Method and system for testing and evaluating fault working condition of low-voltage intelligent circuit breaker based on deep neural network

Abstract

The invention relates to a low-voltage intelligent circuit breaker fault condition test and evaluation method and system based on a deep neural network. The method comprises the following steps: collecting real-time operation data of a low-voltage intelligent circuit breaker; performing normalization, denoising and feature extraction on the collected real-time operation data of the low-voltage intelligent circuit breaker in sequence; constructing an overcurrent control circuit for executing a fault current limiting (FCL) function; based on the extracted features, a complete protection closed loop from detection, current limiting, isolation to recovery is realized by using a constructed overcurrent control circuit for executing an FCL function; based on the fault record characteristics and the equipment operation state parameters, intelligent fault current detection is carried out by using a deep neural network method to obtain fault data so as to realize an FCL function; and designing heat capacity based on the system thermal resistance parameter and the fault current parameter, and obtaining a low-voltage intelligent circuit breaker fault condition test evaluation result based on thermal performance through thermal simulation and transient modeling. According to the invention, the performance and reliability of the system under the actual fault condition can be predicted and evaluated.

Description

Technical Field

[0001] This invention belongs to the field of power system automation technology, and relates to a method and system for testing and evaluating the fault conditions of low-voltage intelligent circuit breakers, particularly a method and system for testing and evaluating the fault conditions of low-voltage intelligent circuit breakers based on deep neural networks. Background Technology

[0002] Advances in smart grids have made low-voltage smart circuit breakers a critical component of power systems, with their intelligent performance directly impacting system reliability and power supply quality. Traditional circuit breakers typically rely on mechanical protection and manual inspection, making it difficult to quickly and accurately identify faults. To improve fault response speed and system self-healing capabilities, advanced intelligent diagnostic technologies, such as deep neural networks, must be utilized to analyze data from circuit breakers under different operating conditions.

[0003] There are two obvious problems:

[0004] 1. Inaccurate fault feature identification: Traditional methods rely on preset rules and thresholds to determine faults, which are difficult to adapt to complex power grid environments and changing fault modes.

[0005] 2. Lack of intelligent analysis: Existing fault analysis relies heavily on human experience and lacks automated analysis methods, making it unable to efficiently process large amounts of data. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a fault condition testing and evaluation system and method for low-voltage intelligent circuit breakers based on deep neural networks. This system can utilize deep neural networks to analyze the electrical characteristics and behavioral patterns of low-voltage intelligent circuit breakers under different fault conditions, and perform in-depth analysis of the circuit breaker's response data to predict and evaluate its performance and reliability under actual fault conditions.

[0007] The above-mentioned objective of this invention is achieved through the following technical solution:

[0008] A fault condition testing and evaluation method for low-voltage smart circuit breakers based on deep neural networks includes the following steps:

[0009] Collect real-time operating data of low-voltage intelligent circuit breakers;

[0010] The real-time operation data of the collected low-voltage intelligent circuit breakers are normalized, denoised, and feature extracted sequentially.

[0011] Construct an overcurrent control circuit that performs the FCL function;

[0012] Based on the extracted features, an overcurrent control circuit that performs the FCL function is constructed to achieve a complete protection closed loop from detection, current limiting, isolation to recovery.

[0013] Based on fault record features and equipment operating status parameters, a deep neural network method is used to perform intelligent fault current detection to obtain fault data and realize the FCL function.

[0014] Based on the system thermal resistance parameters and fault current parameters, and the design thermal capacity, the test evaluation results of the fault condition of the low-voltage intelligent circuit breaker based on thermal performance are obtained through thermal simulation and transient modeling.

[0015] Furthermore, the specific method for collecting real-time operating data of the low-voltage intelligent circuit breaker is as follows:

[0016] Deploy high-precision sensors in a 1kV DC power distribution system to continuously monitor and collect voltage and current waveform data of the system output, especially transient data when a fault occurs.

[0017] Furthermore, the specific steps for normalizing, denoising, and extracting features from the collected real-time operating data of the low-voltage intelligent circuit breaker include:

[0018] The raw data in S1 is processed using Hilbert transform to accurately extract instantaneous amplitude, phase, and frequency information that can characterize fault features;

[0019] Normalization is performed on the original data to eliminate the influence of dimensions, and techniques such as wavelet transform are used for noise reduction to improve data quality;

[0020] These standardized data, along with features extracted from the Hilbert transform, are then fed into a deep neural network for training.

[0021] Furthermore, the specific method for constructing the overcurrent control circuit that performs the FCL function is as follows:

[0022] Taking the "pole-to-pole fault" occurring at a 1kV DC load center as an example for analysis, a second-order RLC equivalent circuit model is established, and Kirchhoff's voltage law is used for mathematical derivation:

[0023] The state equations for inductor current and capacitor voltage are as follows:

[0024]

[0025] The relevant equation for fault current is:

[0026]

[0027] in:

[0028]

[0029] Considering the buck voltage effect, the fault current of the buck in a second-order underdamped RLC circuit can be estimated using Kirchhoff's voltage law as follows:

[0030]

[0031] Divide the above equation by Lbus and transform it to the s-domain:

[0032]

[0033] Performing a time-domain transform on the above equation, we get:

[0034]

[0035] Among them, i Lbus Represents fault current, V Cbuck Represents capacitor voltage, C buck For voltage drop capacitor, C F Represents energy storage capacitor, L bus Represents faulty inductor, i Leq For steady-state current, L eq For steady-state inductance, R eq steady-state resistance, V CF Step-down voltage.

[0036] Furthermore, the specific steps for implementing a complete protection closed loop from detection, current limiting, isolation to recovery using the constructed overcurrent control circuit that performs the FCL function based on the extracted features include:

[0037] System initialization: Close SSCB to enable normal system operation and begin continuous measurement of line current;

[0038] The measured current value is compared with the fault current threshold; if the measured current is less than this threshold, the system is determined to be in a normal state, the "operation counter" is reset, and the system returns to initialization to continue monitoring.

[0039] Once the measured current value is determined to exceed the fault threshold, the control system takes into account the inherent "control delay" of the sensor and signal processing links, and then activates the overcurrent control circuit that performs the FCL function, providing fault protection and implementing the SSCB control scheme for the FCL function.

[0040] After the FCL operation is executed, the SSCB will open to isolate the fault and wait for a preset "switching cycle" to complete. After the switching cycle ends, the system will increment the value of the "operation counter" and check whether it has reached the predefined maximum number of operations. If it has not reached the maximum number of operations, the system will automatically close the SSCB again and return to the system initialization restart monitoring process to attempt to restore power. If the counter has reached the maximum value, it indicates that the fault is permanent, and the system will command the SSCB to open permanently and issue an alarm.

[0041] Furthermore, the specific steps for achieving the FCL function by using a deep neural network method to perform intelligent fault current detection based on fault record features and equipment operating status parameters to obtain fault data include:

[0042] The system acquires in real time the four key electrical quantities most relevant to SSCB operation: overcurrent value, current change rate, undervoltage value, and apparent resistance change.

[0043] By utilizing advanced signal processing techniques such as wavelet transform, the original measurement signal is decomposed and reconstructed to extract deeper time-frequency domain features that better characterize the nature of the fault, thereby constructing a complete classification database for training and testing.

[0044] A deep neural network is introduced as a classifier. After training, when real-time data is input into the DNN, a highly reliable classification result is output in a very short time. The intelligent judgment result is transmitted to step 4 in real time as the basis for judging whether the measured current is greater than the fault current threshold in its decision-making process, so as to realize the FCL function.

[0045] Furthermore, the specific method for evaluating the fault condition test results of low-voltage intelligent circuit breakers based on thermal performance, using system thermal resistance parameters, fault current parameters, and design heat capacity through thermal simulation and transient modeling, is as follows:

[0046] By using the case temperature, junction temperature, and thermal impedance data provided in the device datasheet, the maximum power that can be dissipated through each RB-IGCT can be estimated:

[0047]

[0048] Where P(AV)M is the maximum dissipable power, and T vj.max and T c These are the highest junction temperature and case temperature obtained from the device datasheet, R. th(j-c) Thermal resistance of the device from junction to case

[0049] In a DC system, the average current and the root-mean-square current are equal, denoted by I. dc express;

[0050] Therefore Idc The solution is:

[0051]

[0052] Where V T0 For the threshold voltage, r T For the device resistance obtained from the device datasheet

[0053] Estimate the required number of parallel RB-IGCT units by considering reliability and safe operating area constraints; as a rule of thumb, the current capacity of the equipment should be twice the maximum DC current.

[0054]

[0055] Calculate the current of the device under rated operating conditions:

[0056]

[0057] A fault condition testing and evaluation system for low-voltage intelligent circuit breakers based on deep neural networks, comprising:

[0058] The data acquisition module collects real-time operating data from the low-voltage intelligent circuit breaker.

[0059] The preprocessing module performs normalization, noise reduction, and feature extraction on the real-time operation data of the collected low-voltage intelligent circuit breakers in sequence.

[0060] Overcurrent control circuit construction module, which constructs an overcurrent control circuit that performs the FCL function;

[0061] The SSCB control scheme formation module, based on the extracted features, utilizes the constructed overcurrent control circuit that executes the FCL function to realize a complete fault protection closed loop for the solid-state circuit breaker SSCB from detection, current limiting, isolation to recovery;

[0062] The FCL function implementation module uses a deep neural network method to perform intelligent fault current detection based on fault record characteristics and equipment operating status parameters, in order to obtain fault data and realize the FCL function.

[0063] The fault condition test and evaluation module, based on system thermal resistance parameters, fault current parameters, and design heat capacity, obtains fault condition test and evaluation results of low-voltage intelligent circuit breakers based on thermal performance through thermal simulation and transient modeling.

[0064] Furthermore, the acquisition module is also used for:

[0065] Deploy high-precision sensors in a 1kV DC power distribution system to continuously monitor and collect voltage and current waveform data of the system output, especially transient data when a fault occurs.

[0066] A computer-readable storage medium having program instructions stored thereon, which, when executed by a processor, implement a method for testing and evaluating fault conditions of a low-voltage smart circuit breaker based on a deep neural network.

[0067] The advantages and positive effects of this invention are as follows:

[0068] This invention proposes a fault condition testing and evaluation system and method for low-voltage intelligent circuit breakers based on deep neural networks, which is used to evaluate and test the performance of low-voltage intelligent circuit breakers under different fault conditions. Attached Figure Description

[0069] Figure 1 The DC microgrid of the present invention is represented as a schematic diagram of a low-voltage DC power distribution system;

[0070] Figure 2 A schematic diagram of the single busbar single circuit breaker configuration of the present invention for a 1kV DC region;

[0071] Figure 3 This is the equivalent circuit diagram of the rotating low-voltage DC power distribution system under DC load failure according to the present invention;

[0072] Figure 4 This is a flowchart illustrating the implementation of the whole-pack system with fault current control according to the present invention. Detailed Implementation

[0073] The structure of the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that these embodiments are descriptive and not limiting.

[0074] A fault condition testing and evaluation method for low-voltage smart circuit breakers based on deep neural networks includes the following steps:

[0075] Step 1: Collect real-time operation data of the low-voltage intelligent circuit breaker;

[0076] The specific method for step 1 is as follows:

[0077] Deploy high-precision sensors in a 1kV DC power distribution system to continuously monitor and collect voltage and current waveform data of the system output, especially transient data when a fault occurs.

[0078] In this embodiment, step 1 is the foundation for the system to achieve intelligent diagnosis and protection, and its core task is to comprehensively collect real-time operating data of the low-voltage intelligent circuit breaker. Furthermore, step 1 is directly linked to step 5, providing analytical material for both.

[0079] Step 2: Normalize, denoise, and extract features from the real-time operation data of the low-voltage intelligent circuit breaker collected in Step 1;

[0080] The specific steps of step 2 include:

[0081] (1) The raw data in S1 is processed by Hilbert transform to accurately extract instantaneous amplitude, phase and frequency information that can characterize the fault features;

[0082] (2) Normalize the original data to eliminate the influence of dimensions, and use wavelet transform and other techniques to remove noise and improve data quality.

[0083] (3) Input these standardized data, along with the features extracted from the Hilbert transform, into a deep neural network for training.

[0084] In this embodiment, steps 2 and 4 are closely linked, providing a precise basis for action judgment.

[0085] Step 3: Construct an overcurrent control circuit that performs the FCL function;

[0086] The specific method for step 3 is as follows:

[0087] Taking the "pole-to-pole fault" occurring at a 1kV DC load center as an example for analysis, a second-order RLC equivalent circuit model is established, and Kirchhoff's voltage law is used for mathematical derivation:

[0088] The state equations for inductor current and capacitor voltage are as follows:

[0089]

[0090] The relevant equation for fault current is:

[0091]

[0092] in:

[0093]

[0094] Considering the buck voltage effect, the fault current of the buck in a second-order underdamped RLC circuit can be estimated using Kirchhoff's voltage law as follows:

[0095]

[0096] Divide the above equation by Lbus and transform it to the s-domain:

[0097]

[0098] Performing a time-domain transform on the above equation, we get:

[0099]

[0100] Among them, i LbusRepresents fault current, V Cbuck Represents capacitor voltage, C buck For voltage drop capacitor, C F Represents energy storage capacitor, L bus Represents faulty inductor, i Leq For steady-state current, L eq For steady-state inductance, R eq steady-state resistance, V CF Step-down voltage;

[0101] In this embodiment, the overcurrent control circuit constructed in step 3 to perform the FCL function can accurately calculate the peak fault current contributed by the step-down capacitor and the energy storage capacitor. This overcurrent control circuit is constructed using a solid-state circuit breaker with a reverse-blocking integrated gated thyristor as its core. When the deep neural network in step 2 diagnoses the fault and issues a command, this circuit can be triggered in a very short time, quickly engaging the current-limiting impedance to forcibly limit the fault current within a safe range. The design of step 3 and step 6 must be considered in conjunction to ensure that the heat generated by the RB-IGCT during FCL function execution is effectively dissipated, and the junction temperature does not exceed the safe limit, thereby ensuring the reliability and lifespan of the protection device itself during operation.

[0102] Step 4: Based on the features extracted in Step 2, use the overcurrent control circuit that performs the FCL function constructed in Step 3 to provide a control scheme for SSCB that provides fault protection and implements the FCL function.

[0103] The specific steps of step 4 include:

[0104] (1) System initialization: Close SSCB to enable normal system operation and start continuous measurement of line current;

[0105] (2) Compare the measured current value with the more accurate fault current threshold dynamically provided by step 5; if the measured current is less than this threshold, the system is determined to be in a normal state, and after resetting the "operation counter", it returns to step (1) to continue monitoring.

[0106] (3) Once the measured current value is determined by step 5 to exceed the fault threshold, the control system takes into account the inherent “control delay” of the sensor and signal processing links, and then starts the overcurrent control circuit that performs the FCL function constructed in step 3, providing fault protection and implementing the SSCB control scheme of the FCL function.

[0107] (4) After the FCL operation is executed, the SSCB will open to isolate the fault and wait for a preset "switching cycle" to complete. After the switching cycle ends, the system will increase the value of the "operation counter" and check whether it has reached the predefined maximum number of operations. If it has not reached the maximum number of operations, the system will automatically close the SSCB again and return to step (1) to restart the monitoring process and try to restore power. If the counter has reached the maximum value, it indicates that the fault is permanent. The system will command the SSCB to open permanently and issue an alarm.

[0108] First, the solid-state circuit breaker (SSCB) is closed to put the system into normal operation, and the line current is continuously monitored and its rate of change is calculated in real time using a high-precision current sensor. Then, the measured current value and its rate of change are compared with the adaptive fault threshold dynamically generated by the deep neural network in step 5. If an abnormal current is detected but the fault threshold is not reached, the system only activates the warning mode and records the abnormal characteristics without triggering protection actions. Once the deep neural network determines that the current characteristics exceed the fault threshold, the control system, after comprehensively considering the inherent control delays of sensor sampling, signal processing, and communication links, immediately sends a trigger command to the overcurrent control circuit constructed in step 3. Upon responding to the command, the circuit drives... The active module enables the RB-IGCT to conduct rapidly within 40μs, applying the preset current-limiting impedance, thereby effectively suppressing the rise of fault current and limiting its peak value to a safe range. After completing the current limiting operation, the control system disconnects the SSCB to isolate the fault area and simultaneously initiates the intelligent reclosing logic. After experiencing a preset arc-suppression cycle, the SSCB is automatically reclosed. The system's built-in operation counter accumulates and records the number of reclosing operations. If the fault characteristics disappear within the preset maximum number of operations, the system returns to normal operation. Otherwise, it is determined to be a permanent fault, and the system keeps the SSCB in the open state and issues an alarm signal, thus realizing a complete protection closed loop from detection, current limiting, isolation to recovery.

[0109] In this embodiment, a crucial anti-maloperation mechanism effectively distinguishes between normal fluctuations and genuine faults. Once the measured current is determined to exceed the fault threshold, the control system does not immediately trip permanently. Instead, it considers the inherent "control delay" in sensor and signal processing components and then initiates FCL operation. The core of this operation is that the SSCB quickly engages the current-limiting impedance to suppress the sharp rise in fault current. After the FCL operation is executed, the SSCB opens to isolate the fault and waits for a preset "switching cycle" to complete. This is intended to give transient faults a chance to resolve themselves. After the switching cycle ends, the system increments the value of the "operation counter" and checks whether it has reached the predefined maximum number of operations. If not, the system automatically recloses the SSCB and returns to the restart monitoring process to attempt to restore power. If the counter has reached its maximum value, it indicates that the fault is permanent. The system will command the SSCB to open permanently and issue an alarm, thereby preventing equipment damage due to repeated attempts to connect. Ultimately, closed-loop control from fault detection and intelligent decision-making to physical action is achieved, and the recursive trial mechanism significantly improves the resilience of power supply.

[0110] Step 5: Implement an intelligent fault current detection method using deep neural network methods to achieve FCL function;

[0111] The specific steps of step 5 include:

[0112] (1) The system acquires in real time the four key electrical quantities most relevant to SSCB operation: overcurrent value, current change rate (di / dt), undervoltage value, and apparent resistance change.

[0113] (2) Using the advanced signal processing technology of wavelet transform in step 2, the original measurement signal is decomposed and reconstructed, and the deep time-frequency domain features that can better characterize the nature of the fault are extracted, thereby constructing a complete classification database for training and testing.

[0114] (3) A deep neural network is introduced as a classifier. After training, when real-time data is input into the DNN, a highly reliable classification result is output in a very short time. The intelligent judgment result is transmitted to step 4 in real time as the basis for judging whether the measured current is greater than the fault current threshold in its decision-making process, so as to realize the FCL function.

[0115] A deep neural network method is used to implement an intelligent fault current detection method to achieve the FCL function. The input data for this step are the raw current waveform time series data from a high-precision current sensor, system voltage signal, historical fault record characteristics, and equipment operating status parameters. The output data are real-time fault type classification results, fault severity assessment, dynamic action threshold suggestions, and predictive trend signals of fault development.

[0116] This step aims to ensure that the transient heat generated by the RB-IGCT when limiting fault current is effectively dissipated, maintaining the junction temperature within a safe range. Input data includes: the RB-IGCT's transient power loss curve, thermal resistance parameters, heat capacity of the cooling system, fault current parameters, ambient temperature, and maximum allowable junction temperature. Through thermal simulation and transient modeling, the output results are: the minimum heat capacity required for the cooling system, the real-time junction temperature change curve, heat distribution characteristics, and a circuit breaker reliability assessment conclusion based on thermal performance, thereby ensuring the stable operation of the device under repeated faults.

[0117] In this embodiment, the deep neural network learns from this vast feature database and can autonomously discover complex, nonlinear fault modes. After training, when real-time data is input into the DNN, it can output a highly reliable classification result in a very short time. This intelligent judgment result is transmitted in real time to step S4 as an intelligent upgrade to the key criterion "whether the measured current is greater than the fault current threshold" in its decision-making process. In this way, step S5 not only provides step S4 with a more accurate and faster fault detection capability, but also, due to its powerful pattern recognition capability, fundamentally reduces the risk of system malfunction, ensuring that the FCL function can be activated promptly and accurately when it is truly needed.

[0118] The entropy of the measured data is set as:

[0119]

[0120] The probability of the selected category i is expressed as P(y i ) represents the remaining classifier attributes. N represents the gain of attribute a with respect to dataset S as .

[0121]

[0122] Where Value(A) represents each value of attribute A. S(v) represents a subset of A. This is a variant of the C4.5 method DT. C4.5 selects attributes based on gain ratio information. Decisions are pruned to avoid data overfitting. Data partitioning can be represented as:

[0123]

[0124] For a, if the sample S is divided into v values, the gain can be calculated as follows:

[0125]

[0126] Step 6: Design the heat capacity to ensure that the heat generated by the RB-IGCT when performing the FCL function can be effectively dissipated and the junction temperature does not exceed the safety limit, thereby completing the reliability assessment of the circuit breaker under extreme fault conditions.

[0127] The specific method for step 6 is as follows:

[0128] By using the case temperature, junction temperature, and thermal impedance data provided in the device datasheet, the maximum power that can be dissipated through each RB-IGCT can be estimated:

[0129]

[0130] Where P(AV)M is the maximum dissipable power, and T vj.max and T c These are the highest junction temperature and case temperature obtained from the device datasheet, R. th(j-c) Thermal resistance of the device from junction to case

[0131] In a DC system, the average current and the root-mean-square current are equal, denoted by I. dc express;

[0132] Therefore I dc The solution is:

[0133]

[0134] Where V T0 For the threshold voltage, r T For the device resistance obtained from the device datasheet

[0135] Estimate the required number of parallel RB-IGCT units by considering reliability and safe operating area constraints; as a rule of thumb, the current capacity of the equipment should be twice the maximum DC current.

[0136]

[0137] Calculate the current of the device under rated operating conditions:

[0138]

[0139] In this embodiment, the thermal tolerance of the device itself is accurately assessed, the inherent characteristics of power semiconductor devices are analyzed in depth, and the maximum heat that it can continuously withstand under safe conditions and its ability to dissipate heat from the internal core to the external casing are determined.

[0140] Following this temperature threshold, the critical phase of component selection and system matching begins. The analysis in step S3 reveals the staggering current flowing through the solid-state circuit breaker under extreme fault conditions. The task of thermal design at this point is to calculate, based on this current stress, how many power semiconductor units need to operate in parallel to distribute the total current and resulting heat across each unit, ensuring that even under the most severe conditions, the heat borne by each unit is far below its maximum withstand limit. Determining this number of parallel units represents a direct dialogue and balance between electrical requirements and thermodynamic safety.

[0141] Finally, a solid-state circuit breaker prototype with a heat dissipation system needs to be built and tested under the simulated fault current limiting operation scenario described in step S4. High-precision temperature sensing technology is used to monitor the real-time temperature changes of the power semiconductor device during a series of high-intensity actions, including rapid switching and applying current-limiting impedance. This measured temperature data is the final criterion for verifying its thermal reliability. If the temperature exceeds the safe range, it indicates a deficiency in the thermal design, requiring a closed-loop feedback loop: this result will directly propagate in reverse, requiring step S3 to re-examine the strength of its fault current limiting strategy, or requiring step S4 to optimize its control logic, such as adjusting the duration of the fault current limiting or the allowed number of reclosing operations, to avoid cumulative overheating of the device. Through this close collaboration and iteration across steps, it is ultimately ensured that while providing robust protection, the solid-state circuit breaker's lifeline—thermal reliability—is also solidly guaranteed.

[0142] By using the case temperature, junction temperature, and thermal impedance data provided in the device datasheet, the maximum power that can be dissipated through each RB-IGCT can be estimated:

[0143]

[0144] Where P(AV)M is the maximum dissipable power, and T vj.max and T c These are the highest junction temperature and case temperature obtained from the device datasheet, R. th(j-c) Thermal resistance of the device from junction to case

[0145] In a DC system, the average current and the root-mean-square current are equal, denoted by I. dc It means. Therefore, I dc It can be interpreted as:

[0146]

[0147] Where V T0 For the threshold voltage, r T For the device resistance obtained from the device datasheet

[0148] Estimate the required number of parallel rb-igct units by considering reliability and safe operating area constraints. As a rule of thumb, the current capacity of the equipment should be twice the maximum DC current.

[0149]

[0150] Calculate the current of the device under rated operating conditions:

[0151]

[0152] This invention also provides a method for intelligent short-circuit protection of low-voltage DC microgrids.

[0153] Taking a 1kV DC distribution system as an example, numerical analysis was performed on the established method. Throughout the research, we found that using RB-IGCT to implement SSCB provides effective short-circuit capability, forward and reverse voltage blocking, low on-state voltage drop, and low thermal resistance.

[0154] A short-circuit protection based on rb-igctsscb is adopted, with an on-time of 40μs. To overcome the control delay caused by sensors and measurement units, a fault detection method based on decision trees is employed, drawing upon the development of FCL functions. The developed FDA training accuracy is 98.3%, and the testing time is 0.02μs. Furthermore, the developed protection method and implemented FCL ensure continuous power flow in the DC region by disconnecting only the faulty portion of the DC microgrid.

[0155] The working principle of this invention is:

[0156] (1) Short-circuit protection basis for low-voltage DC microgrids

[0157] In the initial stages of this invention, we delve into the short-circuit protection problem of low-voltage DC microgrids. We first construct a 1kV DC distribution system model, powered by a power converter and connected to multiple loads. Detailed listing of system specifications provides a solid foundation for our subsequent analysis and simulations. The system components include a post-regulated isolated DC-to-DC converter, DC and AC loads, and DC cables; detailed descriptions of these components help us understand the system's behavior under different operating conditions. Numerical simulations performed in the MATLAB / Simulink environment allow us to verify the effectiveness and reliability of the system design. Furthermore, we model solid-state circuit breakers (SSCBs), particularly reverse-blocking integrated gated thyristors (RB-IGCTs), which provides crucial technical support for subsequent short-circuit protection strategies. We emphasize the importance of SSCBs in protecting DC microgrids and provide simulation details of the RB-IGCT model, providing a theoretical basis for achieving accurate short-circuit protection. Finally, we propose a short-circuit protection strategy, including the implementation of fault current limiting (FCL) functionality, which aims to quickly isolate fault areas, limit the impact range, and reduce safety risks. The proposed control scheme, including fault current measurement, fault detection logic, and FCL operation execution, lays a practical foundation for subsequent research.

[0158] (2) Objectives, challenges and control schemes of short circuit protection

[0159] The objectives and challenges of short-circuit protection are further analyzed. The complexity of implementing short-circuit protection in DC microgrids is discussed in detail, particularly the potential for maloperation due to initial current in the capacitor filter during DC operation. Selecting a suitable fault detection threshold is challenging, as downstream SSCBs typically require lower thresholds than upstream SSCBs. Our proposed control scheme utilizes the SSCB for fault protection and implements the fault current cascading (FCL) function. This scheme details the process of measuring current during normal operation, executing FCL when the current exceeds the fault current threshold, and restoring normal operation after fault clearance. This scheme not only addresses the controller delay issue in practical systems but also provides an operational framework for subsequent intelligent fault current detection methods. Furthermore, we introduce a deep neural network-based intelligent fault current detection method to improve the accuracy and response speed of the FCL function. By training a deep neural network to identify fault characteristics and evaluating the effectiveness of the FCL method, we provide a new perspective for achieving more efficient short-circuit protection. Finally, we discuss the impact of FCL operation on the thermal treatment requirements of the RB-IGCT, ensuring that the temperature limits of the equipment do not exceed safe ranges under short-circuit conditions. This discussion provides a consideration of thermal stability for subsequent deep neural network methods.

[0160] (3) Application of deep neural networks in intelligent fault current detection

[0161] This invention proposes a decision tree-based intelligent fault current detection method. This method further improves the accuracy and response speed of short-circuit protection by measuring and processing parameters such as overcurrent, current derivative, undervoltage, and apparent resistance changes. Experimental results and analysis show that the accuracy of this technology reaches 98.3% during the training phase, with a testing time of only 0.02 microseconds, significantly improving the efficiency of short-circuit protection. By comparing the allowance energy for different fault detection thresholds, the effectiveness of the proposed FCL method in reducing fault current is verified, and it exhibits faster detection time and lower classification error rate compared to existing technologies. Furthermore, by calculating and comparing power loss and junction temperature under different operating conditions, the thermal stability of the RB-IGCT in SSCB applications is verified. The results show that even under short-circuit conditions, the RB-IGCT can operate without exceeding the maximum allowable junction temperature, demonstrating its applicability as an SSCB in DC microgrids. This finding not only verifies the effectiveness of deep neural network methods but also provides a new direction for future short-circuit protection technologies.

[0162] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks, characterized in that: Includes the following steps: Collect real-time operating data of low-voltage intelligent circuit breakers; The real-time operation data of the collected low-voltage intelligent circuit breakers are normalized, denoised, and feature extracted sequentially. Construct an overcurrent control circuit that performs fault current limiting (FCL) function; Based on the extracted features, an overcurrent control circuit that performs fault current limiting (FCL) function is constructed to achieve a complete protection closed loop from detection, current limiting, isolation to recovery. Based on fault record features and equipment operating status parameters, a deep neural network method is used to perform intelligent fault current detection to obtain fault data and realize the FCL function. Based on the system thermal resistance parameters and fault current parameters, and the design thermal capacity, the test evaluation results of the fault condition of the low-voltage intelligent circuit breaker based on thermal performance are obtained through thermal simulation and transient modeling.

2. The method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks according to claim 1, characterized in that: The specific method for collecting real-time operational data of low-voltage intelligent circuit breakers is as follows: Deploy high-precision sensors in a 1kV DC power distribution system to continuously monitor and collect voltage and current waveform data of the system output, especially transient data when a fault occurs.

3. The method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks according to claim 1, characterized in that: The specific steps for normalizing, denoising, and extracting features from the collected real-time operation data of the low-voltage intelligent circuit breaker include: The raw data in S1 is processed using Hilbert transform to accurately extract instantaneous amplitude, phase, and frequency information that can characterize fault features; Normalization is performed on the original data to eliminate the influence of dimensions, and techniques such as wavelet transform are used for noise reduction to improve data quality; These standardized data, along with features extracted from the Hilbert transform, are then fed into a deep neural network for training.

4. The method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks according to claim 1, characterized in that: The specific method for constructing the overcurrent control circuit that performs the FCL function is as follows: Taking a "pole-to-pole fault" occurring at a 1kV DC load center as an example, this paper analyzes the problem by establishing a second-order RLC equivalent circuit model and applying Kirchhoff's voltage law for mathematical derivation. The state equations for inductor current and capacitor voltage are as follows: The relevant equation for fault current is: -2(C F L bus -C buck L bus )f(t) -C F R eq -C F L bus L eq +C buck L bus L eq in: k＝1,2,3,4 Considering the buck voltage effect, the fault current of the buck in a second-order underdamped RLC circuit can be estimated using Kirchhoff's voltage law as follows: Divide the above equation by Lbus and transform it to the s-domain: Performing a time-domain transform on the above equation, we get: Among them, i Lbus Represents fault current, V Cbuck Represents capacitor voltage, C buck For voltage drop capacitor, C F Represents energy storage capacitor, L bus Represents faulty inductor, i Leq For steady-state current, L eq For steady-state inductance, R eq steady-state resistance, V CF Step-down voltage.

5. The method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks according to claim 1, characterized in that: The specific steps for implementing a complete protection closed loop from detection, current limiting, isolation to recovery using the constructed overcurrent control circuit that performs the FCL function based on the extracted features include: System initialization: Close SSCB to enable normal system operation and begin continuous measurement of line current; The measured current value is compared with the fault current threshold; if the measured current is less than this threshold, the system is determined to be in a normal state, the "operation counter" is reset, and the system returns to initialization to continue monitoring. Once the measured current value is determined to exceed the fault threshold, the control system takes into account the inherent "control delay" of the sensor and signal processing links, and then activates the overcurrent control circuit that performs the FCL function, providing fault protection and implementing the SSCB control scheme for the FCL function; After the FCL operation is executed, the SSCB will open to isolate the fault and wait for a preset "switching cycle" to complete. After the switching cycle ends, the system will increment the value of the "operation counter" and check whether it has reached the predefined maximum number of operations. If it has not reached the maximum number of operations, the system will automatically close the SSCB again and return to the system initialization restart monitoring process to attempt to restore power. If the counter has reached the maximum value, it indicates that the fault is permanent, and the system will command the SSCB to open permanently and issue an alarm.

6. The method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks according to claim 1, characterized in that: The specific steps for achieving the FCL function by using a deep neural network method to perform intelligent fault current detection based on fault record features and equipment operating status parameters to obtain fault data include: The system acquires in real time the four key electrical quantities most relevant to SSCB operation: overcurrent value, current change rate, undervoltage value, and apparent resistance change. By utilizing advanced signal processing techniques such as wavelet transform, the original measurement signal is decomposed and reconstructed to extract deeper time-frequency domain features that better characterize the nature of the fault, thereby constructing a complete classification database for training and testing. A deep neural network is introduced as a classifier. After training, when real-time data is input into the DNN, a highly reliable classification result is output in a very short time. The intelligent judgment result is transmitted to step 4 in real time as the basis for judging whether the measured current is greater than the fault current threshold in its decision-making process, so as to realize the FCL function.

7. The method for testing and evaluating fault conditions of low-voltage intelligent circuit breakers based on deep neural networks according to claim 1, characterized in that: The specific method for evaluating the fault condition test results of low-voltage intelligent circuit breakers based on system thermal resistance parameters, fault current parameters, and design heat capacity, through thermal simulation and transient modeling, is as follows: By using the case temperature, junction temperature, and thermal impedance data provided in the device datasheet, the maximum power that can be dissipated through each RB-IGCT can be estimated: Where P(AV)M is the maximum dissipable power, and T vj.max and T c These are the highest junction temperature and case temperature obtained from the device datasheet, R. th(j-c) Thermal resistance of the device from junction to case In a DC system, the average current and the root-mean-square current are equal, denoted by I. dc express; Therefore I dc The solution is: Where V T0 For the threshold voltage, r T For the device resistance obtained from the device datasheet Estimate the required number of parallel RB-IGCT units by considering reliability and safe operating area constraints; as a rule of thumb, the current capacity of the equipment should be twice the maximum DC current. Calculate the current of the device under rated operating conditions:

8. A fault condition testing and evaluation system for low-voltage intelligent circuit breakers based on deep neural networks, characterized in that: include: The data acquisition module collects real-time operating data from the low-voltage intelligent circuit breaker. The preprocessing module performs normalization, noise reduction, and feature extraction on the real-time operation data of the collected low-voltage intelligent circuit breakers in sequence. Overcurrent control circuit construction module, which constructs an overcurrent control circuit that performs the FCL function; The SSCB control scheme formation module, based on the extracted features, utilizes the constructed overcurrent control circuit that executes the FCL function to realize a complete fault protection closed loop for the solid-state circuit breaker SSCB from detection, current limiting, isolation to recovery; The FCL function implementation module uses a deep neural network method to perform intelligent fault current detection based on fault record characteristics and equipment operating status parameters, in order to obtain fault data and realize the FCL function. The fault condition test and evaluation module, based on system thermal resistance parameters, fault current parameters, and design heat capacity, obtains fault condition test and evaluation results of low-voltage intelligent circuit breakers based on thermal performance through thermal simulation and transient modeling.

9. The fault condition testing and evaluation system for low-voltage intelligent circuit breakers based on deep neural networks according to claim 8, characterized in that: The acquisition module is also used for: Deploy high-precision sensors in a 1kV DC power distribution system to continuously monitor and collect voltage and current waveform data of the system output, especially transient data when a fault occurs.

10. A computer-readable storage medium having stored thereon program instructions that, when executed by a processor, implement the fault condition testing and evaluation method for a low-voltage intelligent circuit breaker based on a deep neural network as described in any one of claims 1 to 7.

Images