's ground insulation state is predicted. A second fault threshold is updated based on the ground insulation state reference value and the operating condition, and a second fault detection result of the inverter is determined based on the first sampled data, the ground insulation state reference value, and the second fault threshold. In this application, a first-stage threshold determination is performed by inverting the first ground impedance and using the first fault threshold. Then, under the condition that no obvious fault is detected, the second fault threshold is dynamically updated based on the ground insulation state reference value, and a second-stage determination is performed. This enables the fault detection process to have adaptive operating condition capabilities, thereby improving the reliability of inverter fault detection. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0053] Figure 1 A circuit structure diagram of an inverter provided in this application embodiment. Figure 1 ;

[0054] Figure 2 A flowchart illustrating an inverter fault detection method provided in this application embodiment. Figure 1 ;

[0055] Figure 3 A flowchart illustrating an inverter fault detection method provided in this application embodiment. Figure 2 ;

[0056] Figure 4 This application provides a schematic diagram of a process for applying an coded diagnostic pulse.

[0057] Figure 5 A flowchart illustrating a reference value for predicting the insulation state to ground provided in an embodiment of this application;

[0058] Figure 6 A flowchart illustrating a fault type prediction method provided in an embodiment of this application;

[0059] Figure 7 This is a schematic diagram of the structure of an inverter fault detection system provided in an embodiment of this application;

[0060] Figure 8This is a schematic diagram of the structure of an inverter fault detection device provided in an embodiment of this application;

[0061] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0062] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0063] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0064] In inverter fault detection, ground fault detection methods such as bridge measurement and leakage current detection are commonly used. These methods rely on shutdown detection or fixed threshold judgment during operation to detect the insulation status of the DC side. Therefore, these fault detection methods are ill-suited to parameter drift caused by bus voltage fluctuations, switching actions, and environmental changes, resulting in low accuracy in fault detection.

[0065] In view of this, this application proposes an inverter fault detection method. First, a first-stage threshold determination is performed through a first-to-ground impedance inversion and a first fault threshold screening. Then, under the condition that no obvious fault is detected, a ground insulation state reference value is predicted based on the operating conditions. The ground insulation state reference value includes: the secondary-side equivalent inductance voltage reference value of the potential-induced degradation (PID) circuit, the leakage current reference value, the ground impedance reference value, and the prediction uncertainty. A second fault threshold is dynamically updated based on the ground insulation state reference value, and a second-stage determination is performed based on the first sampled data, the ground insulation state reference value, and the updated second fault threshold. This enables the fault detection process to have operating condition adaptive capabilities, realizing inverter fault detection in grid-connected continuous operation scenarios and improving the reliability of inverter fault detection.

[0066] It should be understood that the inverter fault detection method of this application embodiment can be used in any inverter fault detection scenario. The DC power output from the photovoltaic array passes through the DC-DC (Direct Current-to-Direct Current) boost stage and DC bus energy storage stage within the inverter, and is then converted into industrial frequency AC power and fed into the public power grid via the DC-AC (Direct Current-to-Alternating Current) inverter stage and grid-connected filter circuit. The inverter main circuit integrates a common-mode leakage current detection branch and a PID repair circuit. During non-repair periods, the PID repair circuit can be equivalent to an inductor and resistor network connected in parallel to ground. In this application embodiment, the PID repair circuit is the same as the PID circuit.

[0067] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0068] Figure 1 A circuit structure diagram of an inverter provided in this application embodiment. Figure 1 .like Figure 1 As shown, the photovoltaic array 101 is converted into alternating current by the photovoltaic inverter 102 and then connected to the power grid 103.

[0069] The photovoltaic inverter 102 includes a main circuit 201, a sampling module 202, a control module 203, and an artificial intelligence recognition module 204. The main circuit 201 includes a DC / DC circuit 301, a DC bus 302, a DC / AC circuit 303, a grid connection circuit 304, a leakage current detection circuit 305, and a PID circuit 306.

[0070] The photovoltaic array 101 is connected to the input terminal of the DC / DC circuit 301; the output terminal of the DC / DC circuit 301 is connected to the input terminal of the DC / AC circuit 303 through the DC bus 302; the output terminal of the DC / AC circuit 303 is connected to the power grid 103 through the grid-connected circuit 304. The leakage current detection circuit 305 is connected between the grid-connected circuit 304 and the power grid 103.

[0071] In a photovoltaic system, parasitic capacitance and equivalent impedance exist between the photovoltaic array, DC cables, filters, inverter chassis, and the earth (PE). During grid-connected operation of the inverter, common-mode voltage generates common-mode current across the parasitic capacitance. The leakage current detection circuit 305 detects the leakage current by measuring the difference between the currents of each phase. When the impedance to ground decreases to a low level or a large capacitance to ground is suddenly connected to the system, the leakage current will change significantly.

[0072] The PID circuit 306 includes a PID circuit transformer, an inductor L1 on the secondary side, a grounding resistor R1, a repair switch K1, and a diagnostic switch K2. During the daytime, when the photovoltaic inverter is normally connected to the grid or the PID repair function is not running, the repair switch K1 and the diagnostic switch K2 are disconnected, and the secondary side of the PID circuit can be equivalent to a ground branch composed of inductor L1, grounding resistor R1, and related parasitic parameters.

[0073] Optionally, the diagnostic switch K2 can be an independent switching device or can reuse the repair switch K1. The timing logic control of the diagnostic switch K2 to perform closing and opening can generate coded diagnostic pulses on the PID circuit.

[0074] The inverter fault detection system includes a sampling module 202, a control module 203, and an artificial intelligence recognition module 204.

[0075] Sampling module 202 is used to acquire the first sampled data of the inverter. For example, the sampled value u of the equivalent inductor voltage on the secondary side of the PID circuit. L Leakage current sampling value i CM DC bus voltage sampling value u BUS Positive busbar to ground voltage sampling value u P Sampled value of negative busbar to ground voltage u N .

[0076] The control module 203 is used to obtain the inverted first ground impedance based on the first sampled data of the inverter, and to determine the first fault detection result of the inverter based on the first ground impedance and the first fault threshold.

[0077] The artificial intelligence recognition module 204 is used to identify the inverter's operating condition based on the first sampled data when the first fault detection result indicates that no fault is detected in the inverter. The operating condition characterizes the inverter's electrical state and disturbance level. Based on the first sampled data, it predicts the inverter's ground insulation state reference value. The ground insulation state reference value includes: the secondary-side equivalent inductance voltage reference value, leakage current reference value, ground impedance reference value, and prediction uncertainty of the PID circuit. Among them, the secondary-side equivalent inductance voltage reference value is the predicted value of the secondary-side equivalent inductance voltage, and the leakage current reference value is the predicted value of the leakage current.

[0078] The control module 203 is used to update the second fault threshold based on the ground insulation status reference value and the operating conditions, and to determine the second fault detection result of the inverter based on the first sampling data, the ground insulation status reference value and the second fault threshold.

[0079] Optionally, the artificial intelligence recognition module 204 can be integrated into the control module 203, or it can be deployed on a separate processor, edge computing module, power plant controller, or cloud.

[0080] It should be understood that Figure 1 This is merely an example of a circuit structure that can implement the inverter fault detection method of the embodiments of this application. The inverter fault detection method of the embodiments of this application can also be implemented on other circuit structures. This application does not limit the circuit structure of the inverter.

[0081] Figure 2 A flowchart illustrating an inverter fault detection method provided in this application embodiment. Figure 1 .like Figure 2 As shown, an inverter fault detection method may include, for example, the following steps:

[0082] S201. Obtain the first ground impedance based on the first sampling data of the inverter, and determine the first fault detection result of the inverter based on the first ground impedance and the first fault threshold.

[0083] The first sampled data consists of multi-source electrical quantities and state quantities during inverter operation, including relevant sampled information characterizing the inverter's operating state and its insulation state to ground. For example, the first sampled data may include at least one of the following: DC bus voltage, inverter output power, leakage current, and the sampled value of the equivalent inductance voltage on the secondary side of the PID circuit.

[0084] The first ground impedance is the estimated result of the inverter's current ground insulation impedance obtained by inverting the first sampled data.

[0085] The first fault threshold is the fault judgment boundary of the fixed threshold set in the first stage. The first fault threshold is used to determine whether the inverter has a complete grounding fault. A complete grounding fault refers to a fault in which the positive or negative busbar of the DC side of the photovoltaic inverter is directly connected to the protective ground at a near-low resistance, and the impedance to ground drops sharply to close to 0Ω, forming a large leakage current to ground.

[0086] The first fault detection result is used to output whether a fault is detected in the inverter based on a first fault threshold. The first fault detection result indicates whether a complete ground fault is detected in the inverter, or whether a complete ground fault is not detected in the inverter.

[0087] For example, the control module receives the first sampled data, preprocesses it, and then sends it to the impedance inversion unit. The impedance inversion unit calculates the current first impedance to ground based on the equivalent loop relationship to ground. In one possible embodiment, after acquiring the first sampled data, a circuit solution model is established by combining the circuit topology parameters of the inverter's current PID circuit. The collected leakage current, the sampled value of the equivalent inductor voltage on the secondary side of the PID circuit, and the DC bus voltage are substituted into the circuit solution model to perform impedance inversion calculation and obtain the first impedance to ground.

[0088] After obtaining the first impedance to ground, the control module compares it with a first fault threshold. When the first impedance to ground is less than or equal to the first fault threshold, a first fault detection result is generated, indicating that a fault has been detected in the inverter, and a protection alarm signal or fault record information is directly output.

[0089] Optionally, if the first fault detection result indicates that a fault exists in the inverter, the control module can perform fault handling actions such as waveform blocking, disconnecting the grid-connected switch, and alarming for a first-level ground impedance fault in the main circuit.

[0090] When the first impedance to ground is greater than the first fault threshold, a first fault detection result is generated, representing the fault that was not detected, and the subsequent operating condition identification and reference value prediction process is triggered.

[0091] This stage produces initial screening results under the basic protection boundary. Its purpose is to quickly separate fault states with significant drops in insulation impedance from continuous operating data, allowing subsequent dynamic judgments to be based on operating samples that have not triggered hard threshold protection. By processing the first sampled data and performing impedance inversion, and comparing the inverted first-to-ground impedance with the first fault threshold, preliminary identification of significant grounding faults or sudden drops in insulation impedance can be completed under grid-connected operation.

[0092] S202. If no fault is detected in the inverter according to the first fault detection result, the operating condition of the inverter is identified based on the first sampled data.

[0093] Operating conditions characterize the electrical state and disturbance level of the inverter. For example, operating conditions include stable operating conditions and transient suppression conditions.

[0094] For example, if the first fault detection result indicates that no fault is detected in the inverter, the artificial intelligence recognition module extracts information that can characterize the inverter's operating status based on the first sampled data and inputs it into the operating condition recognition unit of the artificial intelligence recognition module to identify the current operating condition.

[0095] Operating conditions can be used to characterize the current electrical state and disturbance level of the inverter, and written into the state register area of ​​the current detection cycle as a reference value for predicting the insulation state to ground and as a condition parameter for updating the second fault threshold.

[0096] Operating conditions serve as context variables describing the current data reliability background. For the same leakage current offset or the same voltage-to-ground change, the corresponding normal fluctuation range differs under different operating conditions. Therefore, it is necessary to first complete the operating condition mapping and then perform fine-grained judgment on the insulation status.

[0097] For example, based on the first sampled data and a preset mapping relationship between various operating conditions and sampled data states, a sampled data state matching the first sampled data is determined, and the operating condition corresponding to the matched sampled data state is taken as the identified operating condition. For example, if the DC bus voltage and inverter output power meet the preset value range of a stable operating state, then the operating condition is a stable operating state.

[0098] For example, the operating condition of the inverter is identified based on the first sampled data and the state of the devices in the inverter's circuit. For instance, if the inverter is in grid-connected mode and all the first sampled data remain stable, the operating condition can be determined to be a stable operating state; if the inverter is in the process of start-up / shutdown, grid-connected / off-grid switching, rapid bus changes, or protection action, the operating condition can be determined to be a transient suppression state.

[0099] Optionally, the device status can be obtained directly based on hardware status bits, or indirectly based on the first sampled data. For example, the inverter's grid-connected status can be obtained based on the hardware status bits obtained by the control module, or it can be determined based on whether the inverter's output power is greater than the power threshold. If the inverter's output power is greater than the power threshold, the inverter's grid-connected status is grid-connected; if the inverter's output power is less than or equal to the power threshold, the inverter's grid-connected status is off-grid.

[0100] In this approach, if the first fault detection result indicates that no fault is detected in the inverter, the operating condition identification is introduced. This allows the electrical state and disturbance level of the inverter to be extracted from the original sampled signal and used as a premise for subsequent dynamic threshold adjustment. This enables the second-stage detection to no longer rely on a single fixed boundary, but to compare with the normal response range consistent with the current operating condition, thereby improving the consistency of judgment under continuous operating conditions.

[0101] S203. Based on the first sampled data, predict the reference value of the inverter's insulation status to ground.

[0102] The reference value for the insulation status to ground is a set of predictive benchmarks for the inverter's healthy insulation response under the current operating conditions, which is dynamically generated by combining the first sampled data with the operating condition information.

[0103] The reference values ​​for insulation status to ground include the secondary-side equivalent inductance voltage reference value, leakage current reference value, ground impedance reference value, and prediction uncertainty of the PID circuit. The secondary-side equivalent inductance voltage of the PID circuit characterizes its electrical response to ground. The secondary-side equivalent inductance voltage reference value characterizes the baseline voltage response that the circuit should exhibit under healthy conditions. The leakage current reference value characterizes the normal leakage level under current operating conditions. The ground impedance reference value characterizes the predicted impedance baseline under healthy insulation conditions. The prediction uncertainty quantifies the reliability of the predicted reference values ​​for the secondary-side equivalent inductance voltage, leakage current, and ground impedance of the PID circuit.

[0104] For example, the artificial intelligence recognition module calls the preprocessed first sampled data and operating conditions, inputs them into a preset prediction model, to obtain the reference values ​​of the secondary equivalent inductance voltage, leakage current, ground impedance, and prediction uncertainty. In one possible embodiment, the prediction uncertainty can be determined based on the dispersion of the prediction results; in another implementation, the prediction uncertainty can also be directly output by the prediction model.

[0105] For example, the artificial intelligence recognition module employs a time-series prediction model, inputting a sampled data sequence and operating condition vector of the first sampled data of a preset length prior to the current sampling time, and outputting the reference values ​​for the secondary equivalent inductance voltage, leakage current, ground impedance, and prediction uncertainty within the next prediction window. Optionally, the ground insulation state reference value may also include the fault category probability and the expected duration to reach a preset threshold.

[0106] It should be understood that the temporal prediction model can be constructed using at least one of the following: one-dimensional convolutional network, recurrent neural network, temporal convolutional network, Transformer, state-space model, and tree model. The specific network structure of the temporal prediction model can adopt the temporal prediction model constructed in related technologies, and this application does not impose any restrictions on it.

[0107] By inputting the first sampled data and the operating conditions into the prediction process, reference values ​​for the secondary equivalent inductance voltage, leakage current, ground impedance, and prediction uncertainty are generated. This allows for the establishment of an insulation state reference surface that matches the operating disturbance under the current conditions without triggering the first stage fault, providing a basis for subsequent updates to the second fault threshold and fault detection.

[0108] S204. Update the second fault threshold based on the ground insulation status reference value and operating conditions, and determine the second fault detection result of the inverter based on the first sampling data, the ground insulation status reference value and the second fault threshold.

[0109] The second fault threshold is the dynamic judgment boundary of the first stage. Unlike the fixed protection boundary of the first fault threshold, the second fault threshold can be dynamically updated according to the current operating conditions, the distribution of reference values, and the prediction uncertainty.

[0110] The second fault detection result is used to output the fault judgment conclusion of the first stage. The second fault detection result can indicate that an incomplete grounding fault is detected in the inverter, or that the inverter has an insulation degradation trend of continuously decreasing impedance to ground, or that there is no fault in the inverter. If there is no fault in the inverter, it means that the inverter insulation is normal.

[0111] For example, the control module calculates a new second fault threshold according to a preset threshold update rule based on the operating conditions and the ground insulation state reference value, and updates the second fault threshold based on the new second fault threshold to adapt the second fault threshold to the current operating state. For example, based on the secondary equivalent inductance voltage reference value, leakage current reference value, ground impedance reference value in the ground insulation state reference value, and the prediction uncertainty, the new second fault threshold is obtained according to a preset threshold formula, and the second fault threshold is updated based on the new second fault threshold.

[0112] After obtaining the second fault threshold, the control module compares the actual response represented by the first sampled data with a reference value for the insulation state to ground, and makes a judgment based on the second fault threshold. When the deviation of the actual response from the reference value meets the fault judgment condition, a second fault detection result indicating that the inverter has a fault is output. When the actual response falls within the normal range constrained by the second fault threshold, a second fault detection result indicating that the inverter insulation is normal is output.

[0113] By updating the second fault threshold based on the ground insulation status reference value and operating conditions, and jointly judging the first sampled data and the ground insulation status reference value under this dynamic threshold, it is possible to further distinguish between operating disturbances and actual insulation anomalies, and complete the detection of inverter faults.

[0114] This application provides an inverter fault detection method. It obtains a first ground impedance based on first sampled data of the inverter, and determines a first fault detection result based on the first ground impedance and a first fault threshold. If the first fault detection result indicates that no fault is detected in the inverter, the operating condition of the inverter is identified based on the first sampled data. Based on the first sampled data, a reference value for the inverter's ground insulation state is predicted. A second fault threshold is updated based on the ground insulation state reference value and the operating condition. Finally, a second fault detection result is determined based on the first sampled data, the ground insulation state reference value, and the second fault threshold. In this application, a first-stage threshold determination is performed by inverting the first ground impedance and using the first fault threshold. Then, under the condition that no obvious fault is detected, the second fault threshold is dynamically updated based on the ground insulation state reference value, and a second-stage determination is performed. This enables the fault detection process to have adaptive operating condition capabilities, thereby improving the reliability of inverter fault detection.

[0115] Optionally, before updating the second fault threshold based on the ground insulation status reference value and operating conditions, the control module can also determine whether to apply a coded diagnostic pulse to the PID circuit based on the prediction uncertainty; if it is determined that a coded diagnostic pulse should be applied to the PID circuit, the coded diagnostic pulse is applied to the PID circuit. The second sampling data of the inverter is collected, and the ground insulation status reference value is updated.

[0116] The coded diagnostic pulse satisfies at least one of the following constraints: amplitude constraint, duration constraint, total energy constraint, and positive and negative volt-second balance constraint.

[0117] For example, after receiving the first sampled data, the control module uses the artificial intelligence recognition module to output a reference value for the insulation state to ground and its prediction uncertainty, and compares the prediction uncertainty with a preset threshold. If the prediction uncertainty indicates that the reference value predicted by the artificial intelligence recognition module is unreliable, a coded diagnostic pulse is applied to the PID circuit.

[0118] Taking a prediction uncertainty in the range of 0 to 1 as an example, where higher prediction uncertainty reflects a higher degree of unreliability, when the prediction uncertainty is within a preset range, the control module generates an coded diagnostic pulse control quantity, which is then applied to the PID circuit via a pulse width modulation drive module, causing the PID circuit to apply the coded diagnostic pulse. For example, the preset range is 0.7.

[0119] Subsequently, the sampling module re-collects the second sampling data, which may include information such as the secondary equivalent inductance voltage, leakage current, voltage to ground, and bus voltage. The second sampling data is then input into the artificial intelligence recognition module, which predicts the reference value of the insulation state to ground based on the second sampling data and updates the reference value of the insulation state to ground.

[0120] It should be understood that the steps of the artificial intelligence recognition module in predicting the reference value of the insulation state to ground based on the second sampled data can refer to the steps of the artificial intelligence recognition module in predicting the reference value of the insulation state to ground based on the first sampled data.

[0121] For example, one or more coded diagnostic pulses that satisfy amplitude constraints, duration constraints, total energy constraints, and positive and negative volt-second balance constraints are applied by a PID circuit.

[0122] For example, a charge-balanced pulse sequence with alternating positive and negative polarities and a total volt-second value of approximately zero is used to reduce the impact of the diagnostic signal on the average DC bus-to-ground potential and the PID effect of the photovoltaic module. The encoding method can employ a pseudo-random binary sequence, a step sequence, a multi-frequency sequence, or a combination of different pulse widths.

[0123] For example, the control module generates coded diagnostic pulse control quantities based on amplitude constraints, duration constraints, total energy constraints, and positive and negative volt-second balance constraints. The amplitude constraint is represented as follows: , The total energy constraint is expressed as The positive and negative volt-second balance constraint is expressed as: .

[0124] in, The diagnostic voltage of the coded diagnostic pulse at sampling time t. U represents the diagnostic current of the encoded diagnostic pulse at sampling time t. max I max and E max For device safety limits, The allowable positive and negative volt-second difference. Diagnostic integration interval [0, ... ], The duration of a single coded diagnostic pulse, i.e., the duration constraint, This represents the total energy of a single coded diagnostic pulse over its entire duration.

[0125] For example, the sampling module continuously acquires second sampling data, which includes the sampled values ​​u of the secondary-side equivalent inductor voltage of the PID circuit before, during, and after the application of the coded diagnostic pulse. L The sampling values ​​of the current and positive bus voltage to ground in the PID circuit are u. P Sampled value of negative busbar to ground voltage u N and leakage current sampling value i CM .

[0126] Optionally, the artificial intelligence recognition module can calculate the time constant, peak value, attenuation rate, correlation coefficient, and frequency domain impedance of the impulse response based on the second sampled data. Based on the time constant, peak value, attenuation rate, correlation coefficient, and frequency domain impedance of the impulse response, a reference value for the insulation state to ground can be predicted.

[0127] The measured impedance is obtained by dividing the diagnostic voltage and diagnostic current at that frequency by a complex number; the impedance characteristics of the ground fault are described by a resistor-capacitor model with a resistor and capacitor in parallel. For the k-th diagnostic frequency, the measured complex impedance of the loop and the complex impedance of the parallel resistor-capacitor grounding model can be expressed as shown in equation (1):

[0128] Equation (1)

[0129] in, The measured complex impedance of the circuit at the k-th diagnostic frequency is... Let be the frequency domain complex form of the diagnostic voltage at the k-th diagnostic frequency point. Let the frequency domain complex form of the diagnostic current at the k-th diagnostic frequency point be . The theoretical complex impedance of the parallel RC grounding model at the k-th diagnostic frequency. For grounding insulation resistance, For parasitic capacitance to ground, For the k-th diagnostic angular frequency, The imaginary unit in the complex frequency domain is the circuit's complex frequency domain operator.

[0130] The grounding resistance is calculated by fitting the real part, imaginary part, and phase of the impedance measured at multiple diagnostic frequency points. With ground capacitance This allows for the differentiation between two operating conditions: decreased resistive insulation and changes in parasitic capacitance. If the active response after applying the coded diagnostic pulse is significantly inconsistent with the model prediction, the prediction uncertainty is set to a preset uncertainty value. The preset uncertainty value indicates that the prediction is unreliable. Taking a prediction uncertainty in the range of 0 to 1 as an example, where a higher prediction uncertainty reflects a higher degree of unreliability, the preset uncertainty value could be set to 0.99.

[0131] For example, the frequency domain impedance can be calculated based on the second sampled data. The secondary side of the PID circuit during non-repair periods can be equivalent to an inductor L1, a grounding resistance R1, and related parasitic parameters. The impedance to ground Z... g The change will cause a change in the DC bus voltage to ground, and induce a voltage across inductor L1. The physical constraint feature generation unit obtains a rough impedance to ground based on this equivalent circuit. Expected inductor voltage and parameter sensitivity.

[0132] When the PID secondary diagnostic circuit is simplified to a first-order series RL circuit within a sampling window, the voltage across inductor L1 satisfies equation (2):

[0133] Equation (2)

[0134] in, Let L1 be the current flowing through the secondary side of the PID circuit at sampling time t. Let be the rate of change of the inductor current with respect to time at sampling time t.

[0135] Let Δu PE R represents the increment of the DC bus voltage to ground relative to the current steady-state value. Σ For the equivalent series resistance in the diagnostic circuit excluding inductor L1, the Laplace domain relation satisfies equation (3):

[0136] Equation (3)

[0137] in, Let Laplace be a complex variable. Let Δu be the Laplace domain transfer function of the secondary side of the PID circuit, representing the input Δu. PE arrive The mapping relationship. It is the voltage across inductor L1 The corresponding Laplace transform. For Δu PE The Laplace transform of .

[0138] When Δu PE When the voltage decay response is approximately a step and the parameters remain constant within the sampling window, the inductor voltage decays according to equation (4):

[0139] Equation (4)

[0140] in, This is the theoretical time constant of the secondary side of the PID circuit.

[0141] The time constant is obtained by fitting the impulse response. After that, you can follow The equivalent resistance of the loop is obtained. For implementations where the impedance to ground is a series quantity in this equivalent loop, the physical estimate can be obtained according to the following formula: Its general representation uses impedance inversion mapping: This mapping takes the loop equivalent resistance, loop inductance, loop intrinsic resistance, parasitic resistance, and capacitance to ground as inputs, adapting to complex real-world circuit topologies. Among them, R... par Given the known equivalent resistance of the cable, windings, and sampling branch, Z is the parasitic capacitance to ground. min For low impedance lower limit, This is an estimate of the equivalent series resistance (excluding L1) in the diagnostic circuit obtained by back-calculation from the fitted time constant. This is the inverted ground impedance.

[0142] This represents the impedance inversion relationship established based on the equivalent circuit of the secondary side of a PID circuit. This impedance inversion relationship can be achieved through analytical calculation, parameter lookup tables, numerical iteration, or a pre-calibrated mapping model, using the inverse mapping of a pre-calibrated topological forward mapping F. This allows us to solve for the impedance to ground. It should be understood that the above equivalent relationship is used to provide physical a priori constraints and does not require the actual circuit to be strictly equivalent to a single series loop.

[0143] Figure 3 A flowchart illustrating an inverter fault detection method provided in this application embodiment. Figure 2 .like Figure 3 As shown, it may include:

[0144] S301. Obtain the inverted first ground impedance based on the first sampled data of the inverter.

[0145] S302. Based on the first ground impedance and the first fault threshold, determine whether the first fault detection result indicates that a fault has been detected in the inverter.

[0146] If so, then execute S309;

[0147] If not, then execute S303.

[0148] S303. Identify the operating conditions of the inverter based on the first sampled data, and predict the reference value of the inverter's insulation status to ground based on the first sampled data.

[0149] S304. Based on the prediction uncertainty, determine whether to apply an coded diagnostic pulse to the PID circuit.

[0150] If so, then execute S305;

[0151] If not, then execute S307.

[0152] S305, control the PID circuit to apply coded diagnostic pulses.

[0153] S306. Collect the second sampling data of the inverter and update the reference value of the insulation status to ground.

[0154] S307. Update the second fault threshold based on the ground insulation status reference value and operating conditions.

[0155] S308. Based on the first sampling data, the reference value of the insulation status to ground, and the second fault threshold, determine whether the second fault detection result indicates that a fault has been detected in the inverter.

[0156] If so, then execute S309;

[0157] If not, then execute S301.

[0158] S309: Send control signals or protection signals to the main circuit of the inverter.

[0159] Optionally, the control module can output control signals or protection signals to the main circuit based on the first fault detection result or the second fault detection result. Protection signals include DC / DC or DC / AC blocking signals, grid-connected switch control signals, and fault alarm signals, etc.

[0160] The method in this application, by triggering active diagnostic excitation when uncertainty increases, makes the PID circuit response information more complete, and re-estimates the insulation state reference value in conjunction with the second sampling data, so that the subsequent second fault threshold can reflect the current operating disturbance and insulation evolution characteristics. Therefore, the threshold update process is synchronized with the field operating conditions, and the constraint design of the coded diagnostic pulse keeps its impact on the normal output of the inverter within a controlled range.

[0161] Optionally, the control module controls the PID circuit to apply coded diagnostic pulses. This can be done when the inverter meets safety constraints. Safety constraints include at least one of the following: inverter protection status flag constraints, DC bus voltage constraints, inverter output power constraints, leakage current constraints, and PID circuit repair function constraints.

[0162] For example, the inverter protection status flag constraint includes that the inverter is not in a protected state. The DC bus voltage constraint includes that the DC bus voltage is within a preset voltage allowable range. The inverter output power constraint includes that the inverter output power is within a preset power allowable range. The leakage current constraint includes that the leakage current is less than the diagnostic allowable upper limit. The PID circuit repair function constraint includes that the PID circuit repair function is not in operation.

[0163] For example, the control module continuously reads the first sampled data and compares the inverter protection status flag, DC bus voltage, inverter output power, and leakage current with preset thresholds. When the inverter is not in protection mode, the DC bus voltage is within the preset voltage allowable range, the inverter output power is within the preset power allowable range, the leakage current is less than the diagnostic allowable upper limit, and the PID circuit's repair function is not running, the control module generates a pulse enable signal to drive the PID circuit to output coded diagnostic pulses.

[0164] The coded diagnostic pulse can be formed by preset amplitude, duty cycle and duration, and superimposed on the PID control quantity, so as to complete the coded excitation of the ground insulation state without changing the normal closed-loop control architecture of the inverter.

[0165] For example, the control module first determines whether the current operating window is suitable for applying diagnostic pulses based on safety constraints. Only when the constraints are met will a pulse control command be issued, and then the coded diagnostic pulse will be output through the PID circuit. If any safety constraint is not met, no coded diagnostic pulse will be applied.

[0166] Figure 4 This is a schematic diagram illustrating the process of applying an coded diagnostic pulse, as provided in an embodiment of this application. Figure 4 As shown, it may include:

[0167] S401. Determine whether the inverter meets the safety constraints.

[0168] If so, then execute S402;

[0169] If not, then execute S403.

[0170] S402, control the PID circuit to apply coded diagnostic pulses.

[0171] S403. Determine the second fault detection result based on the first sampling data, the reference value of the insulation status to ground, and the second fault threshold.

[0172] For example, if the inverter does not meet the safety constraints, the second fault detection result can be determined based on the reference value of the ground insulation state predicted by the first sampling data and the historical second fault threshold or the preset second fault threshold.

[0173] The method in this application ensures that the coded diagnostic pulse enters the control loop only when the inverter's operating state is stable and the risk is controlled, thereby maintaining a consistent coupling relationship between the first sampled data and the diagnostic excitation. With this implementation, the DC bus voltage, inverter output power, leakage current, and protection flags together form the operating boundary for applying the diagnostic pulse. The PID circuit repair function and the diagnostic function are processed separately, enabling the coded diagnostic pulse to be executed under controllable operating conditions and providing a stable sampling basis for subsequent ground insulation state estimation.

[0174] Optionally, the second fault threshold includes a ground impedance threshold. Before updating the second fault threshold based on the ground insulation status reference value and operating conditions, the control module may also determine whether to update the second fault threshold based on the operating conditions; if it is determined that the second fault threshold should be updated, the second fault threshold is updated based on the ground insulation status reference value and operating conditions.

[0175] For example, after receiving the current operating condition, the control module first determines whether the condition has changed relative to the previous sampling time, whether it has entered a preset operating condition category, or whether the disturbance level corresponding to the condition meets preset change conditions. When any of these conditions are met, the control module determines that the second fault threshold needs to be updated. If the operating condition remains stable and the duration meets a preset duration condition, the control module maintains the current second fault threshold unchanged and continues to use the existing threshold to perform ground impedance discrimination. If an update is required, the control module uses the ground insulation state reference value and the current operating condition as the update basis to correct the ground impedance threshold so that it matches the normal insulation response under the current operating condition. This threshold correction can be completed by the control module according to a preset mapping relationship, compensation coefficient, or lookup table relationship, and the updated second fault threshold is output for subsequent fault detection.

[0176] For example, the operating conditions include a stable operating state and a transient suppression state. The artificial intelligence module can identify the inverter's operating condition based on the first sampled data and the state of the components in the inverter's circuitry. For instance, if the inverter is in a grid-connected state and all the first sampled data remain stable, it can be determined to be in a stable operating state; if the inverter is in a start-up / shutdown, grid-connected / off-grid switching, bus rapid change, or protection operation process, it can be determined to be in a transient suppression state. The artificial intelligence module can send the operating conditions to the control module, which determines whether to update the second fault threshold based on the operating conditions. If it is determined that the second fault threshold should be updated, the second fault threshold is updated based on the ground insulation state reference value and the operating conditions.

[0177] For example, when the inverter is connected to the grid and all operating parameters remain stable, it can be determined to be in a stable operating state; when the inverter is in the process of start-up / shutdown, grid-connection / off-grid switching, rapid bus changes, or protection operation, it can be determined to be in a transient suppression state. When the operating condition is in a stable operating state, the second fault threshold is updated. When the operating condition is in a transient suppression state, the second fault threshold is not updated.

[0178] It should be understood that, without updating the second fault threshold, in the process of determining the second fault detection result of the inverter based on the first sampling data, the reference value of the insulation state to ground and the second fault threshold, the second fault threshold used can be a historical second fault threshold or a preset second fault threshold.

[0179] The method in this application first determines whether the threshold needs adjustment based on the operating conditions, and then updates the threshold by combining the reference value of the insulation state to ground when adjustment is needed, thereby causing the second fault threshold to drift with changes in operating conditions. Since the ground impedance threshold can be synchronously corrected with changes in bus voltage, inverter output power and disturbance level, it can better fit the normal insulation range under the current operating conditions, thereby improving the adaptability of ground impedance discrimination and the detection stability under continuous operation.

[0180] Optionally, the second fault threshold is updated based on the ground insulation state reference value and operating conditions. The second ground impedance is obtained from the first sampled data. Anomaly scores are obtained based on at least one of the following: the difference between the second ground impedance and the ground impedance reference value; the difference between the secondary equivalent inductance voltage sample value and the secondary equivalent inductance voltage reference value; and the difference between the leakage current sample value and the leakage current reference value. These anomaly scores are then added to the anomaly score set. Based on the operating conditions, a conditional quantile representing the statistically permissible upper bound of the anomaly score when insulation is normal under operating conditions is obtained from the anomaly score set. A threshold correction amount is obtained based on the conditional quantile and the prediction uncertainty. The second fault threshold is then updated based on the threshold correction amount.

[0181] The second ground impedance is the ground impedance estimate obtained by inversion based on the current sampled values.

[0182] Anomaly scores are used to quantify the degree of deviation of the current state from the normal insulation state.

[0183] The abnormal score set is used to record historical abnormal scores under the same or similar operating conditions.

[0184] Conditional quantiles are used to characterize the statistical upper bound of the percentage of abnormalities when insulation is normal under a given operating condition.

[0185] The threshold correction amount is calculated based on the conditional quantile and the prediction uncertainty, and is output to dynamically correct the second fault threshold.

[0186] The trend of the reference value of ground impedance can be obtained from the continuous prediction results.

[0187] Optionally, the second fault detection result of the inverter can be determined based on the first sampling data, the reference value of the insulation state to ground, and the second fault threshold. The second fault detection result of the inverter can be determined based on at least one of the following: the difference between the second impedance to ground and the reference value of the impedance to ground, the difference between the sampled value of the equivalent inductance voltage of the secondary side and the reference value of the equivalent inductance voltage of the secondary side, the difference between the sampled value of the leakage current and the reference value of the leakage current, the trend of the reference value of the impedance to ground, and the second fault threshold.

[0188] For example, the loop impedance characteristics under operating conditions are extracted based on the first sampled data, and parameter inversion is performed using a parallel RC grounding model to obtain the second ground impedance. The inversion can employ circuit-based parameter inversion methods in related technologies, and this application does not impose any limitations on this.

[0189] For example, the control module can establish an equivalent circuit model to ground based on the sampled values ​​of the secondary side equivalent inductance voltage and leakage current, obtain the second impedance to ground by solving the model parameters, and compare it with the reference value of the impedance to ground. The anomaly score can be composed of the absolute value of the difference, the normalized deviation, or the weighted combination value. When multiple differences are involved at the same time, a single score can be synthesized according to the operating condition weight and written into the anomaly score set.

[0190] After identifying the operating conditions, the control module can extract conditional quantiles from the set of abnormal scores under the same operating conditions, and calculate the threshold correction amount in combination with the prediction uncertainty, and then perform recursive updates on the second fault threshold.

[0191] For example, the anomaly fraction at sampling time t is calculated for inductor voltage error, leakage current error, and impedance error. The expression is shown in equation (5):

[0192] Equation (5)

[0193] in, , and For non-negative weights, σ L σ I and σ Z This represents the scaling parameter for the corresponding prediction error. To find the minimum value, avoid having a denominator of 0. For inductor voltage error, This is due to leakage current error. This represents the impedance error in the logarithmic domain.

[0194] Optionally, the second fault threshold can be determined by historical statistics, Bayesian confidence intervals, or conformal prediction intervals. For example, the second fault threshold can be determined based on operating conditions and prediction uncertainty. For instance, it can be determined based on the conditional quantiles of the anomaly scores of normal samples under the same or similar operating conditions and the current uncertainty. As shown in equation (6):

[0195] Equation (6)

[0196] in, The normal and abnormal fractions under operating condition s(t) quantile, α is the allowable baseline false alarm probability, and k is the preset uncertainty coefficient. Let be the prediction uncertainty at sampling time t.

[0197] For example, the second fault threshold includes a ground impedance threshold, and / or an inductor voltage threshold, and / or a leakage current threshold. The control module can determine the impedance detection result based on the difference between the second ground impedance and a ground impedance reference value, and the ground impedance threshold. For example, if the difference is greater than or equal to the ground impedance threshold, the impedance detection result is a detection result indicating a fault in the inverter, and the second fault detection result is determined as a detection result indicating a fault in the inverter.

[0198] For example, the control module can also determine the inductor voltage sub-detection result based on the difference between the sampled value of the secondary-side equivalent inductor voltage and the reference value of the secondary-side equivalent inductor voltage, and the inductor voltage threshold. The leakage current sub-detection result is determined based on the difference between the sampled value of the leakage current and the reference value of the leakage current, and the leakage current threshold. A second fault detection result is obtained based on the impedance sub-detection result, the inductor voltage sub-detection result, and the leakage current sub-detection result. For example, if any one of the impedance sub-detection result, the inductor voltage sub-detection result, and the leakage current sub-detection result indicates a fault in the inverter, the second fault detection result is determined as the detection result indicating a fault in the inverter.

[0199] For example, if the trend of the change in the reference value of the impedance to ground is a continuous downward trend, then the second fault detection result is determined as the detection result characterizing the inverter having an insulation degradation trend.

[0200] For example, a fusion difference is obtained based on the difference between the second impedance to ground and the reference value of the impedance to ground, the difference between the sampled value of the secondary equivalent inductance voltage and the reference value of the secondary equivalent inductance voltage, and the difference between the sampled value of the leakage current and the reference value of the leakage current. The fusion difference is compared with the second fault threshold. If the fusion difference is greater than or equal to the second fault threshold, the second fault detection result is determined as the detection result characterizing the existence of a fault in the inverter.

[0201] The method in this application embodiment uses the second ground impedance obtained by inversion, the statistical upper bound of the abnormal fraction set, and the prediction uncertainty as the basis for threshold correction. The second fault threshold can be adjusted synchronously with the operating conditions, and the second fault detection result takes into account both the current disturbance and the insulation degradation trend, thereby improving the accuracy of fault detection.

[0202] Optionally, the second fault detection result of the inverter is determined based on the difference between the second ground impedance and the ground impedance reference value, and the changing trend of the ground impedance reference value. If the difference between the second ground impedance and the ground impedance reference value is greater than or equal to the ground impedance threshold, the second fault detection result is determined as a fault detection result characterizing the existence of a ground impedance fault in the inverter.

[0203] If the difference between the impedance to ground and the reference value of the impedance to ground is less than the impedance to ground threshold, and the trend of the reference value of the impedance to ground is a continuous downward trend, then the second fault detection result is determined to be a fault detection result that characterizes the inverter having an insulation degradation trend.

[0204] For example, the control module receives a reference value for ground impedance and a second ground impedance, calculates the difference between the two, and compares it with a ground impedance threshold. When the difference is not less than the threshold, the control module outputs the result as a ground impedance fault detection result and sends an alarm signal. When the difference is less than the threshold, the control module further performs trend determination on the historical ground impedance reference value sequence. If a continuous decrease in the sequence is detected, the current state is determined to be an insulation degradation trend, and the corresponding fault detection result is output. The ground impedance threshold can be stored in a threshold register and updated according to the insulation boundary corresponding to the operating condition. In practical applications, this threshold can also be a segmented threshold under different operating conditions, which is not limited in this application.

[0205] The method based on the embodiments of this application can definitively identify ground impedance anomalies that have reached the fault boundary, and can also determine insulation degradation states that have not yet exceeded the threshold but have shown a continuous decline, thereby achieving graded detection of inverter insulation status. This second fault detection result no longer relies solely on a single threshold comparison, but combines trend information to identify insulation degradation in advance, providing earlier criteria for subsequent alarms, power limiting operation, or maintenance, while reducing sensitivity to transient disturbances, making fault determination more closely reflect the actual insulation change process during photovoltaic grid-connected operation.

[0206] Optionally, the control module can also obtain the inverted third ground impedance based on the first sampled data; construct fast time-scale accumulation criteria, medium-speed time-scale accumulation criteria, and slow time-scale accumulation criteria based on the error signal between the first sampled data and the ground insulation state reference value, and the second fault threshold; for each time scale, perform a recursive accumulation operation with a forgetting coefficient on the positive deviation component of the error signal that exceeds the second fault threshold to obtain the recursive accumulation result for the corresponding time scale; determine the level of insulation degradation trend based on the decreasing slope and decreasing acceleration of the third ground impedance corresponding to the recursive accumulation results of the fast time scale, the medium-speed time scale, and the slow time scale.

[0207] Among them, the recursive cumulative results of the fast time scale are used to identify sudden grounding faults; the recursive cumulative results of the medium time scale are used to identify continuous anomalies; and the recursive cumulative results of the slow time scale are used for long-term insulation degradation analysis.

[0208] The third ground impedance is obtained by inverting the first sampled data. The third ground impedance can be obtained again by inverting the first sampled data, or the second ground impedance can be reused.

[0209] For example, the recursive accumulation operation with a forgetting coefficient can use an exponential decay method to accumulate the positive deviation component. That is, when the error signal is greater than the second fault threshold, the excess part is included in the accumulated value according to the weight of the current sampling time, while the historical accumulated amount is attenuated by the forgetting coefficient and then superimposed to form the recursive accumulation result for the corresponding time scale. The fast time scale corresponds to a smaller time window and weaker forgetting, which can respond to the instantaneous changes caused by ground faults. The medium-speed time scale uses a medium time window, which is suitable for characterizing abnormal deviations that last for several sampling periods. The slow time scale uses a longer time window to track the slow drift and aging process of insulation status. By setting thresholds and accumulation upper limits for different time scales, the recursive results can output sudden ground fault indications, continuous abnormality indications, and long-term insulation degradation indications, respectively.

[0210] For example, the falling slope and falling acceleration of the third pair of ground impedance can be calculated from the difference of the third pair of ground impedance at adjacent sampling times. The falling slope reflects the rate of decrease of impedance over time, and the falling acceleration reflects the trend of the rate of decrease.

[0211] After obtaining the recursive cumulative results for fast, medium, and slow time scales, the control module can combine the slope and acceleration of the third-to-ground impedance within the corresponding time period for joint judgment. When the cumulative result on the fast time scale continues to increase and the slope suddenly drops, a higher level of sudden degradation is output. When the cumulative result on the medium time scale steadily increases and the slope is continuously negative, a continuous anomaly level is output. When the cumulative result on the slow time scale rises slowly and the acceleration shows a continuous negative change, a long-term degradation level is output.

[0212] Optionally, when the insulation degradation trend reaches a preset level, an insulation degradation alarm message can be output.

[0213] For example, a cumulative criterion using multiple time scales (fast, medium, and slow) is employed. The window corresponding to the fast time scale is used to identify sudden grounding, the window corresponding to the medium time scale is used to identify continuous anomalies, and the window corresponding to the slow time scale is used to estimate the slope and acceleration of the impedance drop to ground. The cumulative amount at the current sampling time t for the q-th time scale is... It can be recursively derived according to equation (7):

[0214] Equation (7)

[0215] in, ρ is the difference between the anomaly score and the second fault threshold. q This represents the memory coefficient for the corresponding time scale. The window corresponding to a fast time scale uses a smaller ρ. qWindowing for medium-speed time scales identifies continuous anomalies, while windows for slow-speed time scales use a larger ρ value. q Estimate the decreasing trend of impedance to ground. This is the sampling time preceding sampling time t.

[0216] For example, when the number of times the error exceeds the dynamic threshold, the cumulative amount, or the trend of change meets the preset conditions, the insulation degradation trend level is increased, thereby avoiding false alarms of single-point noise.

[0217] For example, the control module can invoke the artificial intelligence recognition module to perform logarithmic domain filtering or state estimation on the ground impedance reference value to obtain a ground impedance trend sequence. Logarithmic domain processing can improve the resolution of impedance changes in the range of tens of kΩ to hundreds of kΩ.

[0218] Optionally, when the third pair of ground impedance Z is obtained simultaneously phy and the reference value of impedance to ground Z AI In this case, the uncertainties of the two estimates can be fused in the logarithmic domain to obtain the fused ground impedance. As shown in equation (8):

[0219] Equation (8)

[0220] Where, σ phy and σ AI The third pair of ground impedances Z phy and the reference value of impedance to ground Z AI Prediction uncertainty in the logarithmic field Given the initial impedance value after logarithmic domain fusion, its exponential restoration yields the fused ground impedance. When the uncertainty of an estimate for a particular path increases, it indicates a decrease in the reliability of that path's result, and the weight of that path in the fusion calculation will automatically decrease.

[0221] Optionally, the artificial intelligence module can predict the distribution of ground impedance at multiple future times based on the changing trend, rate of change, acceleration of change, environmental factors, and historical similar operating conditions. When the predicted distribution of ground impedance at multiple future times shows a continuous downward trend, or when the ground impedance at multiple future times reaches the first fault threshold and the second fault threshold, the estimated duration and confidence interval for reaching the preset fault threshold are calculated respectively.

[0222] When the expected time to reach the preset fault threshold is less than the maintenance lead time and the prediction confidence meets the conditions, predictive maintenance information is generated. The information may include at least one of the following: equipment identification, fault level, estimated location, third impedance to ground, reference value of impedance to ground, rate of descent, expected time to reach the preset fault threshold, and main evidence channels.

[0223] The method in this application, by distributing the positive deviation component of the error signal to recursive accumulation channels at different time scales and combining it with the rate of change and acceleration of change of the third ground impedance for graded determination, enables the ground insulation state to obtain corresponding trend identification results under three scenarios: sudden fault, continuous abnormality, and slow degradation, thereby improving the distinguishability and stability of insulation degradation trend grading.

[0224] Optionally, the first sampling data includes high-frequency local feature data and low-frequency trend feature data. The high-frequency local feature data includes the sampled values ​​of the equivalent inductance voltage on the secondary side of the PID circuit and the leakage current. The low-frequency trend feature data includes the sampled values ​​of the DC bus voltage, the positive bus-to-ground voltage, the negative bus-to-ground voltage, the inverter output power, the pulse width modulation switching status, and environmental parameter samples.

[0225] Optionally, the artificial intelligence recognition module predicts the reference value of the inverter's insulation state to ground based on the first sampled data. It can extract high-frequency local features from the high-frequency local feature data through the first branch of the dual-branch time-series prediction model; extract low-frequency trend features from the low-frequency trend feature data through the second branch of the dual-branch time-series prediction model; merge the high-frequency local features and low-frequency trend features in the feature fusion layer of the dual-branch time-series prediction model to obtain fused features; based on the fused features, the multiple output heads of the dual-branch time-series prediction model output: the reference value of the secondary equivalent inductance voltage, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty.

[0226] High-frequency local feature data correspond to transient fluctuation components.

[0227] Low-frequency trend characteristic data corresponds to slowly changing components.

[0228] The DC bus voltage sampling value, positive bus voltage to ground sampling value, negative bus voltage to ground sampling value, inverter output power, pulse width modulation switch status, and environmental parameter sampling value are used together to characterize operating condition disturbances, load changes, and environmental effects. The environmental parameter sampling value may include temperature, humidity, or other external quantities related to insulation status.

[0229] For example, the dual-branch timing prediction model can be deployed in the artificial intelligence recognition module, with its input terminals electrically connected to the voltage sampling unit, leakage current sampling unit, power sampling unit, switch state acquisition unit, and environmental sampling unit. Each sampling channel can be aligned with a synchronous clock before entering the preprocessing module. The preprocessing module performs normalization, missing value imputation, and sliding window slicing on data of different dimensions, and then sends them to the corresponding branches for timing feature extraction.

[0230] During training, the dual-branch time-series prediction model can establish supervisory labels for the secondary side equivalent inductance voltage, leakage current, and impedance to ground based on historical normal insulation samples. The model parameters are optimized together with prediction error and uncertainty constraints, so that the output reference value can be consistent with the actual electrical response under the current operating conditions.

[0231] The network structure of the dual-branch temporal prediction model can be implemented using one-dimensional convolutional networks, recurrent neural networks, temporal convolutional networks, Transformers, state-space models, tree models, etc., constructed in related technologies. The network layer type, number of layers, hidden unit size, and window length in the model can be arbitrarily set, and this application does not impose any restrictions on them.

[0232] For example, the input consists of a first sampled data sequence of length W and a condition vector prior to the current sampling time. The output includes the reference values ​​for the secondary equivalent inductor voltage, leakage current, ground impedance, fault category probability, prediction uncertainty, and the estimated time to reach the preset fault threshold within the next prediction window. The input-output relationship is shown in equation (9):

[0233] Equation (9)

[0234] in, Let s(t) be the first sampled data sequence of length W, s(t) be the operating condition vector consisting of grid connection status, power, PWM mode, and environmental quantities, H be the prediction step number, and f be the first sampled data sequence of length W. θ This is a two-branch time series prediction model with parameter θ. The reference value for the output secondary-side equivalent inductance voltage. The reference value for the output leakage current. The output impedance to ground is the reference value. For the output fault category probability, The output is the prediction uncertainty. This is the estimated time it will take to reach the preset fault threshold.

[0235] Figure 5 This is a schematic flowchart illustrating a method for predicting a reference value for insulation state to ground, as provided in an embodiment of this application. Figure 5 As shown, the first branch of the two-branch timing prediction model processes the sampled value u of the equivalent inductor voltage on the secondary side. L and leakage current sampling value i CM High-frequency local features of high-frequency local feature data. The second branch processes the DC bus voltage sampling value u. BUS Positive busbar to ground voltage sampling value u P Sampled value of negative busbar to ground voltage u NLow-frequency trend characteristics are generated from low-frequency trend data such as inverter output power, pulse width modulation switching status, and environmental parameter sampling values. The feature fusion layer can splice, map, or weightedly fuse the two outputs to form a fused feature that simultaneously includes local disturbances and trend background. The two branches are merged at the feature fusion layer, and multiple output heads output the secondary equivalent inductance voltage reference value, leakage current reference value, ground impedance reference value, fault category probability, prediction uncertainty, and the estimated time to reach the preset fault threshold.

[0236] Optionally, high-frequency local features and low-frequency trend features can be merged in the feature fusion layer of the dual-branch time-series prediction model to obtain fused features. High-frequency local features, low-frequency trend features, and equivalent circuit priors can be merged in the feature fusion layer of the dual-branch time-series prediction model to obtain fused features. The equivalent circuit prior includes physical constraint features such as the third-to-ground impedance, expected inductor voltage, parameter change direction, and parameter sensitivity. These physical constraint features are generated by a physical constraint feature generation unit. This unit receives the operating condition identification results and, combined with the circuit relationships between the secondary-side equivalent inductance, resistance, DC bus-to-ground voltage, and grounding impedance of the PID circuit, as well as sampled data, outputs estimated physical constraint features such as the third-to-ground impedance, expected inductor voltage, parameter change direction, and parameter sensitivity.

[0237] The model training loss can include the secondary equivalent inductance voltage prediction loss, leakage current prediction loss, ground impedance estimation loss, fault classification loss, physical consistency loss, and uncertainty loss, and their weighted relationship is shown in Equation (10):

[0238] Equation (10)

[0239] in, λ is the total loss during model training. u , λ i , λ Z , λ c , λ p and λ σ The weights are non-negative, and ε is a positive number to prevent division by zero and logarithmic overflow. Predict the loss for the equivalent inductance voltage on the secondary side; Predict losses based on leakage current; To estimate the loss for ground impedance; Classify losses based on faults; This represents a loss of physical consistency. To predict uncertainty loss, physical consistency loss can include an output change direction constraint, which applies to all training samples, ensuring that the direction of change in the model's output impedance to ground aligns with the direction of change after load compensation. LThe response changes in the same direction. Under samples with applied coded diagnostic pulses, the physical consistency loss can also include the circuit equation error constraints corresponding to equations (2) to (4).

[0240] For example, the prediction uncertainty loss can be represented as shown in equation (11):

[0241] Equation (11)

[0242] Where y is the label value. This represents the corresponding predicted value output by the model. ε is the predicted standard deviation of the model output, and ε is a positive number to prevent logarithmic overflow.

[0243] By using equivalent circuit relationships as model inputs, loss constraints, or output verification conditions, the probability of the model outputting results that violate circuit laws under unseen operating conditions can be reduced.

[0244] The artificial intelligence module learns the nonlinear relationship between the sampled data and the reference value of the insulation state to ground under different operating conditions of the inverter. This allows it to compensate for parasitic parameter changes, device drift and multivariate coupling errors that are difficult to describe accurately by conventional physical models, and outputs the reference value of the insulation state to ground.

[0245] The method in this application modeles high-frequency local fluctuations and low-frequency trend changes separately, and then simultaneously outputs multiple reference quantities and uncertainties after fusion, thereby jointly characterizing the correlation between the secondary-side equivalent inductance voltage, leakage current, and ground impedance. Because information from different time scales is separated, extracted, and then fused, the output result can more stably reflect the insulation state reference baseline. When subsequently updating the fault threshold based on this reference baseline, the disturbances caused by bus fluctuations, pulse width modulation switch switching, and environmental changes can be reduced, thereby improving the reliability of fault detection.

[0246] Optionally, the artificial intelligence module can also obtain multi-feature diagnostic results based on the first sampled data, the ground insulation state reference value, and the second sampled data. Based on the multi-feature diagnostic results, the probability of fault occurrence for each fault category is adjusted.

[0247] Multi-feature diagnostic results are used to map instantaneous sampling deviation, ground insulation change trend, and active diagnostic response into fault discrimination features.

[0248] The multi-feature diagnostic results include at least one of the following: secondary-side equivalent inductance voltage difference, degree of voltage imbalance between positive and negative buses and ground, leakage current difference, trend of change of reference value of ground impedance, and pulse response of coded diagnostic pulse. The secondary-side equivalent inductance voltage difference reflects the deviation between the sampled voltage and the reference voltage. The degree of voltage imbalance between positive and negative buses and ground characterizes the asymmetry of the DC-side bus-to-ground potential. The leakage current difference describes the deviation between the measured leakage current and the predicted or reference leakage current. The trend of change of reference value of ground impedance characterizes the temporal evolution of the insulation state. The pulse response of the coded diagnostic pulse characterizes the electrical response characteristics after applying active excitation.

[0249] The fault categories include at least one of the following: internal grounding fault, external DC circuit grounding fault, AC side and filter branch abnormality, parasitic capacitance change, bus transient disturbance, and sensor abnormality.

[0250] For example, after receiving the first sampled data and the reference value for insulation status to ground, the control module can perform time alignment and amplitude normalization on the two, and calculate various diagnostic quantities according to a preset feature mapping relationship. When the difference between the sampled value of the secondary equivalent inductance voltage and its reference value exceeds a preset tolerance, a secondary equivalent inductance voltage difference is formed. When the offset between the positive bus voltage to ground and the negative bus voltage to ground exceeds a balance threshold, the degree of imbalance of the positive and negative bus voltages to ground is formed. When there is a deviation between the sampled value of leakage current and the reference value of leakage current, a leakage current difference is formed. When the reference value of impedance to ground continuously decreases or changes abruptly, its changing trend is extracted.

[0251] The artificial intelligence module can also obtain multi-feature diagnostic results based on the first sampled data and the ground insulation state reference value. Based on these multi-feature diagnostic results, the probability of fault occurrence for each fault category is adjusted. When an coded diagnostic pulse is applied to the PID circuit, the response curves during and after the pulse are collected, and the pulse response is used as the diagnostic feature input. These features can be used individually or combined according to preset weights to form multi-feature diagnostic results.

[0252] For example, after obtaining the multi-feature diagnostic results, the probability of each fault category is redistributed based on the indication strength of each feature for different fault categories. If the imbalance between the positive and negative bus voltages to ground and the leakage current difference increase simultaneously, the probability of an internal grounding fault or an external DC circuit grounding fault is increased. If the trend of the ground impedance reference value shows a continuous decrease and the pulse response shows a significant capacitive hysteresis, the probability of a sudden change in parasitic capacitance is increased. If the multi-channel characteristics fluctuate synchronously in a short period of time, the probability of a bus transient disturbance is increased. If there is a physical contradiction between the voltage, current, and impedance characteristics, the probability of a sensor malfunction is increased. Based on this probability adjustment result, a fault category judgment that better matches the current operating conditions is output.

[0253] For example, the control module can input the first sampled data and the reference value of the insulation state to ground into the diagnostic model to generate a confidence vector corresponding to each fault category, and dynamically adjust the probability of internal grounding faults, external DC circuit grounding faults, parasitic capacitance mutations, sensor anomalies and bus transient disturbances according to the feature matching relationship.

[0254] When the three differences simultaneously reach the threshold and the pulse response meets the first preset judgment condition, it indicates that the electrical response is more consistent with the characteristic combination of internal grounding, thus increasing the probability of internal grounding faults. When the three differences increase synchronously and meet the loop constraints corresponding to external grounding faults, it indicates that the fault is more likely to be located in the external branch of the DC loop, thus increasing the probability of external DC loop grounding faults. When the high-frequency component of the leakage current is abnormal while the impedance trend remains steady and the pulse response shows a time constant or frequency characteristic shift, it indicates that the ground capacitance parameter has abruptly changed, thus increasing the probability of parasitic capacitance abrupt changes. When a single-channel sampling exceeds the normal range and other related characteristics contradict it, or the pulse response does not meet the equivalent circuit constraints, it indicates that the sampling link is inconsistent with the physical relationship, thus increasing the probability of sensor abnormality. When the fluctuation of the secondary equivalent inductance voltage difference can be explained by changes in bus voltage, inverter output power, or pulse width modulation switch state, and the ground impedance reference value does not show a continuous decrease, it indicates that it is closer to operational disturbances than insulation degradation, thus increasing the probability of bus transient disturbances.

[0255] Figure 6 This is a schematic diagram illustrating a fault type prediction process provided in an embodiment of this application. Figure 6 As shown, the multi-feature diagnostic results include the secondary side equivalent inductance voltage difference, the degree of voltage imbalance between the positive and negative buses and ground, the leakage current difference, the trend of the reference value of the ground impedance, and the pulse response of the coded diagnostic pulse. Fault categories include internal grounding faults, external DC circuit grounding faults, AC side and filter branch anomalies, parasitic capacitance mutations, bus transient disturbances, and sensor anomalies.

[0256] For example, when the difference in equivalent inductance voltage on the secondary side, the degree of voltage imbalance between the positive and negative busbars and ground, and the difference in leakage current meet the preset threshold judgment conditions, and the pulse response meets the first preset judgment condition, the probability of the occurrence of internal grounding fault in the equipment is increased.

[0257] For example, when the difference in equivalent inductance voltage on the secondary side, the degree of imbalance between positive and negative bus voltages and ground, and the difference in leakage current all increase, and the trend of change matches the physical constraint relationship of the ground loop corresponding to the external ground fault, the probability of the occurrence of the external DC loop ground fault is increased.

[0258] For example, when the high-frequency component of the leakage current difference meets the high-frequency change judgment condition, the change trend of the ground impedance reference value meets the steady-state judgment condition, and the pulse response meets the preset offset judgment condition of time constant offset or frequency characteristic offset, the probability of parasitic capacitance sudden change is increased.

[0259] For example, when any data in the first sampled data exceeds the preset value range, the deviation of the other associated sampled features from the sampled feature contradicts each other, or the impulse response does not meet the equivalent circuit constraints, the probability of sensor abnormality is increased.

[0260] For example, when the output of a single sensor deviates from the normal range, and the other associated physical characteristics contradict the deviation of the sensor output, or when the impulse response does not meet the equivalent circuit constraints, the probability of sensor abnormality is increased.

[0261] For example, when the fluctuation of the secondary equivalent inductance voltage difference can be corresponded to changes in the bus voltage, inverter output power, or pulse width modulation switch state, and the trend of the change in the ground impedance reference value does not meet the continuous decline judgment condition, the probability of bus transient disturbance occurrence is increased.

[0262] This approach establishes fault probability adjustment based on the joint constraints of three types of information: sampling characteristics, prediction characteristics, and impulse response. This results in more discriminative diagnostic results for internal grounding, external grounding, capacitance mutations, sensor anomalies, and bus transient disturbances. It also distinguishes operational disturbances from actual insulation anomalies, thereby improving the accuracy of fault identification and stability in continuous operation scenarios.

[0263] Optionally, the control module can select the fault category with the highest probability of occurrence based on the fault occurrence probability of each fault category, and output it as the predicted fault category of the inverter failure.

[0264] The method in this application combines first sampled data with a ground insulation state reference value for multi-dimensional feature extraction and maps the features to a probability space of multiple fault categories. This enables the diagnostic results to simultaneously cover information such as voltage imbalance, leakage current changes, insulation impedance evolution, and active excitation response. This makes the fault category determination more consistent with the current insulation state change and improves the ability to distinguish between ground faults, transient disturbances, and sensor anomalies.

[0265] Figure 7 This is a schematic diagram of the structure of an inverter fault detection system provided in an embodiment of this application. Figure 7 As shown, the inverter fault detection system includes a sampling module, an artificial intelligence recognition module, and a control module.

[0266] The sampling module is used to collect the first sampled data from the inverter.

[0267] The control module is used to obtain the inverted first ground impedance based on the first sampled data of the inverter, and to determine the first fault detection result of the inverter based on the first ground impedance and the first fault threshold.

[0268] The artificial intelligence recognition module is used to identify the inverter's operating condition based on the first sampled data when the first fault detection result indicates that no fault is detected in the inverter; the operating condition indicates the inverter's electrical state and disturbance level; based on the first sampled data, it predicts the inverter's ground insulation state reference value; the ground insulation state reference value includes: the secondary equivalent inductance voltage reference value of the potential-induced decay PID circuit, the leakage current reference value, the ground impedance reference value, and the prediction uncertainty.

[0269] The control module is used to update the second fault threshold based on the ground insulation status reference value and operating conditions, and to determine the second fault detection result of the inverter based on the first sampling data, the ground insulation status reference value and the second fault threshold.

[0270] Optionally, the inverter fault detection system also includes a data storage unit, which stores the data collected by the sampling module, the intermediate and output data processed by the artificial intelligence recognition module, and the intermediate and output data processed by the control module.

[0271] Optionally, the control module may output control signals or protection signals to the main circuit based on the first fault detection result or the second fault detection result.

[0272] It should be understood that the inverter fault detection system can implement any of the methods in the above embodiments. Below, specific embodiments of multiple inverter fault detection systems for fault detection are illustrated.

[0273] For example, taking a high-resistance grounding online early warning system under fixed operating conditions as an example, the inverter is operating stably in grid-connected mode, with an initial equivalent impedance to ground of approximately 500kΩ. As the insulation of components or cables becomes damp, the estimated impedance to ground continues to decrease over several hours. Although u L Peak and i CM None of them reached the fixed hard protection threshold, but the window corresponding to the slow time scale confirmed the impedance decreasing trend. The system outputs a predictive warning when the ground impedance enters the warning range or the expected time is less than the maintenance lead time.

[0274] For example, taking false alarm suppression under the condition of power surge as an example, shadowing causes short-term changes in PV power and DC bus voltage, u L A sudden change occurs. The operating condition identification unit classifies this period as a bus transient. The timing prediction model obtains the corresponding normal prediction value based on the bus voltage change rate, PWM modulation ratio, and output power changes. Therefore, uL If the error does not exceed the second fault threshold and the estimated ground impedance does not decrease continuously, the system will not output a ground fault.

[0275] For example, taking the operating condition as a distinction between resistive insulation degradation and parasitic capacitance change, passive monitoring detects an increase in leakage current, but the model gives similar probabilities for external grounding and parasitic capacitance increase. When the busbar and power are stable and the leakage current is below the preset diagnostic allowable value, the system applies alternating positive and negative coded pulses through a PID circuit. If the pulse response mainly shows a change in time constant while the dissipation characteristics remain basically unchanged, the probability of abrupt parasitic capacitance change increases; if the steady-state amplitude, dissipation, and ground impedance estimate all indicate a decrease in impedance, the probability of external grounding fault increases.

[0276] For example, taking the operating conditions of internal grounding and external grounding as an example, when the fault is located outside the detection range of the leakage current sensor, u L The error and the imbalance of the positive and negative bus voltages to ground are significant, while i CM The increment is small, but the system increases the probability of internal grounding faults. When the insulation of the external DC cable breaks, creating a path to ground, u L u P u N and i CM The occurrence of time-dependent synchronization anomalies increases the probability of external grounding faults in the system. In both cases, if the first fault threshold is reached, the protection action is unaffected by the fault identification process and results, and the control module directly executes ground impedance fault protection.

[0277] For example, taking the operating conditions of sensor malfunction and degraded operation as an example, when the leakage current sensor experiences zero-point drift, i CM Single channel slowly shifts, but u L u P u N If the pulse response to the applied coded diagnostic pulse does not indicate a decrease in impedance to ground, the system increases the probability of sensor anomaly and prompts for calibration. If the model input channel is missing or inference times out, the system exits the artificial intelligence recognition process and continues operation using a preset fixed threshold detection method.

[0278] The above describes the methods and systems of this application. The apparatus provided in the embodiments of this application will be described below.

[0279] Figure 8 This is a schematic diagram of the structure of an inverter fault detection device provided in an embodiment of this application. Figure 8 As shown, the inverter fault detection device 800 may include, for example, a first detection module 801, an operating condition identification module 802, a prediction module 803, and a second detection module 804. Optionally, it may also include an active identification module.

[0280] The first detection module 801 is used to obtain the inverted first ground impedance based on the first sampling data of the inverter, and to determine the first fault detection result of the inverter based on the first ground impedance and the first fault threshold.

[0281] Operating condition identification module 802 is used to identify the operating condition of the inverter based on the first sampled data when the first fault detection result indicates that no fault is detected in the inverter; the operating condition indicates the electrical state and disturbance level of the inverter.

[0282] Prediction module 803 is used to predict the reference value of the inverter's insulation state to ground based on the first sampled data; the reference value of the insulation state to ground includes: the reference value of the equivalent inductance voltage on the secondary side of the potential-induced decay PID circuit, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty;

[0283] The second detection module 804 is used to update the second fault threshold based on the ground insulation status reference value and operating conditions, and to determine the second fault detection result of the inverter based on the first sampling data, the ground insulation status reference value and the second fault threshold.

[0284] One possible implementation involves the active identification module, before updating the second fault threshold based on the ground insulation state reference value and operating conditions, specifically performing the following: determining whether to apply a coded diagnostic pulse to the PID circuit based on the prediction uncertainty; if it is determined that a coded diagnostic pulse should be applied to the PID circuit, controlling the PID circuit to apply the coded diagnostic pulse; wherein the coded diagnostic pulse satisfies at least one of amplitude constraints, duration constraints, total energy constraints, and positive and negative volt-second balance constraints; and acquiring the second sampling data of the inverter to update the ground insulation state reference value.

[0285] One possible implementation is that the first sampled data includes DC bus voltage, inverter output power, and leakage current; the active identification module is specifically used to: control the PID circuit to apply coded diagnostic pulses when the inverter meets the safety constraints; the safety constraints include at least one of the following: inverter protection status flag constraints, DC bus voltage constraints, inverter output power constraints, leakage current constraints, and PID circuit repair function constraints.

[0286] In one possible implementation, the second fault threshold includes a ground impedance threshold; before updating the second fault threshold based on the ground insulation status reference value and operating conditions, the second detection module 804 is specifically used to: determine whether to update the second fault threshold based on the operating conditions; if it is determined that the second fault threshold should be updated, update the second fault threshold based on the ground insulation status reference value and operating conditions.

[0287] One possible implementation involves the following: the first sampling data includes the sampled values ​​of the secondary-side equivalent inductance voltage and leakage current of the PID circuit; the second detection module 804 is specifically used to: obtain the inverted second ground impedance based on the first sampling data; obtain an anomaly score based on at least one of the differences between the second ground impedance and the ground impedance reference value, the differences between the secondary-side equivalent inductance voltage sampled value and the secondary-side equivalent inductance voltage reference value, and the differences between the leakage current sampled value and the leakage current reference value, and add the anomaly score to the anomaly score set; based on the operating conditions, obtain the conditional quantile representing the statistically permissible upper bound of the anomaly score when the insulation is normal under the operating conditions from the anomaly score set; obtain a threshold correction amount based on the conditional quantile and the prediction uncertainty; update the second fault threshold based on the threshold correction amount; and determine the second fault detection result of the inverter based on at least one of the differences between the second ground impedance and the ground impedance reference value, the differences between the secondary-side equivalent inductance voltage sampled value and the secondary-side equivalent inductance voltage reference value, the differences between the leakage current sampled value and the leakage current reference value, the changing trend of the ground impedance reference value, and the second fault threshold.

[0288] One possible implementation is that the second fault threshold includes a ground impedance threshold; the second detection module 804 is specifically used to: if the difference between the second ground impedance and the ground impedance reference value is greater than or equal to the ground impedance threshold, then determine the second fault detection result as a fault detection result indicating that the inverter has a ground impedance fault; if the difference between the ground impedance and the ground impedance reference value is less than the ground impedance threshold, and the changing trend of the ground impedance reference value is a continuous decreasing trend, then determine the second fault detection result as a fault detection result indicating that the inverter has an insulation degradation trend.

[0289] In one possible implementation, the second detection module 804 is specifically used to: obtain the inverted third ground impedance based on the first sampling data; construct a fast time-scale accumulation criterion, a medium-speed time-scale accumulation criterion, and a slow time-scale accumulation criterion based on the error signal between the first sampling data and the ground insulation state reference value, and a second fault threshold; for each time scale, perform a recursive accumulation operation with a forgetting coefficient on the positive deviation component of the error signal that exceeds the second fault threshold to obtain the recursive accumulation result for the corresponding time scale; wherein, the recursive accumulation result of the fast time scale is used to identify sudden grounding faults; the recursive accumulation result of the medium-speed time scale is used to identify continuous anomalies; the recursive accumulation result of the slow time scale is used for long-term insulation degradation analysis; and determine the level of insulation degradation trend based on the decreasing slope and decreasing acceleration of the third ground impedance corresponding to the recursive accumulation results of the fast time scale, the medium-speed time scale, and the slow time scale.

[0290] One possible implementation involves the following: the first sampling data includes high-frequency local feature data and low-frequency trend feature data; the high-frequency local feature data includes the sampled values ​​of the secondary-side equivalent inductor voltage and leakage current of the PID circuit; the low-frequency trend feature data includes the sampled values ​​of the DC bus voltage, positive bus-to-ground voltage, negative bus-to-ground voltage, inverter output power, pulse width modulation switching status, and environmental parameter sampling values; the prediction module 803 is specifically used to: extract high-frequency local features from the high-frequency local feature data through the first branch of the dual-branch timing prediction model; extract low-frequency trend features from the low-frequency trend feature data through the second branch of the dual-branch timing prediction model; merge the high-frequency local features and low-frequency trend features in the feature fusion layer of the dual-branch timing prediction model to obtain fused features; based on the fused features, the multiple output heads of the dual-branch timing prediction model output: the reference value of the secondary-side equivalent inductor voltage, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty, respectively.

[0291] One possible implementation is that the prediction module 803 is specifically used to: obtain multi-feature diagnostic results based on the first sampled data, the ground insulation state reference value, and the second sampled data; the multi-feature diagnostic results include at least one of the following: the difference in equivalent inductance voltage on the secondary side, the degree of voltage imbalance between the positive and negative busbars and ground, the difference in leakage current, the trend of the change in the ground impedance reference value, and the pulse response of the coded diagnostic pulse; and adjust the probability of occurrence of each fault category based on the multi-feature diagnostic results; the fault categories include at least one of the following: internal grounding fault of the equipment, external DC circuit grounding fault, AC side and filter branch abnormality, parasitic capacitance change, busbar transient disturbance, and sensor abnormality.

[0292] The apparatus provided in this application embodiment can execute the above method embodiment, and its implementation principle and technical effect are similar, so it will not be described again here.

[0293] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0294] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 9 As shown, the electronic device may include at least one processor 901 and a memory 902.

[0295] The memory 902 is used to store programs. Specifically, the program may include program code, which includes computer operation instructions.

[0296] Memory 902 may include high-speed RAM memory, and may also include non-volatile memory.

[0297] The processor 901 is used to execute computer execution instructions stored in the memory 902 to implement the method of the foregoing method embodiments. The processor 901 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0298] Optionally, the electronic device may also include a communication interface 903. In specific implementations, if the communication interface 903, the memory 902, and the processor 901 are implemented independently, the communication interface 903, the memory 902, and the processor 901 can be interconnected via a bus to complete communication between them.

[0299] Optionally, in a specific implementation, if the communication interface 903, memory 902, and processor 901 are integrated on a single chip, then the communication interface 903, memory 902, and processor 901 can communicate through an internal interface.

[0300] This application also provides a computer-readable storage medium, which may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), and a random access memory (RAM). Specifically, the computer-readable storage medium stores program instructions, which are used to implement the actions of the above-described method implementation.

[0301] This application also provides a program product including executable instructions stored in a readable storage medium. At least one processor of an electronic device can read the executable instructions from the readable storage medium, and the at least one processor executes the executable instructions to cause the electronic device to perform the actions described in the method embodiments.

[0302] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0303] When an integrated unit / module is implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0304] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), and portable hard drives.

[0305] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0306] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM and RAM.

Claims

1. A method for detecting inverter faults, characterized in that, include: The first ground impedance is obtained by inverting the first sampling data of the inverter, and the first fault detection result of the inverter is determined based on the first ground impedance and the first fault threshold. If the first fault detection result indicates that no fault is detected in the inverter, the operating condition of the inverter is identified based on the first sampling data. The operating conditions characterize the electrical state and disturbance level of the inverter; Based on the first sampled data, predict the reference value of the inverter's insulation status to ground; The reference values ​​for the insulation state to ground include: the reference value for the equivalent inductance voltage on the secondary side of the potential-induced decay PID circuit, the reference value for the leakage current, the reference value for the impedance to ground, and the prediction uncertainty. The second fault threshold is updated based on the ground insulation status reference value and the operating condition, and the second fault detection result of the inverter is determined based on the first sampling data, the ground insulation status reference value and the second fault threshold.

2. The method according to claim 1, characterized in that, Before updating the second fault threshold based on the ground insulation state reference value and the operating conditions, the method further includes: Based on the prediction uncertainty, determine whether to apply an coded diagnostic pulse to the PID circuit; When it is determined that the coded diagnostic pulse is applied to the PID circuit, the PID circuit is controlled to apply the coded diagnostic pulse; wherein the coded diagnostic pulse satisfies at least one of amplitude constraint, duration constraint, total energy constraint, and positive and negative volt-second balance constraint. The second sampling data of the inverter is collected to update the reference value of the insulation status to ground.

3. The method according to claim 2, characterized in that, The first sampled data includes DC bus voltage, inverter output power, and leakage current; the control of the PID circuit to apply the coded diagnostic pulse includes: When the inverter meets the safety constraints, the PID circuit is controlled to apply the coded diagnostic pulse. The safety constraints include at least one of the following: inverter protection status flag constraints, DC bus voltage constraints, inverter output power constraints, leakage current constraints, and PID circuit repair function constraints.

4. The method according to any one of claims 1-3, characterized in that, The second fault threshold includes a ground impedance threshold; before updating the second fault threshold based on the ground insulation state reference value and the operating conditions, the method further includes: Based on the operating conditions, determine whether to update the second fault threshold; The step of updating the second fault threshold based on the ground insulation state reference value and the operating conditions includes: If it is determined that the second fault threshold needs to be updated, the second fault threshold is updated based on the ground insulation status reference value and the operating conditions.

5. The method according to any one of claims 1-3, characterized in that, The first sampling data includes the sampled value of the equivalent inductance voltage on the secondary side of the PID circuit and the sampled value of the leakage current; updating the second fault threshold according to the ground insulation state reference value and the operating conditions includes: The inverted second ground impedance is obtained based on the first sampled data; An anomaly score is obtained based on at least one of the following: the difference between the second impedance to ground and the reference value of the impedance to ground, the difference between the sampled value of the secondary equivalent inductance voltage and the reference value of the secondary equivalent inductance voltage, and the difference between the sampled value of the leakage current and the reference value of the leakage current. The anomaly score is then added to the anomaly score set. Based on the operating conditions, obtain the conditional quantiles from the abnormal score set that characterize the upper limit of the statistical tolerance of abnormal scores when the insulation is normal under the operating conditions. The threshold correction amount is obtained based on the conditional quantiles and the prediction uncertainty; Update the second fault threshold according to the threshold correction amount; The step of determining the second fault detection result of the inverter based on the first sampled data, the ground insulation state reference value, and the second fault threshold includes: The second fault detection result of the inverter is determined based on at least one of the following: the difference between the second impedance to ground and the reference value of the impedance to ground; the difference between the sampled value of the secondary equivalent inductance voltage and the reference value of the secondary equivalent inductance voltage; the difference between the sampled value of the leakage current and the reference value of the leakage current; the changing trend of the reference value of the impedance to ground; and the second fault threshold.

6. The method according to claim 5, characterized in that, The second fault threshold includes a ground impedance threshold; based on the difference between the second ground impedance and the ground impedance reference value, and the changing trend of the ground impedance reference value, the second fault detection result of the inverter is determined, including: If the difference between the second impedance to ground and the reference value of the impedance to ground is greater than or equal to the impedance to ground threshold, then the second fault detection result is determined to be a fault detection result characterizing that the inverter has an impedance to ground fault. If the difference between the ground impedance and the ground impedance reference value is less than the ground impedance threshold, and the ground impedance reference value shows a continuous downward trend, then the second fault detection result is determined to be a fault detection result that characterizes the inverter having an insulation degradation trend.

7. The method according to any one of claims 1-3, characterized in that, The method further includes: The inverted third ground impedance is obtained based on the first sampled data; Based on the error signal between the first sampled data and the reference value of the insulation state to ground, and the second fault threshold, a fast time scale cumulative criterion, a medium time scale cumulative criterion, and a slow time scale cumulative criterion are constructed. For each time scale, a recursive accumulation operation with a forgetting coefficient is performed on the positive deviation component of the error signal that exceeds the second fault threshold to obtain the recursive accumulation result for the corresponding time scale; wherein, the recursive accumulation result of the fast time scale is used to identify sudden grounding faults; the recursive accumulation result of the medium time scale is used to identify continuous anomalies; and the recursive accumulation result of the slow time scale is used for long-term insulation degradation analysis. The level of insulation degradation trend is determined based on the recursive cumulative results of the fast time scale, the recursive cumulative results of the medium time scale, and the recursive cumulative results of the slow time scale corresponding to the decreasing slope and decreasing acceleration of the third ground impedance.

8. The method according to any one of claims 1-3, characterized in that, The first sampling data includes high-frequency local feature data and low-frequency trend feature data; the high-frequency local feature data includes the sampled value of the equivalent inductance voltage on the secondary side of the PID circuit and the sampled value of the leakage current; the low-frequency trend feature data includes the sampled value of the DC bus voltage, the sampled value of the positive bus to ground voltage, the sampled value of the negative bus to ground voltage, the inverter output power, the pulse width modulation switch status, and the sampled value of environmental parameters; The step of predicting the reference value of the inverter's insulation state to ground based on the first sampled data includes: High-frequency local features of the high-frequency local feature data are extracted using the first branch of the dual-branch time-series prediction model. The low-frequency trend features of the low-frequency trend feature data are extracted through the second branch of the dual-branch time series prediction model. The high-frequency local features and the low-frequency trend features are merged in the feature fusion layer of the dual-branch time series prediction model to obtain fused features; Based on the fusion characteristics, the following are output by the multiple output heads of the dual-branch timing prediction model: the reference value of the secondary side equivalent inductance voltage, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty.

9. The method according to claim 2, characterized in that, The method further includes: Based on the first sampling data, the ground insulation state reference value, and the second sampling data, a multi-feature diagnostic result is obtained; the multi-feature diagnostic result includes at least one of the following: the secondary side equivalent inductance voltage difference, the degree of voltage imbalance between the positive and negative busbars and ground, the leakage current difference, the trend of the ground impedance reference value, and the pulse response of the coded diagnostic pulse; Based on the multi-feature diagnostic results, the probability of occurrence of each fault category is adjusted; the fault category includes at least one of the following: internal grounding fault, external DC circuit grounding fault, AC side and filter branch abnormality, parasitic capacitance change, bus transient disturbance, and sensor abnormality.

10. An inverter fault detection system, characterized in that, The inverter fault detection system includes: a sampling module, an artificial intelligence recognition module, and a control module; The sampling module is used to collect the first sampling data of the inverter; The control module is used to obtain the inverted first ground impedance based on the first sampled data of the inverter, and to determine the first fault detection result of the inverter based on the first ground impedance and the first fault threshold. The artificial intelligence recognition module is used to identify the operating condition of the inverter based on the first sampled data when the first fault detection result indicates that no fault is detected in the inverter; the operating condition indicates the electrical state and disturbance level of the inverter; and predicts the reference value of the inverter's insulation state to ground based on the first sampled data; the reference value of the insulation state to ground includes: the reference value of the equivalent inductance voltage on the secondary side of the potential-induced decay PID circuit, the reference value of the leakage current, the reference value of the impedance to ground, and the prediction uncertainty; The control module is used to update the second fault threshold according to the ground insulation state reference value and the operating condition, and to determine the second fault detection result of the inverter according to the first sampling data, the ground insulation state reference value and the second fault threshold.

Images