Photovoltaic inverter insulation resistance intelligent detection system and method

Abstract

The invention discloses a photovoltaic inverter insulation resistance intelligent detection system and method, and particularly relates to the technical field of photovoltaic equipment operation safety monitoring, and the method comprises the following steps: generating frozen window shielding interference data through collecting a bus voltage in a state transition stage and constructing a dynamic convergence model; identifying an abnormal trend based on resistance derivative change in a stable state, inputting a dual-path decision process by combining environment and power grid parameters, and dynamically outputting an operation state level; control strategies such as remote prompt, power drop adjustment or output disconnection are executed according to the state grade, and accurate identification and grading response control of the insulation state are achieved; according to the invention, by constructing a voltage disturbance shielding mechanism, a trend derivative identification and delay confirmation strategy and a double-path safety state decision process with dynamic switching capability, high accuracy of insulation abnormity judgment, stability of trend identification and adaptability of a decision strategy are realized, and robustness and application range of a detection system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic equipment operation safety monitoring, and more specifically, to a photovoltaic inverter insulation resistance intelligent detection system and method. Background Art

[0002] As the core energy conversion device in the photovoltaic power generation system, the operation safety of the photovoltaic inverter is closely related to the electrical insulation performance. In order to ensure the long-term stable operation of the inverter and prevent leakage, breakdown or equipment damage caused by insulation degradation, it is usually necessary to monitor its insulation resistance parameters to the ground in real time during operation, and issue alarms or protection controls based on the detection results.

[0003] In the prior art, the method for judging the insulation resistance of the inverter mostly adopts a single-cycle sampling and fixed threshold comparison mode, that is, the currently detected insulation resistance value is compared with the preset safety limit value, and if it is lower than the threshold, it is judged as an abnormal state. However, this method has the following problems: during the startup or shutdown of the inverter, the bus voltage has a short-term jump due to the capacitor charging and discharging effect, which can easily cause instantaneous distortion of the insulation resistance value. If there is no isolation mechanism, it is easy to trigger misjudgment. Secondly, the inverter usually operates in outdoor scenes with frequent environmental fluctuations. The ambient temperature and humidity have a significant impact on the insulation characteristics. Their changes can cause periodic disturbances in the resistance measurement value, making the fixed threshold strategy lack stability and adaptability.

[0004] In addition, under complex working conditions such as drastic fluctuations in grid-connected conditions, unstable grid quality, and obvious nonlinear changes in loads, the single-point insulation resistance value cannot effectively identify trend degradation and is difficult to timely reflect the evolution of the insulation state. In particular, in the following typical usage scenarios, existing detection methods are more likely to fail: industrial and commercial distributed power stations where inverters are frequently started and stopped; tropical and coastal areas with high humidity and temperature or large contrast between day and night environments; complex grid-connected environments connected to fluctuating power grids and coexisting inductive loads or remote feeding equipment. The insulation monitoring method based on static judgment not only has limited accuracy, but also cannot identify potential trends, making it difficult to provide a decision-making basis for hierarchical response control and remote operation and maintenance management. Therefore, a photovoltaic inverter insulation resistance intelligent detection system and method are proposed in this paper to solve the above problems. Summary of the invention

[0005] To achieve the above object, the present invention provides the following technical solutions: A photovoltaic inverter insulation resistance intelligent detection method comprises the following steps: Obtain the inverter operating status, enter the state transition phase after detecting the start or shutdown event trigger, record the bus voltage change data in real time, and synchronize the state change event with the voltage sampling sequence; By means of voltage dynamic convergence modeling, extract the corresponding peak points, the starting points of voltage stable sections, and the convergence time nodes in the bus voltage curve, calculate the start and end time intervals of the freezing window, and shield the participation of insulation resistance data in the judgment logic within the freezing window; After the end of the freezing window, continuously collect the insulation resistance values, perform a first derivative calculation on the resistance increments of multiple consecutive sampling periods. When the rate of change of the resistance value continuously exceeds the set threshold for three or more consecutive periods, it is determined as an abnormal change trend and enters the status evaluation process; Input the results of the resistance change trend, ambient temperature, ambient humidity, load power, and grid voltage stability parameters into the safety status decision-making process. The safety status decision-making process includes two structures: a rule judgment path and a weight fusion path. Perform path selection during initial configuration or when the switching conditions are met, where the switching conditions are comprehensively judged based on the change amplitudes and stability indicators of multiple input variables. The safety status decision-making process outputs the operating status level, including normal status, trend warning status, restricted operation status, and emergency shutdown status; Execute the corresponding control instructions according to the operating status level. Maintain continuous monitoring in the normal status, report parameter records and issue remote prompts in the trend warning status, lower the output power to the set threshold in the restricted operation status, and immediately disconnect the output circuit in the emergency shutdown status to ensure electrical insulation safety.

[0006] In a preferred embodiment, the bus voltage change data in the state transition stage obtains a continuous voltage value sequence through a preset sampling frequency. The voltage dynamic convergence modeling method includes calculating the voltage difference between adjacent sampling points and constructing a voltage change rate curve, locating the peak point and the starting point of the convergence stable interval on this curve, and identifying the convergence time node based on the condition that the voltage change rate is lower than the stable threshold in multiple consecutive sampling periods; The start and end time intervals of the freezing window are dynamically calculated according to the time distance between the peak point and the convergence time node to automatically set the invalid interval for insulation resistance detection during the interference shielding period.

[0007] In a preferred embodiment, the result of the resistance change trend is obtained by performing a first derivative calculation on the insulation resistance data within consecutive sampling periods. The calculation method is the difference between the resistance value in the current period and the resistance value in the previous period divided by the sampling period length, and the result is the resistance change rate; When the resistance change rate exceeds the set rate threshold for three or more consecutive sampling periods, it is determined as an abnormal trend state. This rate threshold is determined by the voltage withstand characteristics of the insulating material, ambient temperature and humidity conditions, and the modeling results of historical data, and any one of a fixed empirical value, a segmented dynamic value, or a logical interval set based on a fuzzy boundary is selected for setting.

[0008] In a preferred embodiment, the calibration behavior of the resistance change trend has a delay confirmation mechanism. This mechanism starts two additional sampling periods to continue observing the resistance change after identifying a preliminary abnormal trend. If the resistance change rate in the new sampling periods maintains the same direction and continues to exceed the aforementioned rate threshold, the trend is confirmed. If the rate in any new period is lower than the threshold or the change direction is opposite, the calibration behavior is cancelled and restored to the normal state.

[0009] In a preferred embodiment, when the safety state decision process executes the weight fusion path, a multi-variable weighted scoring function is used to output the state level. This scoring function takes the resistance change trend result, ambient temperature, ambient humidity, load power, and grid voltage stability parameters as input variables. Each variable is uniformly normalized and converted into a score value before input; the scoring function multiplies each score value by a predefined weight factor and then sums them to form a state level score. This score is distributed between zero and one and is mapped to a normal state, a trend warning state, a restricted operation state, or an emergency shutdown state; the weight factors are configured through initialization parameters or generated by training with historical operation data.

[0010] In a preferred embodiment, the remote prompt behavior in the trend warning state includes generating a status information packet containing five types of data: the current insulation resistance change trend, the average ambient temperature and humidity within the past thirty seconds, the current load power, the grid voltage fluctuation range, and the start and end times of the freeze window, and sending it to the remote monitoring platform through the data upload channel; The power reduction behavior in the restricted operation state includes calculating the power reduction target value, which is the product of the current output power and the power reduction factor. The power reduction factor is jointly determined by the maximum change rate in the resistance change trend result and the current ambient temperature and humidity parameters. Its value is less than 1 and has multiple levels; if the grid voltage stability parameter is simultaneously lower than the stability threshold, the target power value further introduces a preset voltage attenuation compensation factor to perform an additional reduction.

[0011] In a preferred embodiment, the output loop interruption control in the emergency shutdown state is implemented through parallel dual-path logic. One path receives the emergency state signal output by the safety state decision process, and the other path receives auxiliary signal confirmation, including a verification event where the resistance value is lower than the insulation safety lower limit for three consecutive periods and the change rate direction is the same. The disconnection operation is only executed when both paths are activated.

[0012] In a preferred embodiment, the start and end intervals of the freezing time window can be generated by different calculation methods according to a preset strategy. The strategy includes one of the following three methods: The first method is to collect the voltage change rate curve. When it is detected that the change rate is lower than the voltage stability threshold value within five consecutive sampling periods, this interval is determined as the voltage convergence section. The starting point of the freezing window is the peak point of the curve, and the ending point is the starting point of this convergence section, forming a freezing interval between the peak point and the convergence point; The second method is to collect the maximum voltage fluctuation value of the bus voltage within the voltage stable section. When the fluctuation amplitude does not exceed the set offset threshold, the time interval corresponding to the consecutive sampling periods within the collected voltage stable section is used as the freezing window; The third method is to statistically calculate the average convergence time based on historical operating condition data and add a fixed safety margin as the window length to form a freezing interval starting from the peak point.

[0013] In a preferred embodiment, the safety status decision-making process includes two structures: a rule determination path and a weight fusion path. The rule determination path outputs the status level based on the conditional combination logic between input variables; the weight fusion path outputs the status level in a weighted calculation manner of normalized scores and weight factors; the two paths have a dynamic switching mechanism. The switching condition is that when the change amplitudes of at least two types of input variables exceed their respective stability thresholds within consecutive observation periods, or the variances of all variables exceed the set composite standard, it automatically switches from the rule determination path to the weight fusion path.

[0014] In a preferred embodiment, a photovoltaic inverter insulation resistance intelligent detection system includes a status recognition module, an electrical modeling module, a trend evaluation module, a safety decision module, and a remote interaction module. Each module collaborates to construct an intelligent recognition and protection response process for insulation performance, specifically including: The status recognition module is used to obtain the inverter operating status and bus voltage data in real time and identify the status transition periods corresponding to startup and shutdown events; The electrical modeling module is used to perform convergence modeling on the bus voltage change curve, extract the peak point and the starting point of the stable section, generate a freezing time window to shield the insulation resistance data during this period, and calculate the resistance change rate to construct a change trend curve; The trend evaluation module is used to judge the resistance change rate within multiple consecutive sampling periods. If the threshold conditions are continuously met, it is marked as an abnormal trend and input into the subsequent safety determination process together with environmental temperature and humidity, load power, and grid stability parameters; The safety decision module includes two decision-making structures: a rule determination path and a weight fusion path, and has a path switching mechanism. It selects the decision-making path according to the fluctuation degree of input variables and outputs the operating status level. The status levels include four categories: normal, trend warning, restricted operation, and emergency shutdown; The remote interaction module is used to generate a status data packet containing trend results, environmental information, electrical parameters, and the frozen window time period in the trend warning state or a higher-level state, and send it to the remote monitoring platform to achieve remote anomaly perception and system operation and maintenance linkage.

[0015] Technical effects and advantages of the present invention: By constructing a frozen window mechanism based on dynamic convergence modeling of bus voltage, the present invention effectively solves the problem of detection errors caused by voltage fluctuations during the startup or shutdown process of the inverter. The start and end times of the frozen window are dynamically calculated using the time interval between the peak point and the convergence time node, so that the insulation resistance data does not participate in the judgment logic during the disturbance period, ensuring that the detection behavior is only carried out during the voltage stable stage, thereby improving the accuracy of insulation anomaly identification, significantly reducing the false judgment probability caused by capacitor charging and discharging and instantaneous disturbances, and enhancing the anti-interference ability and adaptability of the system.

[0016] Through the resistance change trend derivative judgment mechanism and the delay confirmation strategy, the present invention realizes sensitive identification and stable calibration of the degradation trend of insulation performance. The derivative value exceeding the rate threshold in consecutive multiple sampling periods triggers the trend judgment, and a two-period delay confirmation is introduced to ensure the persistence and direction consistency of the trend change, excluding false positive judgments caused by short-term fluctuations. This trend determination logic enables the detection process to have the ability of time series identification, is applicable to complex working conditions where chronic insulation aging and sudden anomalies coexist, helps to intervene in risk control in advance, and enhances the forward-looking of system safety level judgment.

[0017] The present invention constructs a dual-structure safety state decision-making process composed of a rule judgment path and a weight fusion path, and is equipped with a dynamic switching mechanism based on multi-variable fluctuation indicators, enabling the state recognition strategy to have flexible adaptability. Under the condition of stable operation of variables, the rule path with high calculation efficiency is preferentially used to ensure the response speed; when there are severe disturbances in multiple variables and the risk of rule failure increases, the system automatically switches to the fusion path, relying on normalized scoring and weighted functions to achieve refined judgment. This dual-path structure improves the accuracy and robustness of judgment while ensuring real-time performance, and can widely adapt to the insulation safety management requirements in different photovoltaic scenarios. Brief Description of the Drawings

[0018] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the drawings; Figure 1 It is the schematic diagram of a method for intelligent detection of insulation resistance of a photovoltaic inverter in the present invention.

[0019] Figure 2 It is the schematic diagram of a system for intelligent detection of insulation resistance of a photovoltaic inverter in the present invention. Detailed Embodiments

[0020] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0021] Referring to Figure 1 - Figure 2 the following embodiments are obtained: Embodiment 1: The present invention provides an intelligent detection method for the insulation resistance of a photovoltaic inverter. First, a detection pre - mechanism based on operation - state perception and voltage - dynamic - feature recognition is established. During the state - transition stage of inverter startup or shutdown, by collecting a continuous voltage - value sequence and performing dynamic - convergence modeling, the peak point and the starting point of the voltage - stable section are extracted, the start - and - end time interval of the freeze window is calculated, and the influence of resistance data on the subsequent judgment logic is shielded within this window, thus effectively avoiding misjudgment caused by the change of bus surge voltage.

[0022] After entering the effective detection stage, the method continuously samples and calculates the derivative of the insulation - resistance data to construct the resistance - change trend, and a delay - confirmation mechanism is set up to identify a continuous abnormal - change trend. The judgment result and multiple parameters such as ambient temperature and humidity, load power, and grid - voltage stability are jointly input into the safety - state decision - making process. The safety - state decision - making process includes two structures: a rule - judgment path and a weight - fusion path, and has the ability of dynamic - path switching based on the fluctuation amplitude and stability index of the input variables. According to the operation - state level output by the decision - making process, the operation state is divided into a normal state, a trend - warning state, a restricted - operation state, and an emergency - shutdown state.

[0023] Each operation state corresponds to specific control behaviors. Among them, in the trend - warning state, a data packet containing five types of information will be generated and uploaded to the remote monitoring platform. In the restricted - operation state, the control is adjusted by calculating the power - reduction target value. In the emergency - shutdown state, a dual - path verification logic is used to control the interruption of the output loop. In addition to the standard calculation method for the start - and - end time of the freeze window, it also supports being generated based on the voltage - fluctuation amplitude or historical - convergence time modeling, realizing multi - strategy adaptation ability and improving the robustness and stability of the detection system for complex operation scenarios. The specific steps are as follows: Obtain the operating status of the inverter. After detecting the triggering of a start or shutdown event, enter the state transition phase, record the data of the bus voltage change in real time, and synchronize and bind the state change event with the voltage sampling sequence in time; construct a complete operating status perception framework to enable the detection behavior to have the ability to identify electrical state switching. By capturing state switching events such as start and shutdown, and combining the continuous sampling data of the bus voltage, the system can synchronously match the event nodes with the voltage fluctuation behavior, thus providing an accurate reference on the time axis for subsequent identification of voltage disturbance characteristics. This time synchronization mechanism constitutes the triggering basis of the detection process, enables the entire detection strategy to have dynamic response capabilities, can be tightly coupled with the inverter operating cycle, and avoids misstarting the judgment logic during the state transition phase.

[0024] Through the voltage dynamic convergence modeling method, extract the corresponding peak points, the starting points of the voltage stable sections, and the convergence time nodes in the bus voltage curve, calculate the start and end time intervals of the freezing window, and shield the participation of the insulation resistance data in the judgment logic within the freezing window; introduce a modeling mechanism for voltage disturbances to analytically identify the non-steady-state characteristics of the voltage during state changes and determine the effective coverage interval of the interference period. By extracting the peak points and the convergence sections, the system can accurately identify the key process of the bus voltage transitioning from severe fluctuations to stability, construct a freezing window to temporarily exclude the insulation resistance detection results in this stage, and prevent abnormal disturbances of the resistance value caused by physical phenomena such as bus capacitor charging and discharging from interfering with the judgment results. This mechanism provides a "data isolation period" for the judgment logic, ensures that the detection behavior is carried out under the premise of controllable signal quality, and enhances the immunity of the detection system to electrical disturbances.

[0025] After the end of the freezing window, continuously collect the insulation resistance values, perform a first derivative calculation on the resistance increments of multiple consecutive sampling periods. When the rate of change of the resistance value continuously exceeds the set threshold for three or more consecutive periods, it is determined as an abnormal change trend and enters the state evaluation process; extract the trend index from the dynamic change behavior of the insulation resistance to replace the traditional static threshold judgment logic, thereby realizing the early identification of potential degradation processes. By constructing a resistance change rate index through derivative calculation, the system can judge whether the insulation performance is in a steady state, a slow change state, or a continuously decreasing state. If the resistance value shows an over-limit behavior of the same direction rate for multiple consecutive periods, it indicates that there is a risk of structural degradation of the equipment, and it is necessary to enter the comprehensive evaluation process of the next stage in a timely manner. This strategy emphasizes the trend recognition ability, improves the sensitivity of the system to the insulation performance degradation process, and avoids single-point abnormal misjudgment and delayed response.

[0026] The results of the resistance change trend, ambient temperature, ambient humidity, load power, and grid voltage stability parameters are jointly input into the safety state decision-making process. The safety state decision-making process includes two structures: a rule judgment path and a weight fusion path. Path selection is performed during initial configuration or when the switching conditions are met, where the switching conditions are comprehensively judged based on the change amplitudes and stability indicators of multiple input variables. The safety state decision-making process outputs the operating state level, including normal state, trend warning state, restricted operation state, and emergency shutdown state; an intelligent judgment system with multi-source fusion and switchable paths is constructed to comprehensively evaluate the insulation state in multiple dimensions. The electrical trend information is combined with key variables such as environmental parameters, load information, and grid characteristics and input. Through the set judgment logic and weighted fusion model, the current operating state is classified into levels. The system uses the rule judgment path to handle simple scenarios with clear rules and the weight fusion path to process dynamic working conditions with complex variable interactions, and can switch paths according to the input fluctuation degree to ensure that the judgment strategy maintains accuracy and stability in various operating environments. This mechanism improves the intelligence and adaptability of state judgment and is the core of implementing dynamic protection strategies.

[0027] Corresponding control instructions are executed according to the operating state level. In the normal state, continuous monitoring is maintained. In the trend warning state, parameter records are reported and remote prompts are issued. In the restricted operation state, the output power is reduced to the set threshold. In the emergency shutdown state, the output loop is immediately disconnected to ensure electrical insulation safety. The state recognition result is linked with the actual operation control logic to construct a closed-loop control mechanism from recognition to response. In the normal state, the system maintains routine monitoring without intervention. In the trend warning state, the system generates data packets and transmits them to the remote monitoring platform to achieve remote alarm and trend retention. In the restricted operation state, the output is actively reduced according to the calculated power target value to avoid further deterioration caused by excessive electrical pressure. In the emergency shutdown state, the system ensures the accuracy of the interruption behavior through a dual-channel judgment logic and promptly cuts off the electrical output path to ensure the safety of personnel and equipment. This step accurately maps the judgment result to a hierarchical response behavior and is the key link to achieve risk closed-loop control and adaptive operation strategies.

[0028] In the intelligent detection of the insulation resistance of a photovoltaic inverter, it is first necessary to obtain the operating state of the inverter, that is, to collect and identify information such as the control instructions, output behavior, or status register of the photovoltaic inverter to determine whether the current inverter is in a "start", "run", or "shutdown" state. When a start or shutdown event trigger is detected, that is, when the operating state changes from a non-operating state to an operating state or vice versa, the system automatically enters the state transition stage. The state transition stage refers to the period when electrical quantities such as bus voltage and current fluctuate violently due to opening and closing, which has the characteristics of strong disturbance and high instability and is a key window for judging the dynamic behavior of the system.

[0029] During the state transition phase, the system will record the data of the bus voltage changes in real time. The bus voltage is the voltage value of the main power path connecting the photovoltaic modules and the inverter, reflecting the current electrical state of the equipment. The acquisition of voltage data relies on a preset sampling frequency to obtain a continuous sequence of voltage values, that is, a fixed sampling period is set, such as recording the bus voltage every 5 milliseconds, to form an original voltage sequence with continuous time and clear amplitude changes.

[0030] Subsequently, the system applies a voltage dynamic convergence modeling method to model and analyze the above voltage data. The so-called "modeling method" refers to the structural extraction of the trend of voltage change over time through digital processing means. The process includes calculating the voltage difference between adjacent sampling points, that is, performing a difference operation on two consecutive voltage sampling points to obtain the instantaneous amplitude change of the voltage change; and then constructing a voltage change rate curve with this voltage difference to analyze the severity and trend of voltage fluctuations.

[0031] On this change curve, the system further locates the peak points, that is, the highest or lowest value points that appear during the voltage change process, usually corresponding to the extreme voltage states at the beginning or end of the surge; and at the same time locates the starting point of the convergence steady interval, that is, the first time node when the voltage starts to change stably. The judgment basis for the above two points is: when the voltage change rate is lower than the stable threshold in multiple consecutive sampling periods, that is, the voltage change rate is less than a preset change rate (such as not exceeding five volts per second), the system considers that this section starts to enter the steady state interval, thereby identifying the convergence time node.

[0032] After obtaining the above two key time points, that is, the peak point and the convergence time node, the system calculates the start and end time intervals of the freeze window, and this start and end time is the time interval between the two points. This time interval is called the freeze window, and its function is to automatically set the invalid interval for insulation resistance detection during the interference shielding period. That is to say, within this freeze window, the system will ignore or shield the participation of the insulation resistance detection results to prevent misjudgment of electrical offset caused by severe disturbance of the bus voltage.

[0033] For example, during an inverter startup process, the system samples the bus voltage once every 5 milliseconds, and the voltage is collected to rise rapidly to the peak and then gradually decrease and enter the stable stage. According to the detection results, the peak point appears at the 45th millisecond after startup, and the convergence time node appears at the 190th millisecond. Then the freeze window interval is set from the 45th millisecond to the 190th millisecond after startup. During this period, regardless of whether the insulation resistance changes, it will not be used as the basis for state determination. Only the detection results outside the freeze window are used by the system to judge the insulation trend, ensuring the accuracy and anti-interference ability of the judgment logic.

[0034] This mechanism is connected with the subsequent resistance derivative calculation, resistance trend judgment, and safety status decision-making process, providing a reliable early anti-disturbance strategy for the entire detection method. It also forms a closed-loop logic chain with subsequent responses such as power reduction and disconnection control, and is a basic step to achieve accurate detection and stable operation.

[0035] The resistance change trend result is obtained by performing a derivative calculation on the insulation resistance data within the continuous sampling period. The insulation resistance data refers to the insulation resistance of the inverter to ground collected by the system in each fixed sampling period, generally in megohms. The continuous sampling period refers to multiple continuous time periods obtained by the system at the set sampling frequency, such as 10 milliseconds per period, five consecutive periods, etc.

[0036] The derivative calculation method is to divide the difference between the resistance value of the current cycle and the resistance value of the previous cycle by the length of the sampling cycle. Its physical meaning is the rate of change of the resistance value per unit time, that is, the "resistance change rate". This rate value is a scalar, and the unit is usually megaohms per second. A positive value indicates an increase in insulation, and a negative value indicates a decrease in insulation. The larger the absolute value, the more drastic the change. This calculation method is an approximate estimate of the first-order derivative and is suitable for digital control processing in industrial field environments.

[0037] The system uses three or more consecutive sampling cycles of resistance change rate as the analysis window. If the rate value exceeds the set rate threshold within the window, the system determines it as an abnormal trend state. "Three or more consecutive cycles" means that the abnormal rate behavior remains stable in three sampling periods, eliminating misjudgments caused by short-term random fluctuations; "set rate threshold" is the numerical threshold for judging the establishment of an abnormal trend. The rate must exceed this value continuously to be identified as a trend change.

[0038] The rate threshold is determined by a combination of factors, including: The withstand voltage characteristics of insulating materials: Different materials have different physical responses to insulation attenuation under high voltage, and the withstand voltage drop rate is used as a reference for the lower limit of the threshold; Ambient temperature and humidity conditions: The insulation value decreases faster in high humidity and high temperature environments, so a dynamic threshold needs to be set to adapt to environmental disturbances; Historical data modeling results: After running for a period of time, the system can generate a statistical distribution diagram of the resistance change rate, and then fit a reasonable threshold range.

[0039] To adapt to various working conditions, the threshold can be selected from the following three methods: Fixed experience value: Based on experiments or industry experience, a fixed rate threshold is set, such as a drop of no more than two megaohms per second; Segmented dynamic value: Divide segments by time, environment or device type, and use different thresholds for each segment; Logical interval based on fuzzy boundary setting: Set a slow-varying interval and adopt a fuzzy judgment strategy between "high-confidence trend" and "suspicious trend" in logical judgment to improve the fault tolerance and judgment sensitivity of the algorithm.

[0040] For example, in a certain operating environment, the sampling period is set to every 100 milliseconds. The system continuously detects the resistance values of one hundred megohms, ninety megohms, and seventy-five megohms three times within 300 milliseconds. The calculated result is a decrease of fifty megohms per second, far exceeding the empirical rate threshold of twenty megohms per second. The system then triggers an abnormal trend judgment and enters the subsequent state evaluation process. This method not only improves the forward-looking of detection but also enhances the continuous perception ability of the insulation slow-drop process, which is the key basis for realizing intelligent judgment strategies and multi-level security responses.

[0041] In the intelligent insulation resistance detection method of the photovoltaic inverter of the present invention, to improve the stability and accuracy of trend judgment, the system sets a calibration behavior for the resistance change trend after generating the resistance change trend result. The calibration behavior means that when a certain set of judgment conditions is met, the system determines it as the trend state and uses it as the input basis for the subsequent state evaluation process. To prevent misjudgment caused by transient disturbances or occasional jumps, after initially meeting the judgment conditions of the abnormal trend, it is not immediately calibrated as an effective trend, but a secondary verification is carried out through a delay confirmation mechanism.

[0042] The delay confirmation mechanism is a strategy structure automatically triggered after identifying the initial abnormal trend in the previous stage. Its working method is as follows: Start two additional sampling periods, that is, on the basis of the three consecutive sampling periods that have met the initial judgment conditions, extend two new sampling periods additionally to continue observing the change behavior of the subsequent insulation resistance. These two sampling periods are continuously extended in time and are usually sampled with the same sampling period setting as before. For example, if each period is 100 milliseconds, the delay confirmation period lasts for 200 milliseconds.

[0043] During this delay stage, the system calculates the resistance change rate in each new sampling period in real time. Its definition method is the same as described above, that is, the difference between the resistance value in the current period and the resistance value in the previous period divided by the sampling period length. The resulting value is the insulation resistance change rate per unit time, with the unit of megohm per second.

[0044] The delay confirmation mechanism sets two judgment criteria: One is the direction consistency: The resistance change rates in the two new periods should be consistent with the previously determined trend direction. The trend direction refers to whether the resistance value is continuously decreasing or continuously increasing. For example, if the previous three periods show that the resistance is continuously decreasing, the two new periods should also be decreasing; Second is threshold persistence: the resistance change rates of two new cycles should both continue to exceed the aforementioned rate threshold. The rate threshold is set as publicly disclosed above and is usually calculated based on insulating materials, electrical environments, and historical models.

[0045] If the resistance change rates of new sampling cycles maintain the same direction and continue to exceed the aforementioned rate threshold, the system considers that the resistance change is not an accidental disturbance but a continuous trend, confirms that the trend holds, and inputs it as the formal trend result into the safety state decision-making process. On the contrary, if the rate of any new cycle is lower than the threshold or the change direction is opposite, that is, there is a trend interruption or reverse behavior, the system determines that the previous abnormal trend judgment is a false trigger, cancels the calibration behavior, and resumes to the normal state.

[0046] For example, during a certain detection process, the system sampling period is set to every 100 milliseconds. The resistance values of the first three cycles are 100 megohms, 80 megohms, and 65 megohms in sequence, and the corresponding rates are all in the downward direction and exceed the set threshold of 20 megohms per second, triggering a preliminary abnormal trend judgment. At this time, it enters the delayed confirmation stage. The resistance values of the next two sampling cycles are 50 megohms and 35 megohms respectively, and the rates continue to be negative and the absolute values are greater than the threshold, so the system confirms that the trend holds. If the resistance value of one of the cycles rises to 70 megohms or the rate is less than 10 megohms per second, it is considered that the trend does not hold, and the system resumes to the non-abnormal state. The core significance of this mechanism is to suppress the false alarm rate and enhance robustness. In practical applications, electrical systems are easily affected by factors such as instantaneous interference, environmental fluctuations, or load switching. Without the delayed confirmation mechanism, accidental behaviors are easily misjudged as trend states, resulting in frequent jumps in control responses. By introducing the delayed period judgment, the system bases trend calibration on continuous behaviors, thereby improving the stability, accuracy, and anti-interference ability of state recognition.

[0047] The judgment result of the safety state is used to drive different control strategies, and its judgment process includes two structures: a rule judgment path and a weight fusion path. When the safety state decision-making process executes the weight fusion path, the system no longer judges the state level according to the preset rule logic, but constructs a function model based on multiple parameter inputs and calculates the state in a continuous numerical manner.

[0048] In this path, the system uses a multi-variable weighted scoring function to output the state level. The role of the scoring function is to weight and superimpose multiple input factors according to their importance and then output a unified scoring value, called the state level score, to reflect the comprehensive operating safety level of the current device.

[0049] The scoring function uses the resistance change trend result, ambient temperature, ambient humidity, load power, and grid voltage stability parameters as input variables. Among them: Result of resistance change trend: It refers to the resistance change rate index calculated from the aforementioned derivative, representing the speed and direction of the change in insulation performance; Ambient temperature and ambient humidity: They respectively represent the air temperature and air humidity collected in real time in the inverter operating environment, which have a significant impact on insulation deterioration; Load power: It refers to the real-time power value of the electrical load driven by the current inverter, usually obtained by multiplying voltage and current; Grid voltage stability parameter: It represents the fluctuation amplitude or fluctuation rate of the grid voltage over a period of time, and is used to evaluate the degree of disturbance of the external electrical environment.

[0050] Before input, the above variables are uniformly normalized and converted into score values. Normalization is to uniformly convert different physical quantities (such as temperature, power, voltage fluctuation) into dimensionless ratio values between zero and one, so as to perform weighted processing in the function. Usually, methods such as maximum-minimum normalization, sliding window normalization or empirical range normalization are adopted.

[0051] The scoring function multiplies each score value by a predefined weight factor and then sums them to form a state level score. Among them, the score value is the normalization result, and the weight factor is a coefficient set in advance or obtained through adaptive learning, which is used to characterize the influence degree of each input variable on the state level. For example, if the result of the resistance change trend has a greater influence weight and the ambient humidity has a smaller influence, then the former has a higher weight.

[0052] This score is distributed between zero and one and is mapped to a normal state, a trend warning state, a restricted operation state or an emergency shutdown state. The specific mapping relationship can be set according to the piecewise rule. For example, zero to 0.25 is the normal state; 0.251 to 0.5 is the trend warning state; 0.51 to 0.75 is the restricted operation state; exceeding 0.75 is the emergency shutdown state, and this interval division can be adjusted according to the actual operation safety tolerance.

[0053] The weight factor is configured through initialization parameters or generated by training with historical operation data. This setting method is used to determine the influence proportion of each input variable in the state level scoring function and is the core parameter for constructing the mathematical model of the scoring function. First, the initialization parameter configuration means that in the initial stage of system design or deployment, a set of fixed weight values are obtained in advance through engineering experience, industry standards, or simulation experiments and written into the system initial configuration file. This method is applicable to equipment scenarios where operation data has not been accumulated yet. For example, for an initially installed inverter device, the initial weight factors of each variable can be set according to the following rules: the result of the resistance change trend: 0.45; the ambient temperature: 0.15; the ambient humidity: 0.1; the load power: 0.15; the grid voltage stability parameter: 0.15. Such initial configuration values can be set differently according to equipment type, installation location, or operation and maintenance preferences, and can also be pre-stored in the control chip or set through the user interface parameter menu to ensure the executability of the scoring function.

[0054] Generating by training with historical operation data means that during the long-term operation of the system, the historical corresponding relationship between the input variables and the state level output is recorded, and the allocation ratio of the weight factor is automatically optimized through a training algorithm to make the scoring function more in line with the actual operation characteristics and safety judgment objectives. The specific training method may include the following steps: Data collection: In each detection cycle, the system automatically saves the result of the resistance change trend, the ambient temperature, the ambient humidity, the load power, the grid voltage stability parameter, and the operation state level finally judged by the system at that time; Sample construction: Using the five input variables as feature vectors and the state level as labels, a historical training sample set is constructed; Parameter fitting: Adopting multivariate regression, gradient descent, or other machine learning optimization algorithms to fit the weight factor so that the output result of the scoring function is closest to the historical label state level; Result verification and application: The accuracy of the training result is evaluated through methods such as cross-validation. On the premise of meeting the judgment accuracy requirements, the weight parameters are written into the current model to replace the initialization weight factor and serve as the effective scoring basis during the current operation cycle. For example, after learning from 30 days of operation data, the system finds that the grid voltage stability is more sensitive to the prediction of abnormal states, automatically adjusts its weight from the initial 0.15 to 0.25, and at the same time moderately reduces the weight of the ambient humidity. This behavior does not require manual intervention, and the system can train and automatically switch in the background to achieve the online evolution of the scoring model. Through the combined mechanism of initialization configuration and training generation, the weight factor setting method in the present invention not only meets the usability of the system's initial rapid deployment but also ensures the accuracy and adaptability of the model during long-term operation, thereby enhancing the generalization performance, stability, and prediction accuracy of the state level evaluation function.

[0055] For example, in a certain on-site application, the score for the resistance change trend is 0.85, the score for the ambient temperature is 0.6, the humidity is 0.7, the load power is 0.5, and the voltage stability score is 0.95. The corresponding weight factors are 0.35, 0.1, 0.1, 0.2, and 0.25 respectively. The final state level score is the sum of the products of the above items, which is 0.78, falling into the "emergency shutdown state" interval, and the system immediately enters the disconnection loop response. This method has significant numerical adaptability, model trainability, and continuity of state judgment, avoiding the problem of the rule path failing in the multi-variable coupling state, and providing double guarantees of accuracy and stability for the comprehensive judgment of the dynamic change of insulation performance.

[0056] The judgment result of the state level will trigger corresponding control behavior responses to ensure the safe operation of the equipment and provide operation and maintenance support. Among them, after different operating state levels are identified, the system will respectively execute the remote prompt behavior in the trend warning state and the power reduction behavior in the restricted operation state, corresponding to the intelligent response strategies in the two abnormal state levels of "trend warning state" and "restricted operation state".

[0057] The remote prompt behavior in the trend warning state means that when the state level is judged to be the trend warning state, the system automatically generates a data set reflecting the current electrical state change trend and environmental change parameters for remote monitoring and analysis. The core steps of this behavior include: generating a status information packet containing five types of data: the current insulation resistance change trend, the average ambient temperature and humidity in the past 30 seconds, the current load power, the grid voltage fluctuation range, and the start and end times of the freeze window. Among them: Current insulation resistance change trend: It is the result of the most recent trend identification, including the change direction (rising or falling) and the change rate; Average ambient temperature and humidity in the past 30 seconds: The system collects all ambient temperature and humidity samples in the past 30 seconds and calculates their arithmetic mean to smooth short-term disturbances; Current load power: It is the inverter output power value in the current detection period; Grid voltage fluctuation range: It is the difference between the maximum and minimum values of the grid voltage within the set time window, reflecting the power supply stability of the grid; Start and end times of the freeze window: It is the freeze detection shielding period corresponding to the start or shutdown event, used to support remote state reconstruction.

[0058] The above five types of data are packed to form a status information packet, which is sent to the remote monitoring platform via the data upload channel. The data upload channel can be a local area network, a wired communication link, a wireless cellular network, or other data transmission means; the remote monitoring platform is a background system with data receiving, parsing, visualization, and alarm management functions, used to remotely identify the equipment operation trend, perform predictive maintenance, construct a status evolution curve, etc. Through the triggering of remote prompt actions and data transmission, the operation and maintenance personnel can take proactive intervention before a failure occurs.

[0059] When the status level is judged to be a restricted operation state, the system will execute a power reduction behavior to reduce the electrical output load of the inverter, alleviate the trend of insulation deterioration, and extend the safe operation time of the equipment. This behavior includes: calculating the power reduction target value, which is the product of the current output power and the power reduction factor. Among them: Current output power: It is the real-time output power of the equipment within the current cycle, and the unit can be watt or kilowatt; Power reduction factor: It is a coefficient less than one, used to calculate the target power, representing the ratio limit of the allowable output power.

[0060] The power reduction factor is jointly determined by the maximum change rate in the resistance change trend result and the current ambient temperature and humidity parameters, that is, this factor is not fixedly set, but jointly determined by electrical parameters and environmental factors: The maximum change rate in the resistance change trend result: It refers to the maximum value of the resistance decrease rate in the recent several cycles, representing the severity of insulation degradation; Current ambient temperature and humidity parameters: That is, the temperature and humidity values within the current detection cycle, which have a non-linear impact on the insulation performance.

[0061] Based on the above two input variables, the system can set multiple power reduction levels, which can be set based on a fuzzy inference engine or other preset logics. For example: when the resistance decrease rate is low and the temperature and humidity are normal, the factor is 0.9; when the resistance decreases rapidly and the ambient humidity is high, the factor is reduced to 0.75; if in extreme temperature and humidity conditions at the same time, the factor can be reduced to 0.5.

[0062] If the power grid voltage stability parameter is simultaneously lower than the stability threshold, that is, the power grid voltage fluctuation amplitude exceeds the safe limit allowed by the equipment, the system will introduce a preset voltage attenuation compensation factor on the basis of the target power to perform an additional reduction. The voltage attenuation compensation factor is a predefined coefficient, usually with a value below 0.9, used to further reduce the load pressure when the power grid interference and the internal fault risk of the equipment are superimposed.

[0063] For example, the current output power is 5 kW, the resistance change rate is relatively large, the power reduction factor is set to 0.6, and the target power is 3 kW. At this time, the grid voltage stability is poor, and the compensation factor is 0.9. Then the final output power is 3000 multiplied by 0.9, that is, 2700 W. Through this power reduction behavior, when the system detects that the insulation risk has not reached the disconnection requirement but has a development trend, it actively reduces the system output, which not only maintains the power generation continuity but also controls the risk boundary. It is the key strategy to realize the transfer from "operation with faults" to "active pre-control".

[0064] When the system identifies that the device is in an emergency shutdown state, to ensure electrical insulation safety, it is necessary to immediately interrupt the electrical output path to prevent breakdown, leakage, or personal safety risks caused by further operation. For this purpose, a disconnection mechanism with safety redundancy is constructed, that is, the output loop interruption control is realized through a parallel dual-path logic. The output loop interruption control refers to controlling the conduction state between the AC output part of the inverter and the external load, so that it cuts off the output under specific conditions to achieve electrical isolation. This action is generally completed by the control unit driving a relay, a contactor, or a semiconductor switching device. The disconnection operation in the present invention does not depend on a single judgment result, but is realized through a parallel dual-path logic, that is, the disconnection action will only be executed when the activation conditions of both judgment paths are satisfied simultaneously, thereby constructing logical redundancy to avoid mis-triggering.

[0065] One path receives the emergency state signal output by the safety state decision-making process. This signal is generated by the aforementioned weight fusion path or rule determination path after determining that the system is in the "emergency shutdown state", representing that the system has completed the analysis of all input variables and determined that there is an unacceptable insulation risk in the current state, and immediate hardware disconnection is required. This signal is the first judgment basis and has the decision-making power of global state information. The other path receives the auxiliary signal confirmation. This auxiliary path is independent of the main path judgment logic and is dedicated to performing "underlying verification" on key physical quantities to further ensure that the actual detection result has indeed reached the irrecoverable threshold, thereby preventing unnecessary disconnection operations caused by misjudgment of the main path. The source of this auxiliary signal includes a composite judgment condition: A verification event where the resistance value is lower than the insulation safety lower limit for three consecutive cycles and the change rate direction is the same. Among them, the resistance value being lower than the insulation safety lower limit for three consecutive cycles means that the insulation resistance values collected by the system in three consecutive sampling cycles are all lower than a preset minimum allowable resistance value. For example, this safety lower limit can be set to 20 MΩ; the change rate direction being the same means that the change trends of the resistance values in three cycles are all decreasing and there is no reverse behavior, that is, the derivative values have the same direction. The design of this composite condition is used to judge whether the current insulation state is continuously deteriorating and the value has fallen below the safety bottom line, which is a logical binding of the "trend" and "absolute value" dual indicators.

[0066] The disconnection operation is performed only when both paths are activated, that is, only when the status level judgment path gives an "emergency stop" signal and the auxiliary judgment path confirms that the actual detection parameters meet the dual characteristics of deterioration and over-limit, the system will start the loop interruption control logic. Otherwise, even if the main path outputs an emergency signal, if the auxiliary judgment condition is not met, the disconnection will not be performed temporarily; vice versa.

[0067] An example is as follows: the system sampling period is every 100 milliseconds, and the current output power is four kilowatts. In the last three sampling periods, the resistance values were 18 megohms, 15 megohms, and 12 megohms, respectively, all lower than the safety lower limit of 20 megohms, and the derivative values were all negative, that is, they continued to decline, meeting the auxiliary signal conditions; at the same time, the safety state decision process is calculated through the scoring function, and the output state level is "emergency shutdown state". At this time, both channels are activated, the disconnection logic takes effect, the relay is activated, and the output channel is immediately cut off. Through this dual-channel structure, on the one hand, a fast disconnection response is guaranteed in a real emergency state, and on the other hand, the risk of false disconnection is reduced, especially in the case of short-term fluctuations, environmental interference and other non-persistent risks. The system can avoid unnecessary interruption behavior and improve the accuracy of protection control and system stability.

[0068] In order to effectively shield the transient interference caused by the drastic fluctuation of bus voltage during the startup or shutdown process of the inverter, the system needs to build a freezing time window to temporarily shield the participation of insulation resistance data in the state judgment during this period. In order to adapt to different electrical conditions and operating scenarios, the generation of the freezing time window does not adopt a fixed value setting, but has a strategic dynamic configuration mechanism.

[0069] Specifically, the start and end intervals of the freezing time window can be generated by different calculation methods according to the preset strategy, that is, the system can select different strategy types through the operation parameter setting to dynamically construct the freezing window boundary. The start and end intervals here refer to the starting time and the end time of the freezing window. The window is essentially a time period, which is used to determine whether the resistance data is not accepted during this period.

[0070] The strategy includes one of the following three methods, that is, the system only uses one of the methods to build the window according to the device configuration or operating conditions. The three methods are: The first method is to collect the voltage change rate curve. That is, by performing high-frequency sampling on the bus voltage, calculate the voltage difference between each sampling point and its adjacent sampling point to form the voltage change rate curve. The voltage change rate refers to the change amplitude of the voltage value per unit time, usually measured in volts per second. In this curve, the system determines whether the change rate is lower than the voltage stability threshold within five consecutive sampling periods. The stability threshold is a pre-set rate threshold, representing the maximum voltage fluctuation rate acceptable to the system, such as ten volts per second. If the change rate is lower than this value within five consecutive periods, the system determines this section as the voltage convergence section, that is, the voltage has gradually transitioned from an unstable state to a stable state. At this time, the system sets the starting point of the freeze window as the peak point of the curve, that is, the time point at the maximum or minimum value that appears during this voltage fluctuation process, and sets the end point as the starting point of this convergence section, that is, the first time node of the continuous stable section, thus forming a freeze interval between the peak and the convergence point. This method emphasizes the characteristics of the voltage fluctuation curve and is used to accurately cover the disturbance period.

[0071] The second method is to collect the continuous sampling periods within the voltage stable section and calculate the maximum fluctuation value of the bus voltage within this period; if the fluctuation amplitude does not exceed the set offset threshold, the time interval corresponding to this continuous sampling period is defined as the freeze window. The bus voltage stable section refers to the voltage interval after the system enters stable operation, and the sampling voltage fluctuates slightly within this interval. The system analyzes the difference between the maximum and minimum values of the voltage in this section and defines it as the voltage fluctuation amplitude. If this amplitude does not exceed the set offset threshold, that is, the voltage fluctuation is within the safe allowable range (for example, less than five volts), the system determines that the voltage in this interval is stable. At this time, the system can use the time interval corresponding to the continuous sampling periods within the voltage stable section as the freeze window, that is, use this time period as the safety buffer time after the disturbance ends. This method has looser judgment conditions than the first method and is applicable to scenarios where the fluctuation amplitude of the equipment operating point is not large. The set offset threshold is a factory configuration or user-set value used to define "whether the fluctuation is acceptable".

[0072] The third method is to statistically calculate the average convergence time based on historical operating condition data and add a fixed safety margin as the window length. Historical operating condition data refers to the voltage fluctuation process data recorded after the equipment has experienced multiple startup, shutdown, and other events during past operations. The system statistically analyzes this data and extracts the average value of the time required from the voltage peak to the stable point each time, denoted as the average convergence time. To ensure safe coverage, the system adds a pre-defined fixed safety margin, such as twenty milliseconds, to enhance the shielding boundary. Finally, the freeze window starts from the peak point and extends for the length of this "average convergence time + margin" to form a freeze interval. This method is applicable to equipment after big data training, has on-site adaptive capabilities, and can be used to deploy a general model for similar equipment.

[0073] For example, in a certain on-site environment, the bus voltage undergoes a drastic change after startup. The voltage peak appears at the 60th millisecond, and then the change rate is less than 8 volts per second for five consecutive sampling periods (each period is 10 milliseconds), meeting the conditions of the first method. The system sets the 60th millisecond as the starting point and the 110th millisecond as the ending point, and the freeze window length is 50 milliseconds. The insulation resistance data collected within this window is not involved in the subsequent state level judgment. By providing the above three policy methods and implementing a mutually exclusive selection structure, the system has a flexible freeze window generation ability and can be called according to the actual scenario, user configuration, or control logic. Without increasing the sampling burden, it improves the coverage, determination accuracy, and environmental adaptability of the freeze mechanism for abnormal disturbances.

[0074] When the system identifies whether there is a downward trend in insulation performance, it needs to perform parallel analysis on multiple influencing factors to output an accurate and reliable operating state level. To achieve this goal, a judgment framework with a dual-structured path is constructed, that is, the safety state decision process includes two structures: a rule determination path and a weight fusion path. This structure is flexible in control logic and adaptable in implementation, and is an important part of the present invention to improve the reliability of state recognition. The rule determination path means that the system outputs the state level based on the conditional combination logic between input variables, that is, a set of empirical or expert rules are set to perform logical combination judgment on the specific value relationships between input parameters. For example: if the result of the resistance change trend exceeds the warning threshold and the environmental humidity is higher than the set standard, it is determined as the trend warning state; if the resistance rate drops sharply and the grid voltage fluctuates abnormally at the same time, the restricted operation state is output. This path is applicable to scenarios where the variable change rules are clear and the boundaries can be described regularly. Its advantage is simple calculation and rapid response, and it is suitable for deployment on low-computing power terminals.

[0075] The weight fusion path provides another evaluation ability for the system, and outputs the state level in a normalized score and weight factor weighted calculation method. The specific process includes: uniformly normalizing all input variables (resistance change trend result, environmental temperature, environmental humidity, load power, grid voltage stability parameter) into a score value between 0 and 1; multiplying each score value by the corresponding predefined weight factor and summing them up weighted to output a state level score; the state level score falling into different intervals is mapped to the normal state, trend warning state, restricted operation state, or emergency shutdown state. This path is applicable to scenarios where the interaction between variables is complex, the relationship is non-linear, or it is difficult to describe regularly, and can provide a more continuous state output result, adapting to complex environments.

[0076] To improve the overall adaptability and determination accuracy of the method, the system is designed with two paths having a dynamic switching mechanism, that is, the system can select the applicable path during operation according to the fluctuation of input parameters. The specific switching conditions are that the change amplitudes of at least two types of input variables exceed their respective stability thresholds simultaneously within a continuous observation period, or the variances of all variables exceed the set composite standard. Among them: the change amplitude refers to the maximum change amount of a variable within a period of time; the stability threshold is the maximum allowable fluctuation limit set by the system. For example, the temperature fluctuation exceeds three degrees Celsius, and the load power changes by more than twenty percent; the variable variance represents the degree of dispersion of the values of the variable within a certain time window and measures the volatility; the set composite standard is a set of variance threshold standards set for multiple variables to ensure that the overall judgment is based on multiple dimensions and controllable. The design of automatically switching from the rule determination path to the weight fusion path enables the system not to cause incorrect state judgment due to rule failure in a complex operating environment with unstable, abnormal fluctuations, and ineffective rules.

[0077] Although the rule path has a fast response and a small amount of calculation, it is prone to misjudgment when the variable changes are unstable or the logical relationship is unclear; although the weight path is accurate, it has complex calculations and a high system load. Therefore, the dual-path complementary design can dynamically allocate judgment resources. Photovoltaic inverters are often deployed in areas with variable environments, such as high humidity, high altitude, and areas with severe power grid interference, where multiple parameter disturbances coexist; this mechanism can automatically identify the variable fluctuation characteristics, select a more appropriate state evaluation method, and improve the robustness. When the system runs stably, the rule path is preferred to ensure a fast response; once the variables fluctuate violently, it automatically switches to the weight path to ensure the accuracy and coherence of the judgment result and avoid frequent state jumps. The path switching mechanism can be enabled according to the system configuration under different operation and maintenance strategies, computing resources, or operation models, without being limited to a fixed architecture, improving the generality and engineering practicality of the overall technical solution.

[0078] Specifically, an example is given to illustrate: in a certain operating scenario, the system recognizes that the change trend score of the resistance fluctuates from 0.5 to 0.9 in the current five sampling periods, the environmental temperature fluctuates by six degrees Celsius, and the humidity changes by 30%, which has exceeded the set stability threshold. The system automatically switches from the rule determination path to the weight fusion path, recalculates the state level score in the fusion path, outputs the "emergency shutdown state", and triggers the disconnection logic. If not switched, the rule path may judge it as the "trend warning state" due to not covering the combination relationship of this type of variable, resulting in a response delay.

[0079] Embodiment 2: An intelligent insulation resistance detection system for a photovoltaic inverter, including a state recognition module, an electrical modeling module, a trend evaluation module, a safety decision module, and a remote interaction module. Each module collaborates to construct an intelligent recognition and protection response process for insulation performance, specifically including: The status recognition module is used to obtain the inverter operation status and bus voltage data in real time, and identify the status transition periods corresponding to startup and shutdown events; The electrical modeling module is used to perform convergence modeling on the bus voltage change curve, extract the peak points and the starting points of the stable sections, generate a frozen time window to shield the insulation resistance data during this period, and calculate the resistance change rate to construct a change trend curve; The trend evaluation module is used to judge the resistance change rate within multiple consecutive sampling periods. If the threshold conditions are continuously met, it is calibrated as an abnormal trend, and it is jointly input into the subsequent safety determination process together with the environmental temperature and humidity, load power, and grid stability parameters; The safety decision-making module includes two decision-making structures: a rule determination path and a weight fusion path, and has a path switching mechanism. It selects a decision-making path according to the fluctuation degree of the input variables and outputs the operation status level. The status levels include normal, trend warning, restricted operation, and emergency shutdown; The remote interaction module is used to generate a status data packet including trend results, environmental information, electrical parameters, and the frozen window time period in the trend warning state or a higher-level state, and send it to the remote monitoring platform to realize remote anomaly perception and system operation and maintenance linkage.

[0080] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0081] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0082] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0083] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0084] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claimed rights.

Claims

1. An intelligent detection method for the insulation resistance of a photovoltaic inverter, characterized in that, It includes the following steps: Obtain the operating status of the inverter. After detecting the triggering of a startup or shutdown event, enter the state transition stage, record the bus voltage change data in real time, and time-synchronously bind the state change event with the voltage sampling sequence; Extract the corresponding peak points, the starting points of the voltage stable section, and the convergence time nodes in the bus voltage curve through the voltage dynamic convergence modeling method, calculate the start and end time intervals of the freezing window, and shield the participation of the insulation resistance data in the judgment logic within the freezing window; After the end of the freezing window, continuously collect the insulation resistance values, perform a first derivative calculation on the resistance increments of multiple consecutive sampling periods. When the resistance value change rate continuously exceeds the set threshold for three or more periods, it is determined as an abnormal change trend and enter the state evaluation process; Input the resistance change trend result, ambient temperature, ambient humidity, load power, and grid voltage stability parameters into the safety state decision-making process. The safety state decision-making process includes two structures: a rule judgment path and a weight fusion path. Perform path selection during initial configuration or when the switching conditions are met, where the switching conditions are comprehensively judged based on the change amplitudes and stability indicators of multiple input variables. The safety state decision-making process outputs the operating state level, including normal state, trend warning state, restricted operation state, and emergency shutdown state; Execute the corresponding control instructions according to the operating state level. In the normal state, maintain continuous monitoring. In the trend warning state, report the parameter records and issue a remote prompt. In the restricted operation state, reduce the output power to the set threshold. In the emergency shutdown state, immediately disconnect the output circuit to ensure electrical insulation safety.

2. The intelligent insulation resistance detection method for a photovoltaic inverter according to claim 1, wherein The bus voltage change data in the state transition stage obtains a continuous voltage value sequence through the preset sampling frequency. The voltage dynamic convergence modeling method includes calculating the voltage difference between adjacent sampling points and constructing a voltage change rate curve, positioning the peak point and the starting point of the convergence stable interval on this curve, and identifying the convergence time node based on the condition that the voltage change rate is lower than the stable threshold in multiple consecutive sampling periods; The start and end time intervals of the freezing window are dynamically calculated according to the time distance between the peak point and the convergence time node to automatically set the invalid interval for insulation resistance detection during the interference shielding period.

3. The intelligent detection method for the insulation resistance of a photovoltaic inverter according to claim 2, wherein The resistance change trend result is obtained by performing a first derivative calculation on the insulation resistance data within consecutive sampling periods. The calculation method is the difference between the current period resistance value and the previous period resistance value divided by the sampling period length, and the result is the resistance change rate; When the resistance change rate exceeds the set rate threshold for three or more consecutive sampling periods, it is determined as an abnormal trend state. The rate threshold is determined by the voltage withstand characteristics of the insulating material, ambient temperature and humidity conditions, and the modeling results of historical data, and is set in any one of the ways of selecting a fixed empirical value, a segmented dynamic value, or a logical interval based on a fuzzy boundary.

4. A method for intelligent detection of the insulation resistance of a photovoltaic inverter according to claim 3, characterized in that, The calibration behavior of the resistance change trend has a delayed confirmation mechanism. This mechanism starts two additional sampling periods to continue observing the resistance change after identifying a preliminary abnormal trend. If the resistance change rate in the new sampling period remains in the same direction and continues to exceed the aforementioned rate threshold, the trend is confirmed. If the rate in any new period is lower than the threshold or the change direction is opposite, the calibration behavior is cancelled and restored to the normal state.

5. The intelligent insulation resistance detection method for a photovoltaic inverter according to claim 4, wherein, When the safety state decision process executes the weight fusion path, a multi-variable weighted scoring function is used to output the state level. This scoring function takes the resistance change trend result, ambient temperature, ambient humidity, load power, and grid voltage stability parameter as input variables. Each variable is uniformly normalized and converted into a scoring value before input; the scoring function multiplies each scoring value by a predefined weight factor and then sums them to form the state level score. This score is distributed between zero and one and is mapped to a normal state, trend warning state, restricted operation state, or emergency shutdown state. The weight factor is configured through initialization parameters or generated by training with historical operation data.

6. The intelligent insulation resistance detection method of a photovoltaic inverter according to claim 5, wherein The remote prompting behavior in the trend warning state includes generating a status information packet containing five types of data: the current insulation resistance change trend, the average ambient temperature and humidity within the past thirty seconds, the current load power, the grid voltage fluctuation amplitude, and the start and end times of the freeze window, and sending it to the remote monitoring platform through the data upload channel. The power reduction behavior in the restricted operation state includes calculating the power reduction target value, which is the product of the current output power and the power reduction factor. The power reduction factor is jointly determined by the maximum change rate in the resistance change trend result and the current ambient temperature and humidity parameters, and its value is less than 1 and has multiple levels; if the grid voltage stability parameter is simultaneously lower than the stability threshold, the target power value further introduces a preset voltage attenuation compensation factor to perform an additional reduction.

7. A method for intelligent detection of the insulation resistance of a photovoltaic inverter according to claim 6, characterized in that, The output loop interruption control in the emergency shutdown state is implemented through parallel dual-path logic. One path receives the emergency state signal output by the safety state decision process, and the other path receives auxiliary signal confirmation, including a verification event where the resistance value is lower than the insulation safety lower limit for three consecutive periods and the change rate direction is the same. The disconnection operation is only executed when both paths are activated.

8. The intelligent detection method for the insulation resistance of a photovoltaic inverter according to claim 7, wherein The start and end intervals of the freeze time window can be generated using different calculation methods according to the preset strategy. The strategies include one of the following three methods: The first method is to collect the voltage change rate curve. When it is detected that the change rate is lower than the voltage stability threshold value within five consecutive sampling periods, this interval is determined as the voltage convergence section. The starting point of the freeze window is the peak point of the curve, and the ending point is the starting point of this convergence section, forming a freeze interval between the peak and the convergence point; the second method is to collect the maximum bus voltage fluctuation value within the voltage stable section. When the fluctuation amplitude does not exceed the set offset threshold, the time interval corresponding to the continuous sampling periods within the collected voltage stable section is used as the freeze window; the third method is to statistically calculate the average convergence time based on historical operating conditions data and add a fixed safety margin as the window length, starting from the peak point to form the freeze interval.

9. The intelligent detection method for the insulation resistance of a photovoltaic inverter according to claim 8, wherein, The safety status decision-making process includes two structures: a rule judgment path and a weight fusion path. The rule judgment path outputs the status level based on the conditional combination logic between input variables; The weight fusion path outputs the status level through a weighted calculation method of normalized scores and weight factors. The two paths have a dynamic switching mechanism. The switching condition is that when the change amplitudes of at least two types of input variables exceed their respective stability thresholds within consecutive observation periods, or when the variances of all variables exceed the set composite standard, it automatically switches from the rule judgment path to the weight fusion path.

10. A photovoltaic inverter insulation resistance intelligent detection system, based on the photovoltaic inverter insulation resistance intelligent detection method according to any one of claims 1-9, characterized in that, It includes a status recognition module, an electrical modeling module, a trend evaluation module, a safety decision-making module, and a remote interaction module. Each module collaborates to construct an intelligent recognition and protection response process for insulation performance, specifically including: The status recognition module is used to obtain the inverter operating status and bus voltage data in real time and identify the status transition periods corresponding to start and shutdown events; The electrical modeling module is used to perform convergence modeling on the bus voltage change curve, extract the peak points and the starting points of the stable sections, generate a frozen time window to shield the insulation resistance data during this period, and calculate the resistance change rate to construct a change trend curve; The trend evaluation module is used to judge the resistance change rate within multiple consecutive sampling periods. If the threshold conditions are continuously met, it is marked as an abnormal trend and jointly input into the subsequent safety judgment process together with the environmental temperature and humidity, load power, and grid stability parameters; The safety decision-making module includes two decision-making structures: a rule judgment path and a weight fusion path, and has a path switching mechanism. It selects the decision-making path based on the fluctuation degree of input variables and outputs the operating status level. The status levels include four categories: normal, trend warning, restricted operation, and emergency shutdown; The remote interaction module is used to generate a status data packet containing trend results, environmental information, electrical parameters, and the frozen window time period in the trend warning state or a higher-level state, and send it to the remote monitoring platform to achieve remote anomaly perception and system operation and maintenance linkage.

Images