Method and device for detecting leakage current of direct-current street lamp based on wavelet transformation

Abstract

The invention relates to the field of current detection, in particular to a method and a device for detecting leakage current of a direct-current street lamp based on wavelet transformation. Comprising the following steps: initializing wiring, carrying out load-environment adaptation, carrying out multi-channel sampling, carrying out wavelet decomposition, reconstructing a transient component and a steady component, calculating an RMS value, a three-phase equivalent leakage current effective value, a WPT total energy entropy, a node energy ratio, a ripple coefficient and a transient pulse width, calculating a dynamic threshold, and inputting five-dimensional leakage current characteristics into an SVM classification model to identify a fault type and locate the fault type. And grading alarm. The method has the advantages that the load type is automatically recognized, the sampling range is dynamically adjusted, and sampling saturation or insufficient sensitivity caused by load switching is avoided; the leakage current detection accuracy is improved; the threshold is corrected in combination with a multi-dimensional correction coefficient, and the problem that the false alarm rate of a fixed threshold is high is solved; the fault classification accuracy is improved, the fault position is accurately positioned, and the operation and maintenance efficiency is improved; the device is convenient to use, visible in data and high in practicability.

Description

Technical Field

[0001] This invention relates to the field of current detection, and in particular to a method and apparatus for detecting leakage current of DC streetlights based on wavelet transform. Background Technology

[0002] With the large-scale adoption of LED DC streetlights in urban roads, tunnels, and landscape lighting, the complexity of their outdoor operating environment (rain, fog, temperature fluctuations, day-night load switching) and leakage current problems caused by DC bus insulation faults are becoming increasingly prominent. Leakage current in DC streetlight systems not only affects the reliable operation of equipment but may also cause fires due to continuous heating from resistive leakage or generate electric sparks due to transient pulse leakage, threatening personal safety. Existing leakage current detection technologies have significant shortcomings and are difficult to meet the operation and maintenance needs of outdoor streetlights.

[0003] Traditional detection devices rely solely on fixed thresholds to determine the effective value of leakage current, failing to distinguish between "transient capacitive leakage current due to cable dampness," "LED driver ripple interference," and "persistent resistive leakage current due to insulation damage." This results in a false alarm rate exceeding 30% in rainy or foggy weather. While some literature mentions transient current feature extraction, it is not adapted to the transient interference characteristics of outdoor streetlights, as it is only applicable to medium-voltage lines in nuclear power plants. In contrast, previous literature, although using wavelet analysis to process fault signals, focused on short-circuit faults in AC transmission lines and did not optimize for transient / steady-state separation of leakage current in DC streetlights.

[0004] Existing technologies are mostly designed for specific closed scenarios and environmental parameters. The nuclear power plants mentioned in existing literature only consider the correction of thresholds by humidity, and coal mines focus on gas and dust concentration. However, DC streetlights operating outdoors need to cope with specific environmental factors such as "rain and fog humidity, cable temperature, and day and night light load patterns". Existing devices do not include these parameters, resulting in a disconnect between threshold judgment and actual operating conditions.

[0005] Existing patents can only output "leakage alarm" signals and cannot achieve "fault type identification + fault location". Although existing literature mentions load current and threshold matching, it does not involve fault classification. In contrast, the SVM classification model in existing literature is only for AC line short circuit faults and cannot be transferred to DC street light leakage current (resistive, capacitive, pulse type) classification. Maintenance personnel spend more than 2 hours checking each 1km of line for each light, which is extremely inefficient.

[0006] DC street light loads are diverse, including LED lights, intelligent controllers, emergency power supplies, etc. The leakage current characteristics of different loads are significantly different. However, existing devices use a fixed sampling range, which results in insufficient sensitivity to low ripple loads or sampling saturation to high pulse loads.

[0007] Therefore, it is an urgent problem to solve how to combine specific load types, environmental influences, fault type identification, and fault location to provide a leakage current detection solution that is suitable for DC streetlights in outdoor scenarios. Summary of the Invention

[0008] To overcome the above-mentioned shortcomings, the present invention aims to provide a method and device for detecting leakage current of DC streetlights based on wavelet transform, thereby solving the problems of high false alarms, difficult operation and maintenance, and poor adaptability in DC streetlight leakage current detection.

[0009] This invention achieves the above objective through the following scheme: a method for detecting leakage current of DC streetlights based on wavelet transform, comprising the following steps:

[0010] (1) Identify the load type and collect environmental parameters, and output the load correction coefficient. Environmental correction factor And adjust the sampling range and sampling rate;

[0011] (2) Collect leakage current signals from the positive and negative terminals of the bus, the PE line, and the load terminal at the adjusted sampling rate;

[0012] (3) Perform 5-level db4 wavelet packet decomposition on the collected leakage current signal to obtain the transient component. With steady-state components ;

[0013] (4) Calculation RMS value Three-phase equivalent leakage current RMS value WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width ;

[0014] (5) Calculate the dynamic threshold The grounding performance level is output according to the dynamic judgment rules. If the grounding performance is good, step (9) is executed. If the grounding performance needs to be re-inspected or the grounding performance is unqualified, step (6) is executed. As the baseline threshold, This is a stationarity correction term. This is a three-phase imbalance correction term;

[0015] (6) The effective value of the three-phase equivalent leakage current WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The five-dimensional leakage current features are input into the trained SVM classification model to identify leakage current fault types and locate fault point information based on street light topology.

[0016] (7) Determine the steady-state components If the first threshold is exceeded, output a continuous leakage current alarm and information on the leakage current fault type and fault point.

[0017] (8) Determine transient components If the peak value exceeds the second threshold, then output an instantaneous leakage current alarm and record the pulse width and timestamp.

[0018] (9) Upload the test results and display them on the screen. The test results include the ground performance level, fault type, distance to the fault point, environmental parameters, and load parameters.

[0019] As a preferred embodiment, a method for detecting leakage current of DC streetlights based on wavelet transform further includes the step of: assessing the insulation level, specifically including: disconnecting the load, and then repeating steps 2-8 to obtain the no-load leakage current value.

[0020] Preferably, in step (3), the coefficients of the nodes in layers 1-4 are reconstructed into transient components. The 0 coefficient of the 5th layer node is reconstructed into a steady-state component. .

[0021] Preferably, in step (4), the calculation of the five-dimensional leakage current characteristics specifically includes:

[0022] (4.1) Calculate the effective value of the three-phase equivalent leakage current. Assume the three-phase leakage current is , , Then the mean Imbalance Three-phase equivalent leakage current RMS value ;

[0023] (4.2) Calculate the energy percentage of nodes Energy of the 5th layer node Total energy Node energy percentage ;

[0024] (4.3) Calculate the total energy entropy of WPT Energy probability , ;

[0025] (4.4) Calculate the ripple system ,in It is the maximum value of the ripple. It is the minimum value of the ripple;

[0026] (4.5) Transient pulse width ,in It is the pulse start, This indicates the end of the pulse.

[0027] Preferably, the stability correction term The calculation formula is:

[0028] .

[0029] Preferably, the three-phase imbalance correction term The calculation formula is:

[0030] .

[0031] Preferably, the dynamic judgment rule in step (5) is:

[0032] like and and Output: "Good grounding performance";

[0033] like or or Output: "Grounding performance needs to be re-inspected";

[0034] like or or Output: "Grounding performance is unqualified".

[0035] Preferably, the training process of the trained SVM classification model is step (6.1), which specifically includes:

[0036] (6.1.1) Extract M sets of sample data from the historical fault database. Each set of sample data contains five-dimensional leakage current characteristics and corresponding leakage current fault type labels.

[0037] (6.1.2) An improved simulated annealing algorithm is used to optimize the SVM classification model parameters. The optimized parameters include the penalty parameter C and the kernel function width g. The objective function is... ,in To predict the fault type of leakage current, Scoring based on the actual leakage current fault type. The regularization coefficient is . These are the effective values ​​of the three-phase equivalent leakage current. WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The weight coefficients, and satisfying ;

[0038] (6.1.3) Divide the sample data into training set and test set in a 7:3 ratio. Use the training set to train the SVM classification model and use the test set to verify the model accuracy. Stop training when the model classification accuracy is ≥97%.

[0039] Preferably, the fault location process is step (6.2), which specifically includes:

[0040] (6.2.1) The streetlights are numbered 1, 2...N, and the leakage current of the corresponding PE line of the light pole is as follows: , ... ;

[0041] (6.2.2) Calculate the difference in leakage current of the PE line between adjacent light poles. ;

[0042] (6.2.3) Filter out the maximum value The fault point was determined to be located at the first The lamppost and the first Between each light pole, the distance from the fault point to the first... The distance between the light poles ,in For the first The lamppost and the first The length of the cable between the light poles.

[0043] A device for detecting leakage current of DC streetlights based on wavelet transform includes: an interface component, a test circuit, a gear selection circuit, a load adaptation module, an environmental sensing module, a sampling module, a main control unit, a wavelet transform module, a threshold judgment module, a communication module, a fault classification and location module, and a human-machine interaction component. The interface component includes a first interface, a second interface, a first detection clamp, and a second detection clamp. The first interface is connected to a DC bus, the second interface is connected to a load or human body impedance simulation network, the first detection clamp is clamped on the positive bus, and the second detection clamp is clamped on the negative bus. The test circuit is connected to the interface component, the main control unit, and the sampling module. The gear selection circuit, the load adaptation module, and the sampling module are connected sequentially. The sampling module is connected to the wavelet transform module and the main control unit. The wavelet transform module is connected to the environmental sensing module and the main control unit. The main control unit is connected to the environmental sensing module, the threshold judgment module, the fault classification and location module, the communication module, and the human-machine interaction component. The threshold judgment module is connected to the fault classification and location module.

[0044] The beneficial effects of this invention are as follows: 1. Automatically identifies load type and dynamically adjusts sampling range to avoid sampling saturation or insufficient sensitivity caused by load switching; 2. Employs wavelet transform-based separation and optimization of transient / steady-state leakage current of DC streetlights, followed by signal reconstruction to improve leakage current detection accuracy; 3. Uses an optimized dynamic threshold formula, combined with multi-dimensional correction coefficients to correct the threshold, including environmental, stability, three-phase imbalance, and load adaptation factors, to solve the problem of high false alarm rate with fixed thresholds; 4. 5-dimensional features + improved simulated annealing optimized SVM classification model + topological meter-level positioning improve fault classification accuracy, accurately locate fault positions, improve operation and maintenance efficiency, and reduce labor costs; 5. The device is easy to use, provides clear and visible data, and is highly practical. It can quickly, accurately, and safely complete leakage current detection in complex field environments, and has broad application value. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the steps of the method of the present invention;

[0046] Figure 2 This is a schematic diagram of the structure of the device of the present invention;

[0047] Figure 3 This is a schematic diagram of the main interface of the display screen of the device of the present invention;

[0048] Figure 4 This is a schematic diagram of the handheld form of the device of the present invention;

[0049] Figure 5 This is a circuit diagram of the human body impedance simulation network of the device of the present invention;

[0050] In the diagram: 1. Interface component; 2. Test circuit; 3. Gear selection circuit; 4. Load adaptation module; 5. Environmental perception module; 6. Sampling module; 7. Main control unit; 8. Wavelet transform module; 9. Threshold judgment module; 10. Communication module; 11. Fault classification and location module; 12. Human-machine interaction component; 13. Dual test clamp storage compartment; 14. DC bus; 15. Load or human body impedance simulation network; 1-1. First interface; 1-2. Second interface; 1-3. First test clamp; 1-4. Second test clamp; 2-1. Power supply; 2-2. Overvoltage protection unit; 2-3. Isolated DC / DC converter; 12-1. Housing; 12-2. Display screen; 10-1. Communication interface. Detailed Implementation

[0051] The present invention will be further described below with reference to specific implementation examples, but the scope of protection of the present invention is not limited thereto:

[0052] Example 1: As Figure 1As shown, a method for detecting leakage current in DC streetlights based on wavelet transform includes the following steps:

[0053] (1) Identify the load type and collect environmental parameters, and output the load correction coefficient. Environmental correction factor And adjust the sampling range and sampling rate.

[0054] The load adaptation module 4 can automatically identify the load type, and the sampling rate of the sampling module 6 can be automatically adjusted according to the load type: when the load type is LED lights, it outputs a load correction coefficient. Furthermore, when the sampling range is selected as 1:10, the output load correction coefficient is used when the load type is an intelligent controller. Furthermore, when the sampling range is selected as 1:20, the output load correction factor is set when the load type is emergency power supply. And the sampling range is set to 1:100.

[0055] The environmental sensing module collects three key environmental parameters: when the relative humidity (RH) > 85%, it outputs an environmental correction coefficient. When the cable surface temperature T > 40℃, the output environmental correction factor is... When the illumination intensity is <50 lux, the output environmental correction factor is adjusted. .

[0056] (2) Collect leakage current signals of the positive pole, negative pole, PE line and load terminal of the bus at the adjusted sampling rate.

[0057] (3) Perform 5-level db4 wavelet packet decomposition on the collected leakage current signal to obtain the transient component. With steady-state components .

[0058] The acquired leakage current signal was subjected to 5-level db4 wavelet packet decomposition (WPT), where:

[0059] High-frequency nodes in layers 1-2 (nodes 1-3): Reconstructing instantaneous pulse leakage current (such as that generated by lightning strikes or load switching, with a frequency of 1kHz~10kHz).

[0060] 3rd-4th layer intermediate frequency nodes (nodes 4-7): Reconstruct intermittent leakage current (such as capacitive leakage current due to cable moisture, frequency 100Hz~1kHz).

[0061] Level 5 low-frequency node (node ​​0): Reconstruct steady-state leakage current (such as resistive leakage current due to insulation failure, frequency <100Hz).

[0062] The mother wavelet packet function (where n is the node index, n = 0, 1, ..., 2j-1, and j is the decomposition level) satisfies orthogonality: ( (This is the Kronecker function, which is 1 when m=n and 0 otherwise to ensure no redundancy in the decomposition); the recursive formula for the WPT decomposition of the nth node in the j-th layer is:

[0063] (Low-pass filter, approximate component) (High-pass filter, detail components)

[0064] in: Here are the low-pass filter coefficients for the db4 wavelet packet, h=[0.48296,0.83652,-0.22414,-0.12941]; For the high-pass filter coefficients, g = [-0.12941, 0.22414, 0.83652, -0.48296]; The discrete coefficients of the nth node in the j-th layer are denoted as . For time series indexing.

[0065] If transient components are detected Peak value exceeds steady-state component RMS value If the load is three times the load, a transient ground fault record is triggered, and the timestamp, environmental parameters, and load type are marked.

[0066] (4) Calculation RMS value Three-phase equivalent leakage current RMS value WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width .

[0067] Three-phase equivalent leakage current RMS value WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The specific calculation process for the five-dimensional leakage current characteristics includes:

[0068] (4.1) Calculate the three-phase equivalent leakage current: Define the three-phase DC leakage current , , (Corresponding to leakage current of the positive busbar of a three-phase street lamp), average leakage current of three phases Three-phase imbalance Three-phase equivalent leakage current ;

[0069] (4.2) Calculate the energy proportion of WPT nodes: Perform j=5-level WPT decomposition on the three-phase equivalent leakage current to obtain the coefficients of each node. The leakage current energy of the nth node in the jth layer (N is the number of sampling points), total leakage current energy of the j-th layer Node energy percentage ;

[0070] (4.3) Calculate the total energy entropy of WPT: energy probability of each node (satisfy Total energy entropy ;

[0071] (4.4) Calculate the ripple coefficient: Extract the ripple component of the steady-state leakage current signal, and calculate the ripple coefficient. ,in This represents the maximum ripple value. This is the minimum ripple value. This represents the steady-state effective value of the leakage current.

[0072] (4.5) Calculate the transient pulse width: Extract the pulse segment of the transient leakage current signal, and calculate the transient pulse width. ,in The pulse start time, This is the time when the pulse ends.

[0073] (5) Calculate the dynamic threshold The grounding performance level is output according to the dynamic judgment rules. If the grounding performance is good, step (9) is executed. If the grounding performance needs to be re-inspected or the grounding performance is unqualified, step (6) is executed. As the baseline threshold, This is a stationarity correction term. This is a three-phase imbalance correction term.

[0074] The dynamic threshold designed in this invention integrates multi-dimensional correction coefficients to solve the problem of high false alarm rate of existing fixed thresholds.

[0075] The parameters are defined as follows:

[0076] The baseline threshold is 30mA, based on IEC60364-4-41, which is the safe DC leakage current threshold for the human body.

[0077] Stationarity correction term To avoid false alarms due to non-stationary leakage current, the threshold should be increased.

[0078] Three-phase imbalance correction term , The threshold value represents the three-phase imbalance. When the imbalance is high, the threshold value is lowered to increase sensitivity.

[0079] : Environmental correction coefficient, output by environmental perception module 5.

[0080] : Load correction coefficient, output by load adaptation module 4.

[0081] Accordingly, the dynamic judgment rule is:

[0082] like and and Output: "Good grounding performance";

[0083] like or or Output: "Grounding performance needs to be re-inspected";

[0084] like or or Output: "Grounding performance is unqualified".

[0085] (6) The effective value of the three-phase equivalent leakage current WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The five-dimensional leakage current features are used as inputs to a trained SVM classification model to identify leakage current fault types and locate fault point information based on street light topology.

[0086] The fault classification and localization module 11 has a built-in trained SVM classification model. It takes five-dimensional leakage current features as input and outputs three types of leakage current faults.

[0087] Resistive leakage current: characterized by Continuously increasing , (Energy is concentrated at steady-state nodes); marked "urgent maintenance required".

[0088] Capacitive leakage current: characterized by Small fluctuations , (Energy is concentrated at the mid-frequency node); marked "No power outage required, re-inspect after rain".

[0089] Pulse-type leakage current: characterized by peak , , (Energy is concentrated at high-frequency nodes); Mark "Pulse parameters, no maintenance required".

[0090] In one embodiment, the training process of the SVM classification model is step (6.1), which specifically includes:

[0091] (6.1.1) Extract 1000 sets of sample data from the historical fault database, including 333 sets of resistive leakage current samples, 333 sets of capacitive leakage current samples, and 334 sets of pulse leakage current samples. Each set of sample data includes five-dimensional leakage current characteristics and corresponding leakage current fault type labels.

[0092] (6.1.2) An improved simulated annealing algorithm is used to optimize the parameters of the SVM classification model. The optimized parameters include the penalty parameter. Kernel width The objective function is ,in To predict the fault type of leakage current, Scoring based on the actual leakage current fault type. The regularization coefficient is . These are the effective values ​​of the three-phase equivalent leakage current. WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The weight coefficients, and satisfying ;

[0093] (6.1.3) Divide the sample data into training set and test set in a 7:3 ratio. Use the training set to train the SVM classification model and use the test set to verify the model accuracy. Stop training when the model classification accuracy is ≥97%.

[0094] Fault location is based on the street light topology (pole number, adjacent spacing) and the difference in leakage current of the PE line. The process is step (6.2), which specifically includes:

[0095] (6.2.1) The streetlights are numbered 1, 2...N, and the leakage current of the corresponding PE line of the light pole is as follows: , ... ;

[0096] (6.2.2) Calculate the difference in leakage current of the PE line between adjacent light poles. ;

[0097] (6.2.3) Filter out the maximum value The fault point was determined to be located at the first The lamppost and the first Between each light pole, the distance from the fault point to the first... The distance between the light poles ,in For the first The lamppost and the first The length of the cable between the light poles.

[0098] (7) Determine the steady-state components If the first threshold is exceeded, a continuous leakage current alarm, leakage current fault type, and fault location information will be output. The first threshold is... ,when At that time, a continuous leakage current alarm will be output.

[0099] (8) Determine transient components If the peak value exceeds the second threshold, an instantaneous leakage current alarm is output, and the pulse width and timestamp are recorded. If the number of pulses exceeds 5 within one hour, an emergency maintenance notification is triggered. The second threshold is... ,when At that time, an instantaneous leakage current alarm will be output.

[0100] (9) Upload the test results and display them on the display screen 12-2. The test results include the ground performance level, fault type, fault point distance, environmental parameters, and load parameters.

[0101] like Figure 2 As shown, a wavelet transform-based DC street lamp leakage current detection device includes: an interface component 1, a test circuit 2, a gear selection circuit 3, a load adaptation module 4, an environmental sensing module 5, a sampling module 6, a main control unit 7, a wavelet transform module 8, a threshold judgment module 9, a communication module 10, a fault classification and location module 11, and a human-machine interaction component 12; the interface component 1 includes a first interface 1-1, a second interface 1-2, a first detection clamp 1-3, and a second detection clamp 1-4; the first interface 1-1 is connected to a DC bus 14, the second interface 1-2 is connected to a load or human body impedance simulation network 15, and the first detection clamp... Clamps 1-3 are held on the positive busbar, and the second detection clamps 1-4 are held on the negative busbar; the test circuit 2 is connected to the interface component 1, the main control unit 7, and the sampling module 6; the gear selection circuit 3, the load adaptation module 4, and the sampling module 6 are connected in sequence; the sampling module 6 is connected to the wavelet transform module 8 and the main control unit 7; the wavelet transform module 7 is connected to the environmental perception module 5, the main control unit 7, and the threshold judgment module 9; the main control unit 7 is connected to the environmental perception module 5, the threshold judgment module 9, the fault classification and location module 11, the communication module 10, and the human-machine interaction component 12; the threshold judgment module 9 is connected to the fault classification and location module 11.

[0102] The interface component 1 enables "plug-and-play" connection to street light busbars, loads, or human body impedance simulation network 15. The first interface 1-1 supports connection to street light power supply busbars with voltage levels ranging from 24V to 400V. The first and second detection clamps 1-3 and 1-4 are connected to the main unit via 0.5m silicone rubber sheathed cables with a withstand voltage of 1kV and can withstand environments from -40℃ to +85℃. The clamp jaws employ open-type Hall effect sensors with a 28mm aperture, capable of clamping cables with a cross-section ≤240mm². The jaws feature automatic zeroing and self-locking mechanisms, allowing for one-handed operation. A buzzer is installed on the device to promptly issue a danger alarm. After wiring initialization, the device performs a self-test. If the self-test fails, fault handling is initiated; if the self-test passes, subsequent detection sampling and calculations are completed.

[0103] In one embodiment, the first detection clamp 1-3 and the second detection clamp 1-4 keep their jaws closed and return to zero. If the bias is > ±5mA, pressing the button on the side of the detection clamp completes the automatic zeroing, which takes <1s.

[0104] The test circuit 2 includes a power supply 2-1, an overvoltage protection unit 2-2, and an isolated DC / DC converter 2-3. The power supply 2-1 draws power from the first interface 1-1 and supplies power to the internal components of the device. The overvoltage protection unit 2-2 is used to prevent damage to the device when the bus voltage exceeds 400V. The isolated DC / DC converter 2-3 converts the 400V DC bus voltage into 3.3V and 5V internal power supply voltages. The power supply 2-1 is connected to the main control unit 7. In one embodiment, when the power supply 2-1 is closed, the device is powered on, and the display screen 12-2 of the human-machine interface component 12 displays a self-test interface. If no fault is detected after 3 seconds, it automatically enters the "online monitoring" mode. If overvoltage (bus voltage > 400V) or sensor fault is detected, the relay is immediately disconnected and a "fault code" is displayed. The fault codes include E01 = overvoltage and E02 = humidity sensor fault.

[0105] The range selection circuit 3 consists of a relay matrix and a precision voltage divider resistor network, providing three voltage division ratios of 1:1, 1:10, and 1:100, corresponding to voltage ranges of 24V, 220V, and 400V. It is automatically switched by the load adapter module 4, with a switching time of <20ms and no test interruption.

[0106] The load adaptation module 4 collects the voltage ripple and current waveform at the load terminal, automatically identifies the load type, and outputs a load correction coefficient. It also controls the gear selection circuit 3 to switch the sampling range and configures the sampling rate of the sampling module 6.

[0107] In one embodiment, if the load adapter module 4 identifies that the load type is an LED light, then the control gear selection circuit 3 switches the sampling range to 1:10, and the sampling rate of the sampling module 6 is 10kHz.

[0108] In one embodiment, if the load adaptation module 4 identifies the load type as an intelligent controller, the control level selection circuit 3 switches the sampling range to 1:20, and the sampling rate of the sampling module 6 is 16kHz.

[0109] In one embodiment, if the load adapter module 4 identifies that the load type is an emergency power supply, the control range selection circuit 3 switches the sampling range to 1:100, and the sampling rate of the sampling module 6 is 32kHz.

[0110] The environmental sensing module 5 includes a rain / fog sensor, a temperature sensor, and a light sensor, which collect relative humidity, cable temperature, and light intensity, respectively. The rain / fog sensor is a capacitive humidity sensor (SHT30), the light sensor is an NTC thermistor (accuracy ±0.5℃), and the light sensor is a photoresistor (detection range 0~1000 lux). The environmental sensing module 5 is connected to the main control unit 7 via an I2C bus, with a data update frequency of 1Hz. The main control unit 7 outputs an environmental correction coefficient based on the parameters: if RH>85% and T≤40℃ and light intensity>50 lux (heavy rain / fog during the day). If RH≤85% and T>40℃ and light intensity<50 lux (high nighttime temperature); If RH≤85% and T≤40℃ and light intensity<50 lux (normal at night). .

[0111] The sampling module 6 includes a four-channel 24-bit Σ-Δ ADC and a sampling resistor with a rated power of 5W and a rated power of 0.5Ω±0.1%. A TVS diode is connected in parallel with the sampling resistor to suppress surges. The sampling module 6 synchronously acquires leakage current signals from the positive and negative terminals of the bus, the PE line, and the load terminal at an adapted sampling rate. The sampled data is transmitted to the main control unit 7 via the SPI bus. Each frame of data includes a timestamp (with an accuracy of 1ms) to ensure signal time consistency.

[0112] The main control unit 7 uses an STM32H743 MCU to complete... RMS value It can perform five-dimensional leakage current characteristics and dynamic threshold calculation, as well as module collaborative control and data interaction. It has a built-in 1MB Flash for storing historical fault data (last 72 hours).

[0113] The wavelet transform module 8 uses an STM32H743 MCU as its core (with built-in DSP and double-precision FPU) and implements the Daubechies-4 (db4) wavelet basis through the CMSIS-DSP library. It decomposes the leakage current signal into transient and steady-state components. The number of decomposition layers is determined to be 5 layers through frequency band matching. The first and second layers of high-frequency nodes (nodes 1-3) reconstruct the instantaneous pulse leakage current, the third and fourth layers of mid-frequency nodes (nodes 4-7) reconstruct the intermittent leakage current, and the fifth layer of low-frequency nodes (node ​​0) reconstructs the steady-state leakage current.

[0114] The threshold determination module 9 is used to determine whether there is abnormal leakage current, and includes a dynamic threshold determination module, a first threshold determination module, and a second threshold determination module. The dynamic threshold determination module is connected to the main control unit 7, the input terminal of the first threshold determination module is connected to the steady-state component output terminal of the wavelet transform module 8, and the input terminal of the second threshold determination module is connected to the transient component output terminal of the wavelet transform module 8.

[0115] The communication module 10 uses a USB Type-C interface and supports the USB-PD protocol. It is used to upload detection data to external detection devices and receive control commands.

[0116] The fault classification and location module 11 has an integrated SVM classification model and a fault location submodule. The trained SVM classification model identifies resistive / capacitive / pulse leakage current based on 5-dimensional features. The fault location submodule uses a topology method to locate the fault location with a deviation of ≤2.3m.

[0117] The human-machine interface component 12, designed to improve ease of operation and maintenance, includes a housing 12-1 and a display screen 12-2. The housing 12-1 is made of flame-retardant PC+ABS alloy, and the overall potting achieves an IP65 protection rating, allowing for outdoor operation in rainy or foggy weather. The display screen 12-2 is a 3.5-inch OLED touchscreen. Figure 3 As shown, the display is divided into four areas: Area A: Real-time scrolling display of transient components. Waveform, supports two-finger zoom; Area B: Large font display of steady-state RMS value. Energy entropy Grounding performance level, indicated by color: green for good, yellow for requiring re-inspection, and red for unqualified; Zone C: Dynamic threshold. Slide bar, supports manual fine-tuning by the user; D zone: shortcut function keys, including record, zeroing, maintenance mode, and data export; audible and visual alarm unit: built-in buzzer and dual-color LED, outputs 800Hz continuous alarm sound when there is a fault, alarm trigger delay <100ms, dual-color LED displays green to indicate normal and red to indicate fault.

[0118] Figure 4This is a handheld external structural diagram of the device of the present invention, which includes a housing 12-1, a display screen 12-2, a communication interface 10-1, a dual detection clamp storage compartment 13, a first detection clamp 1-3, a second detection clamp 1-4, and a power supply 2-1; the display screen 12-2 is located on the top front of the housing 12-1, the power supply 2-1 is located on the bottom front, and the communication interface 10-1, the first detection clamp 1-3, the second detection clamp 1-4, and the dual detection clamp storage compartment 13 are arranged sequentially on the side.

[0119] like Figure 5 As shown, to conform to the IEC 60479-1 human body electric shock model, the human body impedance simulation network 15 includes: resistor R1, resistor R2, capacitor C1, capacitor C2, and resistor R3. Resistor R1 and capacitor C1 are connected in parallel to form parallel branch 1, and resistor R2 and capacitor C2 are connected in parallel to form parallel branch 2. One end of parallel branch 1 and parallel branch 2 is connected to the second interface 1-2, and the other end is connected in series with resistor R3 and then connected to the ground terminal. Specifically, resistor R1 = 10kΩ, resistor R2 = 10kΩ, capacitor C1 = 100nF, capacitor C2 = 100nF, and resistor R3 = 500Ω. Resistor R1 and resistor R2 simulate skin contact resistance, resistor R3 simulates body resistance, and capacitors C1 and C2 are parasitic capacitances. When a user needs to conduct a "human perception" experiment, the system can quickly switch, and the device automatically lowers the threshold to ensure a more rigorous personal safety assessment.

[0120] In this embodiment, a 220V DC street light line in a certain city is used as the test object. The line is 1km long, contains 20 light poles, with a pole spacing of 50m, and the load is 18W LED lights. The complete testing process is performed as follows:

[0121] Device wiring: Connect the first interface 1-1 to the 220V DC bus 14 (positive, negative, PE line), and connect the second interface 1-2 to the LED load of the 10th light pole; clamp the positive bus with the first test clamp 1-3 and the negative bus with the second test clamp 1-4, and press the zero button to complete the self-test.

[0122] Load adapter module 4 acquires the 10kHz ripple signal of the LED lamp, identifies it as an "LED lamp load," and outputs... Switch the sampling range to 1:10, sampling rate 10kHz; environmental sensing module detects: RH=92% (heavy rain / fog), T=18℃, light intensity=800lux (daytime), output. .

[0123] Sampling module 6 simultaneously acquires four leakage current signals.

[0124] Wavelet transform module 8 performs 5-level db4 decomposition and reconstructs the following: (1) Steady-state component (2) Transient components Peak value = 65mA, at this time There is no instantaneous alarm.

[0125] Main control unit 7 calculates: (1) Calculates the five-dimensional leakage current characteristic value: , , ; , ; (2) Dynamic threshold .

[0126] because , max( If 5% < 10%, the device outputs "Good grounding performance" (green LED lights up); the test results are uploaded to the platform via USB Type-C, and the entire process takes less than 5 minutes, which is significantly better than the 2 hours required for traditional lamp-by-lamp inspection.

[0127] A resistive leakage fault (cable insulation damage) is set between the 5th and 6th light poles, and the test is repeated:

[0128] (1) Steady-state components , , ;

[0129] (2) The SVM classification model outputs "resistive leakage", and the fault location submodule calculates... (Maximum value) The distance from the fault point to the 5th light pole is 21.5m (actual distance is 20m, deviation is 1.5m).

[0130] (3) The device outputs a "continuous leakage alarm". The maintenance personnel can find the fault point within 15 minutes based on the location, which improves the maintenance efficiency by 87.5%.

[0131] The present invention has been subjected to comparative tests on specific road sections. The comparative tests include: comparison of false alarm rate in rainy and foggy weather, comparison of fault location accuracy, and comparison of fault classification accuracy.

[0132] (a) Comparative test of false alarm rate in rainy and foggy weather.

[0133] Test conditions: Ten 1km long DC street light lines were selected (each line contains 20 light poles, and the load is LED lights). The lines were continuously monitored for 24 hours under the following conditions: sunny day (RH=40%, T=25℃), light rain day (RH=70%, T=20℃), and heavy rain and fog day (RH=90%, T=15℃). The number of false alarms of the leakage current detection device was counted (false alarm definition: no actual leakage but output "grounding performance needs to be re-inspected" or "grounding performance is unqualified").

[0134] Test subject: The original device did not have an environmental sensing module; the device of this invention includes an environmental sensing module.

[0135] Test results: See Table 1.

[0136] Test conclusion: The modified device outputs through the environmental perception module... The modified dynamic threshold effectively reduced the false alarm rate caused by environmental interference in rainy and foggy weather, especially in heavy rain and fog.

[0137] Environment type Number of false alarms from the original device (total of 10 lines) False alarm count for this device (total of 10 lines) False alarm rate reduction percentage sunny 5 times 2 times 60% Light rain 18 times 5 times 72.2% Heavy rain and fog 32 times 7 times 78.1%

[0138] Table 1

[0139] (II) Comparison test of fault location accuracy.

[0140] Test conditions: Select a 1km long DC street light line (including 20 light poles, with a spacing of 50m between adjacent light poles). Set up 3 types of faults (resistive leakage, capacitive leakage, and pulse leakage) between light poles 5-6 (20m from light pole 5), between light poles 10-11 (35m from light pole 10), and between light poles 15-16 (10m from light pole 15), for a total of 9 fault scenarios. Each scenario was tested 5 times, and the deviation between the fault location distance and the actual fault distance was recorded.

[0141] Test subject: The device of this invention.

[0142] Test results: See Table 2.

[0143] Test results: Under various fault scenarios, the average fault location deviation of this modified device is ≤2.3m, and the positioning accuracy is ≥85%, which can meet the positioning requirements of DC street light fault operation and maintenance, and significantly shorten the fault troubleshooting time.

[0144] Fault type Actual location of the fault (distance from a certain light pole) Average positioning distance (5 tests) Mean positioning deviation Positioning accuracy resistive leakage 20m from the 5th lamppost At 21.2m on the 5th lamppost 1.2m 94% resistive leakage 35m from the 10th light pole At 36.5m of the 10th light pole 1.5m 95.7% resistive leakage 10m from the 15th light pole At 10.8m of the 15th light pole 0.8m 92% Capacitive leakage 20m from the 5th lamppost At 22.1m on the 5th lamppost 2.1m 90% Capacitive leakage 35m from the 10th light pole At 37.3m on the 10th light pole 2.3m 93.4% Capacitive leakage 10m from the 15th light pole At 11.5m of the 15th light pole 1.5m 85% Pulse-type leakage 20m from the 5th lamppost At 21.8m on the 5th light pole 1.8m 91% Pulse-type leakage 35m from the 10th light pole At the 36.9m mark of the 10th light pole 1.9m 94.6% Pulse-type leakage 10m from the 15th light pole At 11.2m on the 15th light pole 1.2m 88%

[0145] Table 2

[0146] (III) Fault classification accuracy test.

[0147] Test conditions: Extract 300 sets of fault data (100 sets of resistive leakage, 100 sets of capacitive leakage, and 100 sets of pulse leakage) from the historical fault database of DC street lights, input the data into the fault classification and location module of this modified device, and count the number of times the device correctly classifies each type of fault.

[0148] Test subject: The device of this invention.

[0149] Test results: See Table 3.

[0150] Test conclusion: Accuracy above.

[0151] Fault type Number of test samples (groups) Number of times the classification was correct (groups) Classification accuracy resistive leakage 100 98 98% Capacitive leakage 100 97 97% Pulse-type leakage 100 96 96% total 300 291 97%

[0152] Table 3

[0153] Example 2: The difference between this example and Example 1 is that it also includes the following steps: assessing the insulation level, the main control unit 7 sends a Modbus command (function code 0x06) to the street light intelligent controller to put it into maintenance mode and disconnect the load; repeating steps S2 to S8 to obtain the no-load leakage current value. The result indicates that "insulation resistance > 10MΩ, no maintenance required".

[0154] The above description describes specific embodiments of the present invention and the technical principles employed. Any changes made in accordance with the concept of the present invention that do not exceed the spirit of the specification and drawings should still fall within the protection scope of the present invention.

Claims

1. A method for detecting leakage current of DC streetlights based on wavelet transform, characterized in that... Includes the following steps: (1) Identify the load type and collect environmental parameters, and output the load correction coefficient. Environmental correction factor And adjust the sampling range and sampling rate; (2) Collect leakage current signals from the positive and negative terminals of the bus, the PE line, and the load terminal at the adjusted sampling rate; (3) Perform 5-level db4 wavelet packet decomposition on the collected leakage current signal to obtain the transient component. With steady-state components ; (4) Calculation RMS value Three-phase equivalent leakage current RMS value WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width ; (5) Calculate the dynamic threshold The grounding performance level is output according to the dynamic judgment rules. If the grounding performance is good, step (9) is executed. If the grounding performance needs to be re-inspected or the grounding performance is unqualified, step (6) is executed. As the baseline threshold, This is a stationarity correction term. This is a three-phase imbalance correction term; (6) The effective value of the three-phase equivalent leakage current WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The five-dimensional leakage current features are input into the trained SVM classification model to identify leakage current fault types and locate fault point information based on street light topology. (7) Determine the steady-state components If the first threshold is exceeded, output a continuous leakage current alarm and information on the leakage current fault type and fault point. (8) Determine transient components If the peak value exceeds the second threshold, then output an instantaneous leakage current alarm and record the pulse width and timestamp. (9) Upload the test results and display them on the screen. The test results include the ground performance level, fault type, distance to the fault point, environmental parameters, and load parameters.

2. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, It also includes the step of assessing the insulation level, specifically by disconnecting the load and then repeating steps 2-8 to obtain the no-load leakage current value.

3. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, In step (3), the coefficients of the nodes in layers 1-4 are reconstructed into transient components. The 0 coefficient of the 5th layer node is reconstructed into a steady-state component. .

4. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, In step (4), the calculation of the five-dimensional leakage current characteristics specifically includes: (4.1) Calculate the effective value of the three-phase equivalent leakage current. Assume the three-phase leakage current is , , Then the mean Imbalance Three-phase equivalent leakage current RMS value ; (4.2) Calculate the energy percentage of nodes Energy of the 5th layer node Total energy Node energy percentage ; (4.3) Calculate the total energy entropy of WPT Energy probability , ; (4.4) Calculate the ripple system ,in It is the maximum value of the ripple. It is the minimum value of the ripple; (4.5) Transient pulse width ,in It is the pulse start, This indicates the end of the pulse.

5. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, The stationarity correction term The calculation formula is: 。 6. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, The three-phase imbalance correction term The calculation formula is: 。 7. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, The dynamic judgment rule in step (5) is as follows: like and and Output: "Good grounding performance"; like or or Output: "Grounding performance needs to be retested"; like or or Output: "Grounding performance is unqualified".

8. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, The training process of the trained SVM classification model is step (6.1), which specifically includes: (6.1.1) Extract M sets of sample data from the historical fault database. Each set of sample data contains five-dimensional leakage current characteristics and corresponding leakage current fault type labels. (6.1.2) An improved simulated annealing algorithm is used to optimize the SVM classification model parameters. The optimized parameters include the penalty parameter C and the kernel function width g. The objective function is... ,in To predict the fault type of leakage current, Scoring based on the actual leakage current fault type. The regularization coefficient is . These are the effective values ​​of the three-phase equivalent leakage current. WPT total energy entropy Node energy percentage ripple coefficient Transient pulse width The weight coefficients, and satisfying ; (6.1.3) Divide the sample data into training set and test set in a 7:3 ratio. Use the training set to train the SVM classification model and use the test set to verify the model accuracy. Stop training when the model classification accuracy is ≥97%.

9. The method for detecting leakage current of DC streetlights based on wavelet transform according to claim 1, characterized in that, The fault location process is step (6.2), which specifically includes: (6.2.1) The streetlights are numbered 1, 2...N, and the leakage current of the corresponding PE line of the light pole is as follows: , ... ; (6.2.2) Calculate the difference in leakage current of the PE line between adjacent light poles. ; (6.2.3) Filter out the maximum value The fault point was determined to be located at the first The lamppost and the first Between each light pole, the distance from the fault point to the first... The distance between the light poles ,in For the first The lamppost and the first The length of the cable between the light poles.

10. A device for detecting leakage current of DC streetlights based on wavelet transform, characterized in that... include: The interface component (1) comprises an interface (1-1), a test circuit (2), a gear selection circuit (3), a load adaptation module (4), an environmental perception module (5), a sampling module (6), a main control unit (7), a wavelet transform module (8), a threshold judgment module (9), a communication module (10), a fault classification and location module (11), and a human-machine interaction component (12). The interface component (1) includes a first interface (1-1), a second interface (1-2), a first detection clamp (1-3), and a second detection clamp (1-4). The first interface (1-1) is connected to the DC bus (14), the second interface (1-2) is connected to the load or human body impedance simulation network (15), the first detection clamp (1-3) is clamped on the positive bus, and the second detection clamp (1-4) is connected to the DC bus (14). Clamps (1-4) are held in place on the negative busbar; the test circuit (2) is connected to the interface component (1), the main control unit (7), and the sampling module (6); the gear selection circuit (3), the load adaptation module (4), and the sampling module (6) are connected in sequence; the sampling module (6) is connected to the wavelet transform module (8) and the main control unit (7); the wavelet transform module (8) is connected to the environmental perception module (5), the main control unit (7), and the threshold judgment module (9); the main control unit (7) is connected to the environmental perception module (5), the threshold judgment module (9), the fault classification and location module (11), the communication module (10), and the human-machine interaction component (12); and the threshold judgment module (9) is connected to the fault classification and location module (11).

Images