An intelligent insulation impedance detection system and method for an energy storage converter

Abstract

This invention belongs to the field of electrical safety technology and specifically discloses an intelligent insulation impedance detection system and method for energy storage converters. Based on a bridge balance structure, the system integrates temperature and humidity sensors and voltage sensors to collect environmental parameters and bridge imbalance signals in real time. A PID algorithm drives a digital potentiometer and resistor array to dynamically adjust the bridge balance, while a temperature and humidity compensation model is used to eliminate environmental interference. Bandpass filtering and a detection model are used to intelligently analyze preprocessed multidimensional data and output high-precision insulation impedance values. This invention achieves fully automated detection with high accuracy and fast response, making it suitable for complex operating conditions and significantly improving the reliability of insulation performance assessments for energy storage converters.

Description

Technical Field

[0001] The present invention belongs to the field of electrical safety technology, and in particular relates to an intelligent insulation impedance detection system and method for an energy storage converter. Background Art

[0002] Insulation impedance testing is a key step in evaluating the insulation performance of energy storage converters when it comes to the safe operation of electrical equipment. Traditional testing methods primarily use the bridge method, which measures the resistance of the insulating material to determine insulation defects. However, this method has significant limitations:

[0003] Environmental interference: Environmental factors such as temperature, humidity, and electromagnetic interference can easily disrupt bridge balance and reduce measurement accuracy. For example, temperature changes can alter the resistivity of insulating materials, while increased humidity can introduce surface leakage currents. Traditional bridge methods struggle to compensate for these variables in real time, resulting in errors of up to ±10%.

[0004] Insufficient signal processing capabilities: Traditional methods lack efficient feature extraction methods for noisy bridge output signals, making it difficult to distinguish valid signals from interference. Especially in complex electromagnetic environments, the detection results are unreliable.

[0005] As energy storage converters evolve toward higher power, modularization, and non-isolation, these issues become increasingly prominent. While artificial intelligence (AI) technology has demonstrated advantages in signal processing, existing technologies have yet to fully integrate the bridge method with intelligent algorithms, failing to meet the demands of high-precision, real-time detection. Summary of the Invention

[0006] The purpose of the present invention is to overcome the defects in the prior art and provide an intelligent detection system and method for insulation impedance of an energy storage converter.

[0007] A first aspect of the present invention provides an intelligent insulation impedance detection system for an energy storage converter, comprising:

[0008] Bridge balancing structure: Using a Wheatstone bridge topology, the four bridge arms form a quadrilateral loop, with two diagonal arms connected to a DC excitation power supply and a microvolt voltage detection port, respectively. The four bridge arms include at least one variable resistance arm and one measurement arm.

[0009] Data acquisition module: used to collect ambient temperature, humidity and bridge output voltage in real time;

[0010] Control module: connected to the data acquisition module, calculates the bridge balance deviation based on the ambient temperature, humidity and bridge output voltage, and outputs adjustment instructions;

[0011] Automatic adjustment circuit: connected to the control module and the variable resistor arm respectively, responding to the adjustment instruction, controlling the variable resistor arm to adjust the resistance value to achieve dynamic balance of the bridge;

[0012] The insulation impedance detection module is connected to the data acquisition module and outputs an insulation impedance detection value through a detection model based on the bridge output voltage.

[0013] A further solution is that the four bridge arms include:

[0014] The first standard resistance arm and the second standard resistance arm are both high-precision metal film resistors, connected in series between the positive electrode of the excitation power supply and the common node to form a fixed reference bridge arm;

[0015] The variable resistor arm is a digital potentiometer connected in series between the common node and the negative pole of the excitation power supply; the digital potentiometer is also connected to the control module via an I²C or SPI communication interface;

[0016] Measuring arm: connected to the measured circuit of the energy storage converter, located on the same side of the bridge arm as the variable resistor arm, used to access the insulation impedance to be measured.

[0017] A further solution is that the automatic adjustment circuit includes:

[0018] A resistor array connected in parallel at both ends of the variable resistor arm, wherein the resistance values of the resistors in the resistor array are respectively exponentially or linearly distributed to cover the target adjustment range;

[0019] Each resistor in the resistor array is connected in series with a MOSFET switch.

[0020] A further solution is that the data acquisition module includes:

[0021] Temperature sensor: used to collect the ambient temperature of the bridge balance structure in real time;

[0022] Humidity sensor: used to collect environmental humidity in real time;

[0023] Voltage sensor: A microvolt-level sensor based on a high-impedance differential amplifier, connected to the voltage detection port between the two midpoints of the bridge, used to measure the bridge unbalance voltage ΔU.

[0024] A further solution is that the control module includes a data acquisition layer, an algorithm processing layer, and a control output layer;

[0025] The data acquisition layer is connected to the temperature sensor, humidity sensor and voltage sensor respectively, and is used to obtain the ambient temperature, humidity and bridge unbalance voltage ΔU;

[0026] The algorithm processing layer uses the bridge unbalance voltage ΔU as an input variable, compensates ΔU according to the ambient temperature and humidity, and calculates the bridge balance adjustment value through the PDI control algorithm;

[0027] The control output layer generates an adjustment instruction for the digital potentiometer based on the balance adjustment amount calculated by the PID control algorithm, sends the adjustment instruction to the digital potentiometer via the SPI or I²C communication interface, and adjusts the variable resistor arm through the digital potentiometer to make the bridge output voltage approach zero.

[0028] A further solution is that the algorithm processing layer is also connected to a BMS timing control module, which activates the detection circuit of each module in a time-sharing manner to avoid the parallel effect caused by the simultaneous connection of multiple modules to detect internal resistance.

[0029] A further solution is that the insulation impedance detection module includes a signal preprocessing unit and a detection model;

[0030] The signal preprocessing unit filters out the low-frequency drift and high-frequency noise of the bridge unbalanced voltage ΔU through a bandpass filter, retains the effective frequency band of the polarization process of the insulating material, performs normalization processing, and converts it into dimensionless data between 0 and 1;

[0031] The construction process of the detection model is as follows:

[0032] A high-precision insulation resistance meter is used to collect insulation resistance test values under different working conditions in multiple time series. The ambient temperature, humidity, and bridge unbalanced voltage ΔU are also collected simultaneously. The ambient temperature, humidity, and bridge unbalanced voltage ΔU data are organized into a three-dimensional tensor format (T, H, ΔU), where T is the ambient temperature and H is the humidity. A large number of insulation resistance test values and (T, H, ΔU) matching the insulation resistance test values are obtained and manually labeled by experts. After labeling, (T, H, ΔU) are used as the input of a convolutional neural network, and the insulation resistance test values are used as the output of the convolutional neural network for iterative training. A detection model based on the insulation resistance test values output by (T, H, ΔU) is obtained.

[0033] A second aspect of the present invention provides an intelligent detection method for insulation impedance of an energy storage converter, using the above-mentioned system, comprising the following steps:

[0034] System initialization: Start the bridge balancing structure, adjust the variable resistor arm through the digital potentiometer to make the bridge unbalance voltage ΔU ≤ 10μV, and record the initial balance parameters;

[0035] Real-time data acquisition: The data acquisition module synchronously collects ΔU, T, and H to generate a raw signal sequence with a time stamp;

[0036] Dynamic balance adjustment: The control module calculates the adjustment amount through the PID algorithm according to ΔU, T, and H, and drives the automatic adjustment circuit to switch the digital potentiometer or parallel resistor to make ΔU return to balance;

[0037] Signal preprocessing: Band-pass filtering and normalization are performed on ΔU, and it is then fused with T and H to form three-dimensional time series data (T, H, ΔU).

[0038] Insulation impedance detection: Input the pre-processed three-dimensional data (T, H, ΔU) into the detection model and output the real-time insulation impedance detection value.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] Through the deep integration of bridge dynamic balance control, environmental parameter compensation, intelligent signal processing and multi-module collaborative technology, the present invention achieves high-precision, high-reliability and full automation of insulation impedance detection of energy storage converters, providing strong technical guarantee for the safe operation of electrical equipment.

[0041] This method dynamically compensates for bridge voltage imbalance by collecting real-time ambient temperature and humidity data and combining it with a PID control algorithm. This significantly reduces measurement errors caused by temperature and humidity variations and improves insulation impedance detection accuracy. A bandpass filter preprocesses the bridge output signal, effectively filtering out low-frequency drift and high-frequency noise while retaining the characteristic polarization frequency band of the insulation material. A convolutional neural network model is then used to intelligently analyze multidimensional data (T, H, ΔU), addressing signal distortion in complex electromagnetic environments.

[0042] The present invention is based on an automatic adjustment circuit of a digital potentiometer and a high-speed MOSFET resistor array, with a response time of less than , and cooperates with the PID algorithm to achieve rapid zeroing of the bridge (ΔU ≤ 10μV), ensuring the stability of the system under dynamic working conditions and avoiding the measurement deviation caused by manual adjustment lag of the traditional bridge method.

[0043] The present invention uses the BMS timing control module to activate each detection circuit in a time-sharing manner, avoiding mutual interference of internal resistance when multiple modules are connected in parallel, solving the problem of decreased accuracy of multi-channel detection in traditional methods, and is suitable for complex application scenarios of high-power, modular energy storage converters.

[0044] The present invention utilizes a pre-trained CNN detection model to directly output the insulation resistance value without relying on manual experience and judgment, significantly improving detection efficiency. At the same time, the system runs automatically throughout the entire process after initialization, greatly reducing manual intervention and lowering operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The following drawings are merely provided for illustrative purposes only and are not intended to limit the scope of the present invention.

[0046] Figure 1 This is a connection diagram of the insulation impedance intelligent detection system;

[0047] In the figure: 1. Bridge balancing structure; 2. Data acquisition module; 3. Control module; 4. Automatic adjustment circuit; 5. Insulation impedance detection module; 6. BMS timing control module; 7. Standard resistor arm; 8. Variable resistor arm; 9. Measurement arm; 10. Temperature sensor; 11. Humidity sensor; 12. Voltage sensor; 13. Data acquisition layer; 14. Algorithm processing layer; 15. Control output layer; 16. Detection model; 17. Signal preprocessing unit; 18. Resistor array; 19. MOSFET switch. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution, design method and advantages of the present invention more clear, the present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] Example 1

[0050] like Figure 1 As shown, this embodiment provides an intelligent insulation impedance detection system for an energy storage converter, including:

[0051] Bridge balancing structure 1: adopts a Wheatstone bridge topology, with four bridge arms forming a quadrilateral loop, two diagonal portions of which are connected to a DC excitation power supply and a microvolt voltage detection port, respectively. The four bridge arms include at least one variable resistor arm 8 and one measurement arm 9. The four bridge arms include: a first standard resistor arm 7 and a second standard resistor arm 7: both are high-precision metal film resistors, connected in series between the positive pole of the excitation power supply and a common node, forming a fixed reference bridge arm; a variable resistor arm 8: a digital potentiometer, connected in series between the common node and the negative pole of the excitation power supply; the digital potentiometer is also connected to the control module 3 via an I²C or SPI communication interface; a measurement arm 9: connected to the energy storage converter circuit under test, located on the same side of the bridge arm as the variable resistor arm 8, and used to access the insulation impedance to be measured;

[0052] Data acquisition module 2: used to collect ambient temperature, humidity, and bridge output voltage in real time. In this embodiment, data acquisition module 2 includes: a temperature sensor 10: used to collect the ambient temperature of the bridge balance structure 1 in real time; a humidity sensor 11: used to collect the ambient humidity in real time; a voltage sensor 12: a microvolt-level sensor based on a high-impedance differential amplifier, connected to the voltage detection port between the two midpoints of the bridge, used to measure the bridge unbalance voltage ΔU;

[0053] Control module 3: connected to the data acquisition module 2, calculates the bridge balance deviation based on the ambient temperature, humidity, and bridge output voltage, and outputs an adjustment instruction. The control module 3 includes a data acquisition layer 13, an algorithm processing layer 14, and a control output layer 15. The data acquisition layer 13 is connected to the temperature sensor 10, humidity sensor 11, and voltage sensor 12, respectively, to obtain the ambient temperature, humidity, and bridge imbalance voltage ΔU. The algorithm processing layer 14 uses the bridge imbalance voltage ΔU as an input variable, compensates ΔU based on the ambient temperature and humidity, and calculates the bridge balance adjustment value using the PDI control algorithm. The control output layer 15 generates an adjustment instruction for the digital potentiometer based on the balance adjustment value calculated by the PID control algorithm, and sends the adjustment instruction to the digital potentiometer via the SPI or I²C communication interface. The digital potentiometer adjusts the variable resistor arm 8 so that the bridge output voltage approaches zero. The process of compensating ΔU based on the ambient temperature and humidity is as follows:

[0054] Based on the temperature characteristics of the resistivity of the insulating material, the unbalanced bridge voltage ΔU is compensated:

[0055]

[0056] Where α is the temperature coefficient of the insulation material being tested (obtained from the equipment manual, such as:

[0057] epoxy resin α = −0.02 / °C), T0 is the initial calibration temperature (e.g., 25°C);

[0058] Combined with the surface leakage resistance parallel model, the compensation of humidity on the bridge unbalanced voltage ΔU is determined:

[0059] Implement humidity detection H, when H>60%RH, trigger the humidity compensation mechanism;

[0060] According to the formula calculate , where k is a reference parameter, representing the initial magnitude or reference value of the surface leakage resistance under specific conditions, which determines the basic magnitude of the leakage resistance, and β is the humidity influence coefficient, reflecting the effect of humidity H on the surface leakage resistance. The greater the β is, the greater the humidity H is. The faster the rate of decrease with humidity changes, the faster the decrease. Both are empirical parameters obtained by fitting historical data. They are used to quantify the effect of humidity on the surface leakage resistance in insulation resistance, so as to more accurately correct the effect of humidity on insulation resistance. In this embodiment, k = 100 megohms and β = 0.05 / RH, indicating that for every 1% increase in humidity, The exponential decay factor will change accordingly, reflecting the dynamic effect of humidity on leakage resistance;

[0061] Using the parallel model Calculate the equivalent insulation resistance under the influence of humidity ;

[0062] The unbalanced voltage of the bridge is related to the impedance of the bridge arm, Replace the original Recalculate ΔU and compare it with the measured ΔU by formula Correct ΔU to offset the humidity Reduce the interference to ΔU and realize dynamic compensation of humidity to ΔU, which can more accurately reflect the true state of insulation impedance in high humidity environment and improve the reliability of bridge detection;

[0063] Automatic adjustment circuit 4: connected to the control module 3 and the variable resistor arm 8 respectively, responding to adjustment instructions and controlling the variable resistor arm 8 to adjust the resistance value to achieve dynamic balance of the bridge; the automatic adjustment circuit 4 includes: a resistor array 18 connected in parallel at both ends of the variable resistor arm 8, wherein the resistance values of the resistors in the resistor array 18 are arranged according to an exponential or linear law to cover the target adjustment range; each resistor in the resistor array 18 is connected in series with a MOSFET switch 19;

[0064] The insulation resistance detection module 5 is connected to the data acquisition module 2 and outputs an insulation resistance detection value through the detection model 16 based on the bridge output voltage.

[0065] In order to solve the problem of decreased accuracy of multi-channel detection in traditional methods, in this embodiment, the algorithm processing layer 14 is also connected to the BMS timing control module 6, which activates the detection circuit of each module in a time-sharing manner to avoid the parallel effect caused by the simultaneous connection of multiple modules to detect internal resistance.

[0066] In the above, the insulation impedance detection module 5 includes a signal preprocessing unit 17 and a detection model 16;

[0067] The signal preprocessing unit 17 filters out the low-frequency drift and high-frequency noise of the bridge unbalanced voltage ΔU through a bandpass filter, retains the effective frequency band of the polarization process of the insulating material, and performs normalization processing to convert it into dimensionless data between 0 and 1. The construction process of the detection model 16 is as follows:

[0068] A high-precision insulation resistance meter is used to collect insulation resistance test values under different operating conditions in multiple time series. The ambient temperature, humidity, and bridge unbalance voltage ΔU are also collected simultaneously. The ambient temperature, humidity, and bridge unbalance voltage ΔU data are organized into a three-dimensional tensor format (T, H, ΔU), where T is the ambient temperature and H is the humidity. A large number of insulation resistance test values and (T, H, ΔU) matching the insulation resistance test values are obtained and manually labeled by experts. After labeling, (T, H, ΔU) is used as the input of a convolutional neural network, and the insulation resistance test values are used as the output of the convolutional neural network for iterative training. A detection model 16 based on the insulation resistance test values outputted by (T, H, ΔU) is obtained. In this embodiment, the detection model 16 utilizes a convolutional neural network (CNN), and its architecture is designed as follows: Input layer: Receives preprocessed multidimensional time series data. The three feature channels of the multidimensional data (T, H, ΔU) are normalized bridge unbalance voltage, real-time temperature, and real-time humidity. The bridge signal fluctuations containing the time series and the temperature and humidity information at the corresponding time are input into the network, providing raw data for subsequent feature extraction. The first convolutional layer uses 32 1D convolutional filters, each with a kernel size of 5 (processing features from five consecutive time points at a time). The activation function uses ReLU (Reinforced Luminance) to extract local temporal features through convolution operations, such as the voltage change rate within 5ms and the temperature-humidity coupling trend. The ReLU activation function introduces nonlinearity, enabling the network to learn complex feature relationships. The max pooling layer, with a pooling size of 2, reduces the dimensionality of the feature map output by the convolutional layer, retaining the maximum features in local regions (such as the peak of voltage fluctuations). This reduces data dimensionality and computational complexity while also preventing overfitting. The second convolutional layer uses 64 1D convolutional filters with a kernel size of 3, still using ReLU as the activation function, to further extract more detailed temporal features, such as the peak of voltage fluctuations within 3ms and the sudden change point of temperature and humidity. Increasing the number of filters to 64 enables the network to learn richer feature representations. The global average pooling layer performs average pooling on the entire feature map. This compresses the spatial dimension (time series length) into a global feature vector, avoiding overfitting caused by the excessive number of parameters in the fully connected layer while preserving the overall statistical information of the feature map. The fully connected layer contains 128 neurons and uses the Reluctant Unit (ReLU) activation function. It fuses the local and global features extracted by the previous layers, forming an abstract representation related to insulation resistance through nonlinear transformations, providing high-level semantic features for the final output. The dropout layer, with a dropout rate set to 0.3, randomly shuts down some neurons to suppress network overfitting and improve the model's generalization ability on unknown data. The output layer, with one neuron and a linear activation function, performs a linear transformation on the feature vector output by the fully connected layer and directly outputs the insulation resistance prediction value.

[0069] Example 2

[0070] Based on Example 1, this embodiment provides an intelligent detection method for insulation impedance of an energy storage converter, comprising the following steps:

[0071] System initialization: Start the bridge balancing structure and connect the DC excitation power supply to the Wheatstone bridge. The control module sends an initial adjustment command to the digital potentiometer via the I²C or SPI communication interface, setting the variable resistor arm resistance to mid-scale (for example, if the total digital potentiometer resistance is 100kΩ, the initial setting is 50kΩ). The bridge imbalance voltage ΔU at the microvolt voltage detection port is monitored and the digital potentiometer resistance is iteratively adjusted to keep ΔU ≤ 10μV. The current variable resistor arm resistance, ambient temperature measured by the temperature sensor, and ambient humidity measured by the humidity sensor are recorded and stored in the control module register as initial balancing parameters.

[0072] Real-time data acquisition: The data acquisition module drives each sensor to work synchronously:

[0073] The temperature sensor (DS18B20) collects the ambient temperature of the bridge module in real time with an accuracy of ±0.1°C;

[0074] Humidity sensor (SHT30) collects ambient humidity in real time with an accuracy of ±2%RH;

[0075] A voltage sensor based on a high-impedance differential amplifier (input impedance ≥ 10MΩ) measures the unbalanced voltage ΔU at the voltage detection port between the two midpoints of the bridge;

[0076] Add a timestamp to each set of collected (T, H, ΔU) data to generate a continuous raw signal sequence, which is stored in the data buffer and awaits subsequent processing;

[0077] Dynamic balance adjustment: The algorithm processing layer of the control module performs environmental compensation on the collected ΔU:

[0078] When H>60%RH, according to the formula Calculate the surface leakage resistance by

[0079] Corrected the effect of insulation resistance on ΔU;

[0080] Combined with the temperature compensation formula Obtain the error signal e(t) after comprehensive compensation;

[0081] Use PID control algorithm to calculate the adjustment amount:

[0082] Proportion link Quickly respond to current errors;

[0083] Points Phase I Eliminate steady-state errors;

[0084] Differential link D Forecast error change trend;

[0085] Comprehensive output adjustment ;

[0086] The control output layer converts the adjustment amount into a digital potentiometer adjustment instruction or a MOSFET drive signal for the automatic adjustment circuit: if the adjustment amount is to increase the resistance, the digital potentiometer resistance value is increased with a resolution of 0.1%; if fast adjustment is required, the MOSFET of the corresponding resistance value in the automatic adjustment circuit is driven to turn on, and a parallel resistor (such as 10kΩ, 100kΩ) is connected to the variable resistor arm, so that ΔU returns to equilibrium within 10ms (ΔU≤10μV).

[0087] Signal preprocessing: Band-pass filtering and normalization are performed on ΔU, and it is then fused with T and H to form three-dimensional time series data (T, H, ΔU).

[0088] Insulation resistance detection: The preprocessed three-dimensional data (T, H, ΔU) is input into the detection model. The input layer receives three-dimensional time series data. The first convolution layer (32 filters, kernel size 5) extracts local features within 5ms (such as voltage change rate and temperature-humidity coupling trend). The maximum pooling layer (pooling size 2) reduces the dimension and retains the peak features. The second convolution layer (64 filters, kernel size 3) extracts more detailed features within 3ms (such as voltage fluctuation peak). The global average pooling layer compresses the feature map into a global vector. The fully connected layer (128 neurons) fuses high-level features. The dropout layer (dropout rate 0.3) suppresses overfitting. The output layer (1 neuron, linear activation) outputs the real-time insulation resistance detection value.

[0089] The process of adjusting the variable resistor arm is as follows:

[0090] Step 1: Receive adjustment instructions: The control module's PID algorithm calculates the required resistance adjustment based on the bridge unbalance voltage ΔU and the error signal after environmental compensation;

[0091] Step 2: Select the adjustment mode:

[0092] Fine adjustment: The resistance value is adjusted by the digital potentiometer, which is suitable for correcting small deviations;

[0093] Fast adjustment: When the ΔU deviation is large, the resistor array is activated: the control module selects the resistor combination with the closest resistance value in the resistor array based on the size of ΔR, and connects the selected resistor in parallel to the variable resistor arm by turning on the corresponding MOSFET switch, quickly changing the total resistance value;

[0094] Step 3: Perform resistance adjustment and MOSFET switch control: To increase the total resistance, disconnect some parallel resistors (turn off the corresponding MOSFET switches); to decrease the total resistance, connect more parallel resistors (turn on the corresponding MOSFET switches);

[0095] Digital potentiometer adjustment: Send instructions through the I²C / SPI interface to adjust its resistance to the target value.

[0096] Step 4: After adjustment, the voltage sensor re-detects ΔU and feeds it back to the control module. If ΔU still exceeds the threshold (e.g., >10μV), steps 1-3 are repeated until the bridge is balanced.

[0097] Through the above steps, the method achieves high-precision, real-time detection of the insulation impedance of the energy storage converter, effectively compensates for environmental interference, and ensures the reliability of the detection results under complex working conditions.

[0098] While various embodiments of the present invention have been described above, the above descriptions are intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent insulation impedance detection system for energy storage converters, characterized in that: include: Bridge balancing structure: Using a Wheatstone bridge topology, the four bridge arms form a quadrilateral loop, with two diagonal arms connected to a DC excitation power supply and a microvolt voltage detection port, respectively. The four bridge arms include at least one variable resistance arm and one measurement arm. Data acquisition module: used to collect ambient temperature, humidity and bridge output voltage in real time; Control module: connected to the data acquisition module, calculates the bridge balance deviation based on the ambient temperature, humidity and bridge output voltage, and outputs adjustment instructions; Automatic adjustment circuit: connected to the control module and the variable resistor arm respectively, responding to the adjustment instruction, controlling the variable resistor arm to adjust the resistance value to achieve dynamic balance of the bridge; An insulation impedance detection module, connected to the data acquisition module, outputs an insulation impedance detection value based on the bridge output voltage through a detection model; The control module includes a data acquisition layer, an algorithm processing layer, and a control output layer; The data acquisition layer is connected to the temperature sensor, humidity sensor and voltage sensor respectively to obtain the ambient temperature, humidity and bridge unbalance voltage ΔU; The algorithm processing layer uses the bridge unbalance voltage ΔU as the input variable, compensates for ΔU using the ambient temperature and humidity, and calculates the bridge balance adjustment using the PID control algorithm. The dynamic compensation process for ΔU by humidity is as follows: Using the parallel model Calculate the equivalent insulation resistance under the influence of humidity ,in is the surface leakage resistance, is the insulation resistance; The unbalanced voltage of the bridge is related to the impedance of the bridge arm, Replace the original Recalculate ΔU and compare it with the measured ΔU by formula Correct ΔU to offset the humidity Reduce the interference to ΔU and realize dynamic compensation of humidity to ΔU; The control output layer generates an adjustment command for the digital potentiometer based on the balance adjustment value calculated by the PID control algorithm, sends the adjustment command to the digital potentiometer via the SPI or I²C communication interface, and adjusts the variable resistor arm through the digital potentiometer to make the bridge output voltage approach zero; The algorithm processing layer is also connected to the BMS timing control module, which activates the detection circuit of each module in a time-sharing manner to avoid the parallel effect caused by the simultaneous connection of multiple modules to detect internal resistance; The insulation impedance detection module includes a signal preprocessing unit and a detection model; The signal preprocessing unit filters out the low-frequency drift and high-frequency noise of the bridge unbalanced voltage ΔU through a bandpass filter, retains the effective frequency band of the polarization process of the insulating material, performs normalization processing, and converts it into dimensionless data between 0 and 1; The construction process of the detection model is as follows: A high-precision insulation resistance meter is used to collect insulation resistance test values under different working conditions in multiple time series. The ambient temperature, humidity, and bridge unbalanced voltage ΔU are also collected simultaneously. The ambient temperature, humidity, and bridge unbalanced voltage ΔU data are organized into a three-dimensional tensor format (T, H, ΔU), where T is the ambient temperature and H is the humidity. A large number of insulation resistance test values and (T, H, ΔU) matching the insulation resistance test values are obtained and manually labeled by experts. After labeling, (T, H, ΔU) are used as the input of a convolutional neural network, and the insulation resistance test values are used as the output of the convolutional neural network for iterative training. A detection model based on the insulation resistance test values output by (T, H, ΔU) is obtained.

2. The intelligent insulation impedance detection system for energy storage converter according to claim 1, characterized in that: The four bridge arms include: The first standard resistance arm and the second standard resistance arm are both high-precision metal film resistors, connected in series between the positive electrode of the excitation power supply and the common node to form a fixed reference bridge arm; The variable resistor arm is a digital potentiometer connected in series between the common node and the negative pole of the excitation power supply; the digital potentiometer is also connected to the control module via an I²C or SPI communication interface; Measuring arm: connected to the measured circuit of the energy storage converter, located on the same side of the bridge arm as the variable resistor arm, used to access the insulation impedance to be measured.

3. The intelligent insulation impedance detection system for energy storage converter according to claim 2, characterized in that: The automatic adjustment circuit comprises: A resistor array connected in parallel at both ends of the variable resistor arm, wherein the resistance values of the resistors in the resistor array are distributed according to an exponential or linear rule to cover the target adjustment range; Each resistor in the resistor array is connected in series with a MOSFET switch.

4. The intelligent insulation impedance detection system for energy storage converter according to claim 3, characterized in that: The data acquisition module includes: Temperature sensor: used to collect the ambient temperature of the bridge in real time; Humidity sensor: used to collect environmental humidity in real time; Voltage sensor: A microvolt-level sensor based on a high-impedance differential amplifier, connected to the voltage detection port between the two midpoints of the bridge, used to measure the bridge unbalance voltage ΔU.

5. An intelligent detection method for insulation impedance of an energy storage converter, characterized in that: Applying the system described in any one of claims 1 to 4, comprising the following steps: System initialization: Start the bridge balancing structure, adjust the variable resistor arm through the digital potentiometer to make the bridge unbalance voltage ΔU ≤ 10μV, and record the initial balance parameters; Real-time data acquisition: The data acquisition module synchronously collects ΔU, T, and H to generate a raw signal sequence with a time stamp; Dynamic balance adjustment: The control module calculates the adjustment amount through the PID algorithm according to ΔU, T, and H, and drives the automatic adjustment circuit to switch the digital potentiometer or parallel resistor to make ΔU return to balance; Signal preprocessing: Band-pass filtering and normalization are performed on ΔU, and it is then fused with T and H to form three-dimensional time series data (T, H, ΔU). Insulation impedance detection: The pre-processed three-dimensional time series data (T, H, ΔU) is input into the detection model to output the real-time insulation impedance detection value.

Images