A method for detecting insulation of a direct current system

Abstract

The application discloses a kind of DC system insulation detection methods, it is related to the technical field of insulation defect detection, including the following steps, detection node arrangement, installation detection node between positive bus and negative bus of DC system and ground, additional installation reference detection channel, voltage correction, characteristic reference value analysis, insulation drop determination, based on the fluctuation intensity of voltage data analysis after correction, in the system health operation stage, dynamically calculate fault determination threshold, fault branch analysis, with the reference channel differential correction, eliminate common-mode disturbance, accurately extract the insulation characteristics of each node;Based on the dynamic adjustment of fault threshold value of health fluctuation and real-time standard deviation, adaptive system state, avoid false alarm of fixed threshold, improve early sensitivity;After determining insulation drop, first with synchronous difference degree three times reference screening suspected branch, then through positive and negative mismatch score live positioning fault branch, no power-off pull road, give consideration to anti-interference and the effect of investigation efficiency.

Description

Technical Field

[0001] This invention relates to the field of insulation defect detection technology, specifically to a method for detecting insulation in a DC system. Background Technology

[0002] A DC system is an independent power supply device composed of charging equipment, batteries, and a DC power supply panel. It features input and output overvoltage protection and provides stable power to relay protection devices, circuit breaker operation, and signal circuits. It is widely used in power substations, communication base stations, rail transit, and industrial control. The insulation status of its positive and negative poles to ground directly affects equipment operation safety and personal safety. When a single-pole grounding or a decrease in insulation resistance occurs in a DC system, if not addressed promptly, it may develop into a two-pole grounding fault, causing protection devices to malfunction, fail to operate, or even damage the equipment. Therefore, real-time and accurate detection of the insulation status of a DC system is of significant engineering importance.

[0003] Current insulation testing methods mostly rely on calculations based on the original voltage to ground. However, DC system charger ripple, heavy load switching, and global common-mode voltage fluctuations caused by external electromagnetic pulses can directly lead to deviations in the test results, resulting in a high false alarm rate. Furthermore, most methods use fixed thresholds as the criteria for judging insulation faults, which cannot adapt to changes in system operating conditions and reduce the accuracy of the test results. When a decrease in insulation is detected on the bus side, existing technologies lack effective branch-level location methods, often requiring manual circuit-by-circuit switching for troubleshooting, which is inefficient and may cause load power outages. Summary of the Invention

[0004] The purpose of this invention is to provide a method for detecting insulation in a DC system, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a DC system insulation detection method, comprising the following steps: The detection nodes are arranged so that detection nodes are installed between the positive and negative busbars of the DC system and the ground to measure the positive voltage to ground and the negative voltage to ground. On the unloaded output branch with known good insulation, an additional reference detection channel is installed. The reference detection channel measures the standard positive voltage to ground and the standard negative voltage to ground at the same time. Voltage correction: Under normal operating conditions of the DC system, the positive-to-ground voltage and the negative-to-ground voltage simultaneously acquired by the detection node are differentially corrected using the standard positive-to-ground voltage and the standard negative-to-ground voltage acquired by the reference detection channel to obtain the corrected voltage data. Feature benchmark value analysis: Based on the corrected voltage data, calculate the corrected positive-to-ground voltage ratio and negative-to-ground voltage ratio of each detection node as the feature vector of that node. Take the average value of the feature vectors of multiple consecutive sampling points under the system health state to obtain the feature benchmark value of each node. Insulation degradation is determined by analyzing the fluctuation intensity of the corrected voltage data. During the healthy operation phase of the system, the fault determination threshold is dynamically calculated based on the long-term average fluctuation intensity and the standard deviation of the short-term fluctuation intensity of the characteristic proportion of the node. When the absolute value of the deviation of the corrected positive-to-ground voltage proportion from its initial characteristic reference value exceeds the fault determination threshold for multiple consecutive sampling periods, the system is determined to have insulation degradation. At this time, the fault polarity is further determined based on the corrected positive-to-ground voltage proportion and the initial characteristic reference value, including positive insulation degradation and negative insulation degradation. Fault branch analysis: After determining the fault polarity, calculate the synchronization difference between each branch node and the bus node. Branches with a synchronization difference exceeding three times the benchmark value are marked as suspected fault branches. Then, compare the synchronization difference mismatch on both sides of the branch to obtain the synchronization difference score between the positive and negative poles, and set the corresponding threshold to determine whether the branch is a fault branch.

[0006] Optionally, in the voltage correction step, the positive electrode-to-ground voltage collected at the same time by each detection node is subtracted from the standard positive electrode-to-ground voltage collected by the reference detection channel to obtain the positive electrode-to-ground voltage DU of the i-th detection node after time t correction. i,pg (t) is calculated by subtracting the standard negative electrode-to-ground voltage collected by the reference detection channel from the negative electrode-to-ground voltage collected at the same time by the negative electrode-to-ground voltage collected at each detection node at the same time, to obtain the negative electrode-to-ground voltage DU of the i-th detection node after time correction by t. i,ng (t).

[0007] Optionally, in the characteristic reference value analysis step, the ratio of the corrected positive electrode voltage to ground to the corrected total voltage of each detection node is first calculated to obtain the corrected positive electrode voltage ratio P of the i-th detection node at time t. i,pg (t), the corrected total voltage is the sum of the corrected positive-to-ground voltage and the corrected negative-to-ground voltage; The corrected positive voltage percentage P at time t of the i-th detection node is obtained. i,pg Since the total voltage proportion is 1, the negative electrode voltage proportion at time t of the i-th detection node after correction is 1 - P. i,pg (t), and then take the average of the positive voltage percentage of the system under healthy operating conditions for several consecutive sampling periods to obtain the positive characteristic reference value P of the i-th detection node under healthy conditions. i,pg,avg The negative polarity baseline value of the i-th detection node in a healthy state is 1 - P. i,pg,avg .

[0008] Optionally, in the insulation degradation determination step, a sliding time window is set with time t as the center, the maximum and minimum values ​​of the corrected positive electrode to ground voltage within the window are extracted and the difference between the two is calculated as the positive electrode fluctuation amplitude, and the difference between the maximum and minimum values ​​of the corrected negative electrode to ground voltage is calculated as the negative electrode fluctuation amplitude. The average value of the positive electrode fluctuation amplitude and the negative electrode fluctuation amplitude is taken to obtain the fluctuation intensity S(t) at time t. After obtaining the fluctuation intensity S(t) at time t, the fault judgment threshold F(t) at time t is obtained by combining the inherent fluctuation level and recent short-term fluctuation amplitude under the system health state through static benchmark weighting and dynamic fluctuation standard deviation. Then calculate the percentage of positive voltage P at time t of the i-th detection node after correction. i,pg (t) and the positive characteristic benchmark value P of the i-th detection node in a healthy state. i,pg,avg absolute value of deviation | P i,pg (t)-P i,pg,avg | When the fault determination threshold F(t) of time t is exceeded for at least three consecutive sampling periods, the system insulation is determined to be degraded.

[0009] Optionally, after determining that the system insulation has deteriorated, the fault polarity is further determined, as follows: When P i,pg (t)>P i,pg,avg When F(t) is +F, it is determined that the insulation of the negative electrode has decreased; When P i,pg (t)<P i,pg,avg When -F(t), it is determined that the positive electrode insulation has decreased.

[0010] Optionally, in the fault branch analysis step, the corrected positive voltage ratio data of the branch node and the bus node are first obtained, the change in positive voltage ratio of the two nodes at adjacent sampling times is calculated, and they are normalized with the positive characteristic reference value under healthy conditions. Then, the absolute value of the difference after normalization is taken to obtain the absolute value of the deviation. Simultaneously, based on the ratio of the absolute difference in positive voltage percentage between the branch node and the bus node at the same moment to the preset safe difference in voltage percentage, an adaptive weighting term is obtained; combining the absolute deviation term and the adaptive weighting term, the synchronization difference degree D between the branch node j and the bus node is obtained. j,bus ; The value range of detection node i is the total number of voltage detection nodes arranged in the DC system, including bus nodes and branch nodes j; The synchronization difference D between branch node j and bus node is obtained. j,bus Then, the synchronization difference baseline value was set to D. base ; When the synchronization difference between branch node j and bus node Dj,bus >3×D base The branch was initially identified as a suspected faulty branch. When the synchronization difference between branch node j and bus node D j,bus ≤3×D base It was determined to be a normal branch.

[0011] Optionally, after determining a suspected faulty branch, the synchronization difference between branch node j and the bus node is calculated based on the positive and negative characteristics, respectively. The absolute value of the difference between the two is then taken to obtain the positive and negative synchronization difference score DF, and the health calibration value is set to HS. When the positive and negative pole synchronization difference score DF > 5 × health calibration value HS, the branch is determined to be the target fault branch; When the positive and negative electrode synchronization difference score DF≤5×health calibration valueHS, it is judged as a false positive abnormality, and the management personnel are notified to conduct an inspection.

[0012] Optionally, in the step of arranging the detection nodes, the reference detection channel adopts the same hardware sampling circuit structure as the detection node and is installed on an unloaded spare branch that has been pre-confirmed to have good insulation and no external load.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: I. This invention sets up a reference detection channel to synchronously and differentially correct the original ground voltage of each detection node with the reference channel voltage. This can eliminate the influence of global common-mode disturbances such as DC system load fluctuations, switching operations, or external electromagnetic pulses on the detection results, and extract feature signals that are only related to the insulation state of the node itself. Even if the system bus voltage fluctuates significantly for a short period of time, effective features directly related to the insulation state can be extracted, reducing the probability of misjudgment under complex operating conditions.

[0014] Second, the fault judgment threshold of this invention is calculated in real time based on the standard deviation of the long-term average fluctuation intensity and the recent short-term fluctuation intensity under the healthy state of the system. It can automatically adjust with the changes in the system operating state. When the transient disturbance of the system increases, the threshold increases accordingly; when the system is running smoothly, the threshold automatically decreases, which improves the sensitivity of early insulation degradation detection, can adapt to the natural aging characteristics of insulation, avoids the problem of increased false alarm rate of traditional fixed threshold operation over a long period of time, and improves the accuracy of detection results.

[0015] Third, after determining that the insulation of the system has deteriorated, this invention first calculates the synchronization difference between each branch node and the bus node, and marks branches that exceed three times the normal reference value as suspected fault branches; then compares the synchronization difference mismatch on both sides of the suspected branch to obtain the positive and negative pole synchronization difference score DF. When the positive and negative pole synchronization difference score exceeds five times the health calibration value, it is confirmed as the target fault branch. There is no need to disconnect power to each branch. The branch-level location can be automatically completed while the system is energized, which improves the troubleshooting efficiency and avoids the risk of load power failure. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the insulation degradation determination process of the present invention. Figure 3 This is a flowchart of the fault branch analysis of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figures 1 to 3 This embodiment provides a method for detecting insulation in a DC system, including the following steps: The detection node arrangement involves installing voltage detection nodes between the positive and negative buses of the DC system and the ground, thus obtaining bus node information, including positive and negative bus nodes, and measuring the positive-to-ground voltage U at each node. pg (t), measuring the voltage U of the negative terminal to ground. ng (t), on a known well-insulated unloaded output branch, an additional reference detection channel with the same structure as the busbar node is installed. This reference detection channel measures the standard positive terminal to ground voltage U at the same time. ref,pg (t) The standard negative terminal voltage to ground is U ref,ng (t); Voltage correction involves collecting voltage data at each detection node under normal DC system operating conditions. Differential correction is then applied to each detection node using voltage data simultaneously collected via a reference channel, yielding corrected positive-to-ground and negative-to-ground voltages. This process eliminates interference from common-mode voltage fluctuations within the system. The calculation process for the corrected positive-to-ground voltage is as follows:

[0019] DUi,pg (t) is the positive terminal to ground voltage of the i-th detection node after time t correction, the effective voltage signal after eliminating common-mode interference of the system, reflecting the insulation state deviation of the node itself, and the value of i is the total number of voltage detection nodes arranged in the DC system. U i,pg (t) represents the positive terminal voltage to ground of the i-th detection node at time t, which is obtained through real-time measurement and includes the raw sampled data of system common-mode fluctuations and the insulation characteristics of the node itself. U ref,pg (t) represents the positive terminal voltage to ground of the reference detection channel at time t, reflecting the global common-mode voltage fluctuation of the DC system, and is used to offset the influence of system-level voltage disturbances; The calculation process for the corrected negative pole-to-ground voltage is as follows:

[0020] DU i,ng (t) represents the negative electrode voltage to ground after time correction at the i-th detection node (t); U i,ng (t) represents the negative electrode voltage to ground at time t of the i-th detection node, which is obtained through real-time measurement; U ref,ng (t) represents the negative terminal voltage to ground of the reference detection channel at time t; By installing a reference detection channel on a known well-insulated, unloaded output branch, the positive-to-ground insulation resistance of this branch is much higher than the system's insulation requirement threshold. Therefore, the positive-to-ground voltage of this branch is almost unaffected by its own insulation condition; its variation originates only from the common-mode background fluctuations of the entire DC system. The positive and negative voltages to ground measured by the reference detection channel can be approximated as the system's background fluctuations at the current moment. Thus, by using the real-time measured positive-to-ground voltage U at time t of the i-th detection node... i,pg (t) and the negative electrode-to-ground voltage U of the i-th detection node at time t i,ng (t) Subtracting the corresponding background fluctuation amount can eliminate the influence of global common-mode disturbances such as DC system load fluctuations, switching operations or external electromagnetic pulses on the detection results. Even if the system bus voltage fluctuates significantly for a short time, effective features directly related to the insulation state can be extracted, reducing the probability of misjudgment under complex working conditions.

[0021] After obtaining the corrected positive-to-ground voltage, the proportion of positive-to-ground voltage and the proportion of negative-to-ground voltage for each detection node are calculated and used as the initial feature vector for that node. The process is as follows:

[0022] P i,pg(t) represents the corrected positive voltage ratio at time t of the i-th detection node, a normalized characteristic value that eliminates the influence of system voltage fluctuation amplitude and reflects the relative proportion of positive electrode insulation to ground. U i,pg (t) represents the positive terminal voltage to ground of the i-th detection node at time t; U i,ng (t) represents the negative terminal voltage to ground at time t of the i-th detection node; DU i,pg (t)+DU i,ng (t) represents the corrected total system voltage, which is the sum of the positive and negative voltages; θ is a very small positive number, with a value of 10. -9 This is used to prevent numerical overflow caused by a denominator of zero, and does not affect the calculation precision of significant digits during normal operation; The positive electrode-to-ground voltage DU of the i-th detection node after time correction is calculated. i,pg (t) and the corrected total system voltage DU i,pg (t)+DU i,ng The ratio (t) converts the absolute value of the voltage into a relative proportion, eliminating the differences caused by different dimensions and improving the adaptability to systems with different voltage levels. After correction, the positive electrode voltage proportion P at time t of the i-th detection node is obtained. i,pg After (t), since the total proportion is 1, the proportion of the negative electrode voltage at time t of the i-th detection node after correction is 1 - P. i,pg (t); After analyzing the characteristic reference values ​​and obtaining the corrected positive-to-ground voltage ratio and negative-to-ground voltage ratio for each detection node, the average value of the feature vectors of multiple consecutive sampling points under healthy system conditions is taken to calculate the initial characteristic reference for each detection node. The initial characteristic reference includes the positive characteristic reference value and the negative characteristic reference value. The process is as follows:

[0023] P i,pg,avg Let be the positive characteristic baseline value of the i-th detection node in a healthy state; N hea The total number of sampling points during the health status collection period; P during the health status collection period i,pg (t) summation, used to calculate the statistical average and eliminate the interference of short-term fluctuations; T hea This is a set of healthy periods, usually selected during the initial stage of system commissioning or periods when no insulation defects have been confirmed. Obtain the positive characteristic baseline value P of the i-th detection node in a healthy state. i,pg,avgSubsequently, since the insulation levels of the positive and negative terminals of the DC system to ground are basically symmetrical under normal healthy conditions, and the sum of the characteristic reference values ​​of the negative and positive terminals is 1, the characteristic reference value of the negative terminal of the i-th detection node under healthy conditions is 1 - P. i,pg,avg ; By calculating the initial baseline based on the average value during the healthy operation phase, it can adapt to the normal characteristic drift of DC system during long-term operation, such as the natural aging of insulation materials and the slow change of distributed capacitance. The baseline value will naturally iterate with the normal aging of the system throughout its entire life cycle, avoiding the problem of increased false alarm rate over time caused by the traditional fixed threshold method.

[0024] Insulation degradation is determined by analyzing the fluctuation intensity of corrected voltage data, combining the inherent fluctuation level under healthy system conditions and recent short-term fluctuation amplitude, and calculating the fault determination threshold through static benchmark weighting and dynamic fluctuation standard deviation. Insulation degradation is further determined by real-time monitoring of the positive correction characteristic ratio of bus nodes. When this value exceeds the dynamic threshold range for three consecutive sampling periods, insulation degradation is determined to have occurred. If the positive deviation of the characteristic ratio exceeds the threshold, it is determined to be negative insulation degradation; if the negative deviation of the characteristic ratio exceeds the threshold, it is determined to be positive insulation degradation. The process is as follows: First, calculate the fluctuation intensity. With the current time t as the center, set a sliding time window. The window length is set to 5 sampling periods by default. Extract the maximum and minimum values ​​of the corrected positive electrode voltage to ground and the maximum and minimum values ​​of the corrected negative electrode voltage to ground within the window. Then, calculate the difference between the maximum and minimum values ​​of the corrected positive and negative electrode voltages within the window to obtain the positive electrode fluctuation amplitude and the negative electrode fluctuation amplitude. Take the average of the two to obtain the fluctuation intensity S(t) at time t. The fault determination threshold is then calculated, and the specific process is as follows:

[0025] In the above formula, F(t) is the fault judgment threshold at time t, which serves as the critical value for insulation fault judgment. Exceeding this threshold indicates an abnormal shift in the judgment characteristic. By combining static reference fluctuations and dynamic fluctuation increments, dynamic threshold adjustment is achieved, balancing detection sensitivity and anti-interference capability, and improving the accuracy of insulation fault judgment. S avg is the statistical average value of the fluctuation intensity under healthy conditions; represents the long-term average value of the fluctuation intensity S(t) at time t during the healthy operation phase of the system, representing the normal background noise level of the system; β1 is a static weighting coefficient, ranging from 2 to 4, with a default value of 3. It is used to scale the average fluctuation intensity under healthy conditions to form a basic threshold. The smaller the coefficient, the lower the basic threshold and the more sensitive the detection, but it is easily affected by normal load fluctuations or measurement noise interference, resulting in false alarms. The larger the coefficient, the stronger the anti-interference ability, but it may miss early insulation degradation. Therefore, the default compromise value of 3 is taken. β2 is a dynamic weighting coefficient, with a default value of 1.5, used to scale the standard deviation of recent fluctuation intensity. In DC systems, transient processes such as the switching of large equipment and the conversion between equalization and float charging of chargers can cause the fluctuation intensity S(t) at time t to increase in a short period of time, but its standard deviation is usually the statistical average of the fluctuation intensity S under healthy conditions. avg The dynamic weighting coefficient β2 is 0.5 to 2 times that of the threshold. If the dynamic weighting coefficient β2 is too large, the threshold will rise too much, which will easily lead to a decrease in detection sensitivity. If β is too small, it will not be able to effectively deal with severe transient interference and will be prone to false alarms. In practical applications, it can be adjusted according to actual needs. Therefore, the dynamic weighting coefficient β2 is set to 2 for scenarios with more transient interference and 1 for scenarios with high requirements for fault response speed. M is the number of sampling points in the sliding time window, which determines the response speed of the dynamic threshold. The larger the number, the smoother the threshold adjustment, but the slower the response speed. S(t-mΔ) represents the time fluctuation intensity over time t-mΔ, where Δ is the time interval between two adjacent moments used to calculate the standard deviation; S(t-mΔ)-S avg This represents the difference between the fluctuation intensity at a single sampling point and the healthy baseline, reflecting the degree of abnormality of the fluctuation at that moment. The fluctuation intensity S(t) at time t is relative to the statistical average St of the fluctuation intensity under healthy conditions. avg When significant changes occur, such as load switching or increased interference, the fault judgment threshold will increase accordingly to avoid misjudging insulation degradation due to changes in normal operating conditions. Conversely, if the system is more stable recently, the fault judgment threshold will decrease accordingly to improve the detection sensitivity of early insulation degradation. Furthermore, by using the square root quantization of the dispersion of fluctuations within the sliding window, the severity of the current system disturbance can be reflected.

[0026] After obtaining the fault determination threshold F(t) at time t, and combining it with the corrected positive voltage ratio P of the i-th detection node at time t, i,pg (t) and the positive characteristic benchmark value P of the i-th detection node in a healthy state. i,pg,avg The fault diagnosis process is as follows: First, when the corrected positive voltage percentage P at time t of the i-th detection node is... i,pg (t) and the positive characteristic benchmark value P of the i-th detection node in a healthy state. i,pg,avg The absolute value of the deviation is |P i,pg (t)-P i,pg,avg | If the fault determination threshold is exceeded for at least three consecutive sampling periods, the system insulation is determined to be degraded, and the fault polarity is then determined. When P i,pg (t)>P i,pg,avgWhen F(t) is greater than the current positive electrode characteristic ratio, it means that the difference exceeds the fault judgment threshold. It is judged that the negative electrode insulation has decreased. When the negative electrode insulation resistance decreases, the negative electrode voltage to ground decreases relatively, and the positive electrode voltage ratio increases simultaneously. When P i,pg (t)<P i,pg,avg When F(t) is reached, it is determined that the positive electrode insulation has decreased, indicating that the current positive electrode characteristic ratio is lower than the health benchmark and the difference exceeds the fault judgment threshold. When the positive electrode insulation resistance decreases, the positive electrode voltage to ground decreases relatively, and the negative electrode voltage to ground increases relatively. Therefore, the negative electrode voltage ratio increases simultaneously.

[0027] Fault branch analysis involves calculating the synchronization difference between each output branch inlet node and the bus reference node after determining insulation degradation on a specific electrode side. Branches exceeding three times the normal baseline value are marked as suspected faulty branches. The synchronization difference mismatch between the positive and negative poles of the branch is then compared to obtain a synchronization difference score. A fault threshold is then set to determine whether the branch is indeed a faulty branch. The specific process is as follows: The value range of detection node i is the total number of voltage detection nodes arranged in the DC system. Therefore, bus node and branch node j are also included. Thus, the voltage data and characteristic reference values ​​of bus node and branch node j can be obtained through the above steps, and will not be elaborated here. First, obtain the corrected positive voltage percentage data for each output branch node j on the electrode side where insulation degradation occurs, as well as the positive characteristic reference value of that branch node j.

[0028] In the above formula, D j,bus This represents the synchronization difference between branch node j and the bus node. The larger the value, the worse the synchronization between the branch characteristics and the bus characteristics, and the higher the probability of insulation failure. M is the number of sampling points in the sliding time window, and tm is the kth sampling time within the time window; ΔP j,pg (tm) represents the first-order difference of the positive voltage ratio of branch node j at time tm after correction. It refers to the change in the positive voltage ratio of branch node j between two adjacent sampling times, reflecting the instantaneous fluctuation amplitude and trend of the insulation characteristics of the branch within a sampling period. P j,pg,avg The positive characteristic baseline value of branch node j under healthy conditions; ΔP bus,pg (tm) represents the first-order difference of the positive voltage ratio of the bus node at time tm, which indicates the change in the positive voltage ratio of the bus node between two adjacent sampling times. It represents the instantaneous change trend of the global characteristics of the system and serves as a synchronization reference for all branches. P bus,pg,avg This is the baseline value for the positive electrode characteristics under healthy bus node conditions; P j,pg (tm) represents the percentage of the positive voltage at branch node j during time tm after correction; P bus,pg (tm) represents the percentage of positive voltage at the bus node during time tm; For the absolute value of the deviation, the difference between the relative changes of branch node j and bus node is calculated and the absolute value is taken to quantify the degree of difference between the branch fluctuation and the global fluctuation at a single sampling time, and to eliminate the influence of the deviation direction. For branches with normal insulation, their characteristic changes completely follow the global fluctuation of the system, and the difference will be stable near 0; while for branches with insulation degradation, their characteristic changes will be independent of the global fluctuation, and the difference will increase significantly. Furthermore, by summing and averaging the characteristic fluctuation differences between branch node j and bus node at all sampling moments within the sliding time window, the influence of characteristic jumps caused by random noise and instantaneous electromagnetic interference at single moments is eliminated, avoiding misjudgment of synchronization caused by a single abnormal sampling point; at the same time, it fully covers the continuous change process of fault characteristics, improving the accuracy of the results. For adaptive weighting terms, the influence of the deviation at that moment is dynamically adjusted based on the characteristic differences between branch node j and the bus node, where |P| j,pg (tm)-P bus,pg (tm) | Reflects the absolute difference between the positive voltage ratio of branch node j and bus node at the current moment. When the absolute difference is too large, the weight decays rapidly, filtering out the influence of instantaneous abnormal sampling points or non-steady-state interference, and improving the accuracy of the results. P ran The voltage percentage safety difference ranges from 0.2 to 0.5. Under normal DC system operation and without insulation faults, the corrected characteristic percentage is theoretically stable. However, due to factors such as load variations, measurement noise, and residual errors in the reference channel, the actual value will fluctuate within this range. Based on field operation data and simulation analysis, the amplitude of this normal fluctuation usually does not exceed 0.05. Under healthy conditions, the positive voltage percentage difference between any node and the bus node is less than 0.1 in most cases. The voltage percentage safety value P... ran The minimum value of 0.2 is already greater than the upper limit of normal fluctuations, ensuring that in a healthy system, all normal minor differences will not lead to a significant reduction in weight. The upper limit is set at 0.5 because the voltage ratio itself ranges from 0 to 1. A value exceeding 0.5 indicates that there is a significant difference between the two nodes, such as one being close to the positive bus and the other being close to the negative bus. In this case, even if the insulation degradation occurs on this branch, its relative change may be masked by the large static difference. In this case, the influence of this item is directly reduced to avoid misjudgment. The synchronization difference D between branch node j and bus node is obtained. j,bus Subsequently, the synchronization difference benchmark value D is set based on the average synchronization difference between all normal branch bus nodes. base ; When the synchronization difference between branch node j and bus node D j,bus >3×D base The branch was identified as a suspected faulty branch and proceeded to further verification. When the synchronization difference between branch node j and bus node D j,bus ≤3×D base If a branch is identified as normal, but the system has already determined the fault polarity and detected a decrease in insulation on a certain side, but the specific branch has not been located, it may be due to a decrease in the insulation of the busbar itself or a simultaneous decrease in the insulation of all branches. In this case, the management personnel need to be notified to conduct an inspection. Once a branch is identified as a suspected faulty branch, the synchronization difference D between branch node j and the bus node is calculated based on both positive and negative characteristics. j,bus Then, the absolute value of the difference between the two is taken to obtain the positive and negative pole synchronization difference score DF, and the health calibration value is set as HS. The health calibration value is the benchmark value obtained in advance under the system health state. If there is a unilateral insulation degradation in the branch, such as the positive pole insulation to ground deterioration, the synchronization difference degree D between branch node j and bus node is calculated based on the positive pole characteristics. j,bus It will increase, while the negative characteristic calculation shows the synchronization difference D between branch node j and the bus node. j,bus The circuit is still close to a healthy level. At this point, the positive and negative pole synchronization difference score DF will be much larger than the healthy calibration value HS. If there is no fault in the branch, the difference between the two sides will be basically equal, and the positive and negative pole synchronization difference score DF will be extremely small. Therefore, the following setting is made: When the positive and negative pole synchronization difference score DF > 5 × health calibration value HS, the branch is determined to be the target fault branch. At this time, the fault branch number is output. Since there are a lot of normal disturbances in the industrial field that can cause asymmetric characteristics, such as the slight difference in distributed capacitance when the branch load is switched, the inherent measurement error of the sampling circuit, and the slight drift of insulation parameters caused by changes in ambient temperature and humidity, the positive and negative pole synchronization difference score DF caused by non-fault disturbances is usually 1 to 3 times the health average. Setting 5 times the health calibration value HS as the fault branch judgment standard can filter out normal fluctuations and avoid misjudgment. When the positive and negative pole synchronization difference score DF≤5×health calibration valueHS, it is judged as a false positive anomaly, not a real insulation fault. There is short-term electromagnetic interference near the branch, such as inverter start-up and shutdown, switch operation, high-power equipment start-up and shutdown, etc., or the sampling circuit of the branch experiences instantaneous data jump, communication packet loss or noise interference, which leads to an abnormally high synchronization difference within a single window. However, the anomaly has no polarity bias on the positive and negative pole sampling. At this time, the management personnel should be notified immediately for inspection.

[0029] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting insulation in a DC system, characterized in that, Includes the following steps: Step S1: Detection node arrangement. Detection nodes are installed between the positive and negative busbars of the DC system and the ground to measure the positive-to-ground voltage and the negative-to-ground voltage. On the known well-insulated no-load output branch, an additional reference detection channel is installed. The reference detection channel measures the standard positive-to-ground voltage and the standard negative-to-ground voltage at the same time. Step S2: Voltage correction. Under normal operating conditions of the DC system, the positive-to-ground voltage and negative-to-ground voltage simultaneously acquired by the detection node are differentially corrected by the standard positive-to-ground voltage and standard negative-to-ground voltage acquired by the reference detection channel to obtain the corrected voltage data. Step S3: Feature reference value analysis. Based on the corrected voltage data, calculate the corrected positive-to-ground voltage ratio and negative-to-ground voltage ratio of each detection node as the feature vector of that node. Take the average value of the feature vectors of multiple consecutive sampling points under the healthy state of the system to obtain the feature reference value of each node. Step S4: Insulation degradation determination. Based on the analysis of fluctuation intensity of the corrected voltage data, during the healthy operation phase of the system, the fault determination threshold is dynamically calculated according to the long-term average fluctuation intensity and the standard deviation of the short-term fluctuation intensity of the characteristic proportion of the node. When the absolute value of the deviation of the corrected positive pole to ground voltage proportion from its initial characteristic reference value exceeds the fault determination threshold for multiple consecutive sampling periods, the insulation degradation of the system is determined. At this time, the fault polarity is further determined based on the corrected positive pole to ground voltage proportion and the initial characteristic reference value, including positive pole insulation degradation and negative pole insulation degradation. Step S5: Fault branch analysis. After determining the fault polarity, calculate the synchronization difference between each branch node and the bus node. Mark branches with a synchronization difference exceeding three times the benchmark value as suspected fault branches. Then compare the synchronization difference mismatch on both sides of the branch to obtain the synchronization difference score between the positive and negative poles, and set the corresponding threshold to determine whether the branch is a fault branch.

2. The DC system insulation detection method according to claim 1, characterized in that: In the voltage correction step, the positive electrode-to-ground voltage collected at the same time by each detection node is subtracted from the standard positive electrode-to-ground voltage collected by the reference detection channel to obtain the positive electrode-to-ground voltage DU of the i-th detection node after time t correction. i,pg (t) is calculated by subtracting the standard negative electrode-to-ground voltage collected by the reference detection channel from the negative electrode-to-ground voltage collected at the same time by the negative electrode-to-ground voltage collected at each detection node at the same time, to obtain the negative electrode-to-ground voltage DU of the i-th detection node after time correction by t. i,ng (t).

3. The DC system insulation detection method according to claim 2, characterized in that: In the characteristic reference value analysis step, the ratio of the corrected positive electrode voltage to ground to the corrected total voltage of each detection node is first calculated to obtain the positive electrode voltage ratio P of the i-th detection node at time t after correction. i,pg (t), the corrected total voltage is the sum of the corrected positive-to-ground voltage and the corrected negative-to-ground voltage; The corrected positive voltage percentage P at time t of the i-th detection node is obtained. i,pg Since the total voltage proportion is 1, the negative electrode voltage proportion at time t of the i-th detection node after correction is 1 - P. i,pg (t), and then take the average of the positive voltage percentage of the system under healthy operating conditions for several consecutive sampling periods to obtain the positive characteristic reference value P of the i-th detection node under healthy conditions. i,pg,avg The negative polarity baseline value of the i-th detection node in a healthy state is 1 - P. i,pg,avg .

4. The DC system insulation detection method according to claim 3, characterized in that: In the insulation degradation determination step, a sliding time window is set with time t as the center. The maximum and minimum values ​​of the corrected positive electrode voltage to ground within the window are extracted and the difference between the two is calculated as the positive electrode fluctuation amplitude. At the same time, the difference between the maximum and minimum values ​​of the corrected negative electrode voltage to ground is calculated as the negative electrode fluctuation amplitude. The average value of the positive electrode fluctuation amplitude and the negative electrode fluctuation amplitude is taken to obtain the fluctuation intensity S(t) at time t. After obtaining the fluctuation intensity S(t) at time t, the fault judgment threshold F(t) at time t is obtained by combining the inherent fluctuation level and recent short-term fluctuation amplitude under the system health state through static benchmark weighting and dynamic fluctuation standard deviation. Then calculate the percentage of positive voltage P at time t of the i-th detection node after correction. i,pg (t) and the positive characteristic benchmark value P of the i-th detection node in a healthy state. i,pg,avg absolute value of deviation | P i,pg (t)-P i,pg,avg | When the fault determination threshold F(t) of time t is exceeded for at least three consecutive sampling periods, the system insulation is determined to be degraded.

5. The DC system insulation detection method according to claim 4, characterized in that: After determining that the system insulation has deteriorated, the next step is to determine the fault polarity, as follows: When P i,pg (t)>P i,pg,avg When F(t) is +F, it is determined that the insulation of the negative electrode has decreased; When P i,pg (t)<P i,pg,avg When -F(t), it is determined that the positive electrode insulation has decreased.

6. The DC system insulation detection method according to claim 5, characterized in that: In the fault branch analysis step, the corrected positive voltage ratio data of the branch node and the bus node are first obtained. The change in the positive voltage ratio of the two nodes at adjacent sampling times is calculated respectively. The positive voltage ratio is then normalized with the positive characteristic reference value under healthy conditions. The absolute value of the difference after normalization is then taken to obtain the absolute value of the deviation. Simultaneously, based on the ratio of the absolute difference in positive voltage percentage between the branch node and the bus node at the same moment to the preset safe difference in voltage percentage, an adaptive weighting term is obtained; combining the absolute deviation term and the adaptive weighting term, the synchronization difference degree D between the branch node j and the bus node is obtained. j,bus ; The value range of detection node i is the total number of voltage detection nodes arranged in the DC system, including bus nodes and branch nodes j; The synchronization difference D between branch node j and bus node is obtained. j,bus Then, the synchronization difference baseline value was set to D. base ; When the synchronization difference between branch node j and bus node D j,bus >3×D base The branch was initially identified as a suspected faulty branch. When the synchronization difference between branch node j and bus node D j,bus ≤3×D base It was determined to be a normal branch.

7. The DC system insulation detection method according to claim 6, characterized in that: After determining a suspected faulty branch, the synchronization difference between branch node j and the bus node is calculated based on the positive and negative characteristics, respectively. The absolute value of the difference between the two is then taken to obtain the positive and negative synchronization difference score DF, and the health calibration value is set to HS. When the positive and negative pole synchronization difference score DF > 5 × health calibration value HS, the branch is determined to be the target fault branch; When the positive and negative electrode synchronization difference score DF≤5×health calibration valueHS, it is judged as a false positive abnormality, and the management personnel are notified to conduct an inspection.

8. The DC system insulation detection method according to claim 1, characterized in that: In the step of setting up the detection nodes, the reference detection channel adopts the same hardware sampling circuit structure as the detection nodes and is installed on an unloaded spare branch that has been confirmed to have good insulation and no external load.

Images