Control method and system of mcb and rcd collaborative distribution box based on intelligent algorithm

Abstract

The present invention relates to a control method and system for a coordinated MCB and RCD distribution box based on intelligent algorithms, belonging to the field of low-voltage power distribution protection and intelligent fault diagnosis technology. The method includes: S1, acquiring a discrete-time current sequence; S2, processing the discrete-time current sequence using continuous wavelet transform to generate a time-frequency coefficient matrix; S3, calculating the high-frequency energy integral based on the time-frequency coefficient matrix; S4, calculating the time-frequency distribution entropy based on the time-frequency coefficient matrix; S5, calculating the waveform randomness factor based on the discrete-time current sequence; and S6, combining the high-frequency energy integral, time-frequency distribution entropy, and waveform randomness factor to generate an electrical disturbance feature signature through a nonlinear fusion model. This invention significantly reduces the maloperation rate of protection devices.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution protection and intelligent fault diagnosis technology, specifically to a control method and system for MCB and RCD collaborative distribution boxes based on intelligent algorithms. Background Technology

[0002] In low-voltage power distribution systems, traditional line protection devices, such as miniature circuit breakers (MCBs) and residual current devices (RCDs), rely primarily on monitoring a single physical quantity, the effective value of the current or the residual current, for their protection logic. These devices typically use fixed, pre-set operating thresholds for judgment.

[0003] This traditional protection method has inherent limitations. On the one hand, the surge current generated when a motor starts, a frequency converter runs, or a high-power device is switched on may exceed the preset threshold for a short period of time. These surge currents are essentially benign, transient operational disturbances, but traditional protection devices may trip unnecessarily due to their inability to effectively identify their transient characteristics, i.e., false protection, which seriously affects the continuity and reliability of power supply.

[0004] On the other hand, for series or parallel arc faults caused by aging insulation or loose connections, especially in their early stages, the effective value increment of the fault current may not be significant enough to trigger the operating threshold of traditional protection devices. However, such faults pose a significant fire hazard. Traditional methods, focusing only on the current amplitude and ignoring the unique physical characteristics of arc faults in terms of high frequency and randomness, lead to the risk of incomplete protection, posing a serious threat to life and property safety.

[0005] The root cause of the above limitations is that traditional protection technologies lack the ability to perform in-depth analysis of current signals and cannot comprehensively reveal the inherent physical nature of electrical events from multiple dimensions such as energy, time-frequency distribution, and waveform randomness. Therefore, it is difficult to accurately distinguish between dangerous malignant faults and benign operational disturbances in complex power environments.

[0006] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a control method and system for MCB and RCD collaborative distribution boxes based on intelligent algorithms, so as to solve the problems mentioned in the background art.

[0008] The technical solution of the present invention includes: S1, obtaining a discrete-time current sequence;

[0009] S2. Based on the discrete-time current sequence, continuous wavelet transform is applied to generate a time-frequency coefficient matrix.

[0010] S3. Calculate the high-frequency energy integral based on the time-frequency coefficient matrix;

[0011] S4. Calculate the time-frequency distribution entropy based on the time-frequency coefficient matrix;

[0012] S5. Calculate the waveform randomness factor based on the discrete-time current sequence;

[0013] S6. By combining high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor, an electrical disturbance feature signature is generated through a nonlinear fusion model.

[0014] S7. Dynamically generate adaptive tolerance threshold based on discrete-time current sequence;

[0015] S8. If the duration of the electrical disturbance signature being greater than the adaptive tolerance threshold exceeds the preset confirmation window, a synchronous trip command is output to the MCB and RCD; if the electrical disturbance signature is less than or equal to the adaptive tolerance threshold, the current operating state is maintained.

[0016] Preferably, the electrical disturbance feature signature is generated through a nonlinear fusion model, including:

[0017] The maximum-minimum normalization function is used to process the high-frequency energy integral, time-frequency distribution entropy, and waveform randomness factor to generate normalized feature parameters. These parameters are then fused using a nonlinear multiplication-gated model: the normalized time-frequency distribution entropy and waveform randomness factor are weighted and summed, and then multiplied by the normalized high-frequency energy integral as a gating factor to generate the final feature signature.

[0018] Preferably, the boundary values ​​required for the maximum-minimum normalization function are determined by statistical analysis of a pre-labeled electrical event waveform database; the weight coefficients are determined by optimization training using a supervised machine learning algorithm on the electrical event waveform database.

[0019] Preferably, the time-frequency distribution entropy is obtained by calculating the Shannon entropy of the time-frequency energy distribution.

[0020] Preferably, the waveform randomness factor is obtained by calculating the variance of the residual signal formed by the difference between the discrete-time current sequence and the power frequency fundamental component.

[0021] Preferably, dynamically generating the adaptive tolerance threshold includes:

[0022] Set a basic protection threshold; calculate the short-term and long-term exponential moving averages of the line current RMS value in parallel; calculate the normalized difference between the short-term and long-term exponential moving averages; combine the basic protection threshold and the normalized difference, and introduce an exponential decay factor to generate an adaptive tolerance threshold.

[0023] Preferably, an exponential decay factor is introduced to allow the adaptive tolerance threshold to recover from the increased value to the basic protection threshold over time after a load change event is triggered.

[0024] The load change event is triggered when the normalized difference is greater than the preset event trigger threshold.

[0025] The MCB and RCD collaborative distribution box control system based on intelligent algorithms includes:

[0026] The data acquisition module is used to acquire discrete-time current sequences;

[0027] The feature calculation module is used to generate a time-frequency coefficient matrix based on the discrete-time current sequence, and to calculate the high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor based on the time-frequency coefficient matrix and the discrete-time current sequence.

[0028] The feature signature generation module is used to combine high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor to generate electrical disturbance feature signatures through a nonlinear fusion model.

[0029] The dynamic threshold generation module is used to dynamically generate adaptive tolerance thresholds based on discrete-time current sequences.

[0030] The collaborative protection decision module is used to output a synchronous trip command if the duration of the state where the electrical disturbance signature is greater than the adaptive tolerance threshold exceeds a preset confirmation window; if the electrical disturbance signature is less than or equal to the adaptive tolerance threshold, the control system maintains the current operating state.

[0031] This invention provides an improved method and system for controlling MCB and RCD coordinated distribution boxes based on intelligent algorithms, which has the following improvements and advantages compared with the prior art:

[0032] 1. This solution improves the accuracy of electrical fault identification, enabling precise differentiation between real faults and benign transient disturbances, and constructs a multi-dimensional feature analysis framework. Its core lies in generating electrical disturbance feature signatures through a nonlinear fusion model. It not only considers the energy intensity represented by the high-frequency energy integral but also innovatively introduces time-frequency distribution entropy and waveform randomness factors to jointly quantify the randomness and disorder of the disturbance. This design ingeniously simulates the physical essence of real arc faults, namely the concurrence of high energy and high randomness. This solution can accurately identify surge currents with high energy but relatively deterministic time-frequency structure and clearly distinguish them from dangerous arc faults that possess both high energy and high randomness, greatly reducing the false tripping rate of protection devices.

[0033] 2. By introducing a dynamic adaptive protection threshold, this solution significantly enhances the operational continuity and reliability of the power supply system, ensuring that the system's protection capability is not permanently reduced after a legitimate transient impact, and that it still has the ability to quickly disconnect from subsequent real arc faults caused by aging of line insulation. Attached Figure Description

[0034] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0035] Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0037] Example 1:

[0038] Please see Figure 1 This invention provides a control method and system technical solution for MCB and RCD coordinated distribution box based on intelligent algorithms, including: S1, acquiring discrete time current sequence;

[0039] S2. Based on the discrete-time current sequence, continuous wavelet transform is applied to generate a time-frequency coefficient matrix.

[0040] S3. Calculate the high-frequency energy integral based on the time-frequency coefficient matrix;

[0041] S4. Calculate the time-frequency distribution entropy based on the time-frequency coefficient matrix;

[0042] S5. Calculate the waveform randomness factor based on the discrete-time current sequence;

[0043] S6. By combining high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor, an electrical disturbance feature signature is generated through a nonlinear fusion model.

[0044] S7. Dynamically generate adaptive tolerance threshold based on discrete-time current sequence;

[0045] S8. If the duration of the electrical disturbance signature being greater than the adaptive tolerance threshold exceeds the preset confirmation window, a synchronous trip command is output to the MCB and RCD; if the electrical disturbance signature is less than or equal to the adaptive tolerance threshold, the current operating state is maintained.

[0046] To effectively capture the rich transient and non-stationary features in arc fault signals, the complex Morlet wavelet is preferentially selected as the mother wavelet, and its mathematical expression is:

[0047] ;

[0048] in, : represents the mother wavelet function, which is a function of time; : Represents time; : Represents the mathematical constant Pi, the ratio of π to 1200. Represents the bandwidth parameter; e: represents the base of the natural logarithm; j: represents the imaginary unit (j²=-1); The center frequency is represented by this wavelet. This wavelet was chosen because it has good resolution in both time and frequency, making it suitable for analyzing signals such as electric arcs that contain abrupt changes and broadband noise.

[0049] The high frequency defined at S3 refers to the frequency band from 2kHz to 100kHz, which is the typical energy concentration area of ​​arc fault radiated noise in low-voltage power distribution lines; in wavelet transform, this frequency band corresponds to a specific scale. Scope Therefore, high-frequency energy integration The calculation formula is: ,in Let be the time-frequency coefficient matrix, where, a: Represents the high-frequency energy integral; b: Represents the scale in the wavelet transform, which is inversely proportional to the frequency; c: Represents the time position in the wavelet transform. This represents the lower limit of the scale range corresponding to the high-frequency band; This represents the upper limit of the scale range corresponding to the high-frequency band; Represents the time-frequency coefficient matrix; It is a symbol used in calculus for double integration, representing integration over two different variables;

[0050] S8 includes a preset confirmation window. The duration of this preset confirmation window is determined based on experience and experimental data. Its value should be greater than the typical duration of a benign disturbance, such as the initial peak phase of a motor startup surge, and typically less than 100 milliseconds. Simultaneously, it should be less than the fault clearing time required by safety regulations. In this embodiment, the window duration can be set to 150 milliseconds. This means that only when... The state is maintained continuously for 150 milliseconds before it is confirmed as a serious fault requiring tripping, thus effectively avoiding misjudgment of instantaneous spikes;

[0051] The MCB and RCD collaborative distribution box control method based on intelligent algorithms described in this embodiment aims to construct a multi-dimensional and highly robust electrical fault identification and decision-making framework. The method begins by acquiring discrete-time current sequences through high-fidelity sampling, and then uses continuous wavelet transform to upgrade the one-dimensional time-domain signal to a two-dimensional time-frequency coefficient matrix. The significance of this step lies in moving beyond an isolated examination of current amplitude, instead revealing the fine dynamics of current evolution over time at different frequency scales, laying a data foundation for subsequent accurate extraction of fault fingerprint information. Based on this matrix, the method calculates three characteristic quantities with clear physical meaning in parallel: the high-frequency energy integral characterizing the fault energy intensity, the time-frequency distribution entropy quantifying the disorder of energy distribution, and the time-domain waveform randomness reflecting the randomness of the fault. Waveform randomness factor; these three features are combined through a nonlinear fusion model to generate a single scalar – the electrical disturbance feature signature; this scheme abandons the static threshold and instead dynamically generates an adaptive tolerance threshold based on the real-time load changes of the line; protection decisions are based on the continuous comparison between the feature signature and the adaptive threshold, and a time confirmation window is introduced to enhance the anti-interference capability of the decision; this series of interconnected steps enables the entire control method to transcend the limitations of traditional protection devices that rely solely on the effective value of the current. By comprehensively analyzing the energy, time-frequency pattern, and randomness of the fault, it achieves efficient suppression of benign disturbances such as motor starting surges, while maintaining high sensitivity to real arc faults caused by line insulation degradation, thus achieving the dual goals of power supply continuity and electrical safety.

[0052] Example 2:

[0053] Electrical disturbance feature signatures are generated using a nonlinear fusion model, including:

[0054] The maximum-minimum normalization function is used to process the high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor respectively to generate normalized feature parameters. The normalized feature parameters are then fused through a nonlinear multiplication gating model: the normalized time-frequency distribution entropy and waveform randomness factor are weighted and summed, and then multiplied by the normalized high-frequency energy integral as a gating factor to generate the final feature signature.

[0055] The boundary values ​​required for the maximum-minimum normalization function are determined by statistical analysis of a pre-labeled electrical event waveform database; the weight coefficients are determined by optimization training using a supervised machine learning algorithm on the electrical event waveform database.

[0056] Support vector machines are used as the classification algorithm, and radial basis function kernels are selected, i.e. The optimal penalty coefficient was found by performing a grid search on the training set in the database using 10-fold cross-validation. and kernel parameters The objective function for optimization is to maximize the classification margin while minimizing the classification error rate. Represents the radial basis function kernel; Represents the eigenvector; exp: represents the exponential function; Representative eigenvector and Squared Euclidean distance between them; weighting coefficient The determination of the optimal SVM model is based on the recursive feature elimination method, which evaluates the contribution of each normalized feature to the classification result, and then measures and normalizes the contribution.

[0057] The database contains over 10,000 waveform samples, covering: normal operation samples, with loads ranging from no-load to 125% of rated current; benign disturbance samples, including at least five different power levels, such as direct starting of asynchronous motors (0.75kW, 2.2kW, 5.5kW), inverter starting surges, and high-power resistive loads, such as heater switching processes; and real fault samples, simulated by an arc generator conforming to UL1699 standards, including series and parallel arc faults, covering various scenarios such as carbonization paths and conductor gap discharges. All samples were collected on a power distribution test platform with a rated voltage of 220V / 50Hz and a sampling rate of no less than 1MHz. The pre-labeling criteria for waveforms are as follows: any transient process caused by motor starting or load switching that decays to less than 1.5 times the steady-state current within 500 milliseconds is labeled as a surge; while signals conforming to UL1699 standards, with random high-frequency components (2kHz-100kHz) accounting for more than 5% of the total energy and lasting for more than 100 milliseconds, are labeled as arc faults.

[0058] In this embodiment, the generation of electrical disturbance feature signatures is based on a nonlinear fusion model that couples the inherent physical mechanism of arc faults.

[0059] The model is built upon insights into the physical characteristics of arc faults: a dangerous arc fault is not only inevitably accompanied by significant high-frequency energy release, but the energy distribution in the time-frequency domain also exhibits a highly random and disordered state. To effectively distinguish this type of fault from benign surges with high energy but relatively deterministic time-frequency structures, this technical solution proposes a nonlinear multiplicative gating fusion structure. The underlying logic is to use high-frequency energy as a necessary prerequisite for fault determination, and to adjust the weight of randomness features in a gating manner, thereby achieving accurate amplification and identification of fault features.

[0060] Electrical disturbance signature The calculation formula was revised and defined as:

[0061] ;

[0062] Both sides of the equation in this formula are dimensionless, ensuring the consistency of the model output; in the formula, This represents the high-frequency energy integral, whose physical dimension is the same as that of the signal energy, i.e., the square of the current multiplied by the time. ); This represents the time-frequency distribution entropy, which is a dimensionless value. It represents the waveform randomness factor, with the physical dimension being the square of the current; It is the maximum-minimum normalization function, i.e. Its function is to uniformly map three original characteristic parameters with different physical dimensions and numerical ranges to... dimensional interval; The signature representing the electrical disturbance is a dimensionless quantitative score. and These represent the lower and upper boundaries required for normalization, respectively. The method for determining these two values ​​is to conduct a comprehensive statistical analysis on a large-scale electrical event waveform database that is pre-labeled with event types such as normal, surge, and arc fault, thereby finding the empirical minimum and maximum values ​​for each feature. : represents the normalized output value; x: represents the original input feature parameter value; These are three dimensionless weight coefficients. The method for determining this set of weight coefficients is to use a supervised machine learning algorithm, such as a support vector machine, with the goal of maximizing the distance between fault and non-fault samples on the aforementioned waveform database to iteratively optimize the training until the model classification accuracy reaches the optimal level.

[0063] In applications, the normalized value of the high-frequency energy integral... As a multiplicative gating factor, its value directly determines the combined influence weight of the two random characteristics within the parentheses; only when significant high-frequency energy appears in the line, making... When the value is large, the term representing randomness and disorder and Only then will their contributions be significantly amplified, thus leading to The value represents a qualitative leap; the structure effectively filters out low-energy background noise interference; the system outputs a dimensionless, highly condensed quantization score. It accurately reflects the danger level of current electrical disturbances, greatly improving the ability to distinguish between real arc faults and benign surge currents in complex electromagnetic environments, and is the technical basis for reducing the false protection rate and leakage protection rate of distribution boxes.

[0064] Example 3:

[0065] The time-frequency distribution entropy is obtained by calculating the Shannon entropy of the time-frequency energy distribution.

[0066] In this embodiment, the time-frequency distribution entropy The calculation aims to quantify the uncertainty of energy in the time-frequency plane from the perspective of information theory, which is consistent with the highly random characteristics of electric arc faults.

[0067] The time-frequency distribution entropy originates from Shannon entropy in information theory. The technological motivation is to provide a rigorous mathematical measure for the abstract form of chaos. The energy of an electric arc discharge process is randomly distributed across a broad time-frequency spectrum, while the energy distribution of a normal surge is more concentrated. Introducing Shannon entropy can effectively transform this difference in form into a calculable value.

[0068] For calculation The time-frequency coefficient matrix needs to be... Energy Matrix Normalization yields the dimensionless time-frequency probability distribution matrix. :

[0069] ;

[0070] in, Indicates the signal at a specific time point. and scale The energy density in the vicinity, with the denominator being the total energy within the time-frequency analysis window, ensures that all The sum is 1; calculate using the Shannon entropy formula. :

[0071] ;

[0072] in, The representative scale is inversely proportional to the frequency. Represents time and location;

[0073] The output of this calculation process is a dimensionless scalar. ; The higher the value, the more dispersed and unpredictable the energy distribution in the time-frequency plane, which is highly consistent with the physical characteristics of arc faults; Representing the logarithm to the base 2; therefore, the time-frequency distribution entropy, as a key feature parameter, provides a powerful dimension for the system to distinguish chaotic fault signals from structured non-fault signals, significantly enhancing the recognition accuracy of the entire control method.

[0074] Example 4:

[0075] The waveform randomness factor is obtained by calculating the variance of the residual signal formed by the difference between the discrete-time current sequence and the fundamental frequency component.

[0076] Waveform randomness factor in this embodiment Its design aims to quantify, from a time-domain perspective, the degree of random fluctuation in the current waveform that deviates from its ideal power frequency sinusoidal shape;

[0077] The technical motivation is that electric arcs superimpose a large amount of non-periodic random noise onto the power frequency current; by stripping away the deterministic power frequency fundamental component, this hidden randomness can be highlighted in the residual signal; variance, as a classic statistic for measuring the degree of data dispersion, is used as the final indicator to quantify the severity of this random fluctuation.

[0078] The calculation of this factor involves two steps: from the original discrete-time current sequence... In this process, the fundamental frequency component of the power frequency is extracted using a bandpass filter or Fourier transform method. The residual signal is obtained by calculating the difference between the two. Calculate the variance of the residual signal within a time window; this is the waveform randomness factor. The physical dimension of this parameter is the square of the current. ); Represents the original discrete-time current sequence; An index representing a discrete time series;

[0079] In power distribution scenarios, random noise generated by an initial electric arc can affect the residual signal. The fluctuations are much greater than the residual fluctuations caused by the deterministic harmonics generated by the nonlinear load; therefore, the calculated... The value will be significantly higher; this factor constitutes an indicator sensitive to early, weak arc faults, which effectively complements the frequency domain-based features, together constructing a more comprehensive perception of fault modes, thereby improving the early warning and protection capabilities for latent faults.

[0080] Example 5:

[0081] Dynamically generate adaptive tolerance thresholds, including:

[0082] Set a basic protection threshold; calculate the short-term and long-term exponential moving averages of the line current RMS value in parallel; calculate the normalized difference between the short-term and long-term exponential moving averages; combine the basic protection threshold and the normalized difference, and introduce an exponential decay factor to generate an adaptive tolerance threshold.

[0083] An exponential decay factor is introduced to allow the adaptive tolerance threshold to recover from the increased value to the basic protection threshold over time after a load change event is triggered.

[0084] The load change event is triggered when the normalized difference is greater than the preset event trigger threshold.

[0085] The purpose of dynamically generating the adaptive tolerance threshold in this embodiment is to enable the sensitivity of the protection system to intelligently adapt to the real-time operating context of the power grid and resolve the contradiction between protection sensitivity and power supply reliability.

[0086] The design concept of this dynamic threshold stems from an engineering judgment: after a legitimate, large load input, the probability of transient electrical disturbances occurring in the short term is much higher than the probability of a real fault; in order to avoid misjudging such expected surges as faults, a mechanism must be designed that can temporarily raise the protection threshold after a drastic change in load conditions and automatically recover after a period of time.

[0087] Adaptive tolerance threshold The calculation formula is constructed as follows:

[0088] ;

[0089] Both sides of the equation are dimensionless quantities, and... The dimensions are consistent; among them, The adaptive tolerance threshold, which varies with time, is a dimensionless quantity, where t represents time. The coefficient representing the dimensionless influence of load variation. The dimensionless basic protection threshold represents the highest protection sensitivity of the system under stable operating conditions. Its value is determined by analyzing all arc fault samples in the aforementioned waveform database. The values ​​are statistically analyzed, and the lower limit of the confidence interval of their distribution is taken and combined with the safety margin to determine the value. It is the normalized difference reflecting the trend of current growth, calculated as follows: ,in and These are the short-term and long-term exponential moving averages of the effective value of the line current, respectively. This refers to the rated current of the line. This is a dimensionless coefficient representing the influence of load variation. It is an exponential decay factor, and the triggering condition is... , It is a preset dimensionless event trigger threshold; once the condition is met, the timer... Reset to 0 and start the timer. This represents the time elapsed since the most recent load change event was triggered, marking the start of the decay process; Parameter Attenuation constant The unit is as well as The method for determining the value is based on a large number of simulation tests and optimization algorithms conducted using the aforementioned electrical event waveform database. Represents the normalized difference;

[0090] The specific optimization process is as follows: Define an objective function. ,in This represents the number of times all arc fault samples in the database were not detected correctly. It represents the number of times all benign surge samples were incorrectly identified as faults. and The penalty factor can be set, for example, to 10 or 1; then, particle swarm optimization or a genetic algorithm is used within a preset parameter space, for example... , , Perform an iterative search to find the objective function. Minimize the parameter combination { Basic protection threshold The method for determining this is as follows: calculate the values ​​of all arc fault samples in the database. mean of values and standard deviation ,Pick This ensures sensitivity to over 99% of real-world faults;

[0091] When a high-power load starts up instantaneous exceed The system triggers a reset mechanism, making from The boost is rapid, and the peak value of this boost is sufficient to withstand the surge generated during the startup. Peak values ​​are minimized to avoid unnecessary tripping; after startup... Start timing. Under the influence of the attenuation factor, it smoothly recovers to the baseline value of high sensitivity. This dynamic adjustment mechanism enables the protection system to maximize the continuous operation of critical loads without sacrificing its ability to protect against severe failures. Feature signature representing all arc fault samples in the database The mean of the values; express Three times the standard deviation of the value.

[0092] Example 6:

[0093] A control system for a coordinated MCB and RCD distribution box based on intelligent algorithms, comprising:

[0094] The data acquisition module is used to acquire discrete-time current sequences;

[0095] The feature calculation module is used to generate a time-frequency coefficient matrix based on the discrete-time current sequence, and to calculate the high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor based on the time-frequency coefficient matrix and the discrete-time current sequence.

[0096] The feature signature generation module is used to combine high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor to generate electrical disturbance feature signatures through a nonlinear fusion model.

[0097] The dynamic threshold generation module is used to dynamically generate adaptive tolerance thresholds based on discrete-time current sequences.

[0098] The collaborative protection decision module is used to output a synchronous trip command if the duration of the state where the electrical disturbance signature is greater than the adaptive tolerance threshold exceeds a preset confirmation window; if the electrical disturbance signature is less than or equal to the adaptive tolerance threshold, the control system maintains the current operating state.

[0099] The MCB and RCD collaborative distribution box control system based on intelligent algorithms disclosed in this embodiment physically implements the aforementioned control methods through its internal structured functional module division, forming a complete intelligent protection execution entity; the system consists of five closely cooperating modules:

[0100] The data acquisition module integrates a high-bandwidth current sensor and a high-sampling-rate analog-to-digital converter, which is responsible for continuous and high-fidelity sampling of the line current, providing an accurate discrete-time current sequence for subsequent analysis.

[0101] The feature calculation module receives the raw data stream, maps it to the time-frequency domain using the continuous wavelet transform algorithm, and on this basis, calculates three core feature parameters in parallel according to the preset algorithm: high-frequency energy integral, time-frequency distribution entropy, and waveform randomness factor.

[0102] The signature generation module receives the three independent feature parameters mentioned above and applies the aforementioned nonlinear fusion model and weighting coefficients obtained based on data-driven optimization to extract multidimensional information into an electrical disturbance signature that quantifies the degree of disturbance hazard. ;

[0103] The dynamic threshold generation module, also taking the original current sequence as input, analyzes the long-term and short-term trends of the effective current value to calculate an adaptive tolerance threshold that can dynamically adapt to the current load state in real time. ;

[0104] The collaborative protection decision-making module executes the final logical decision, processing the signature at an extremely high frequency. With adaptive threshold Compare; only when Continuously surpass Only when a preset time confirmation window is reached will the module finally determine that it is a serious fault and immediately issue a synchronous trip command to the trip unit of MCB and RCD; in all other cases, the system maintains its current operating state.

[0105] Through this modular design, the system clearly maps complex algorithm processes to functional units. Each module performs its own function and works together efficiently to form an intelligent power distribution protection solution that can accurately identify and respond to real electrical faults while being highly immune to normal operational disturbances.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A control method for a coordinated distribution box of MCB and RCD based on intelligent algorithms, characterized in that, include: S1. Obtain the discrete-time current sequence; S2. Based on the discrete-time current sequence, continuous wavelet transform is applied to generate a time-frequency coefficient matrix. S3. Calculate the high-frequency energy integral based on the time-frequency coefficient matrix; S4. Calculate the time-frequency distribution entropy based on the time-frequency coefficient matrix; S5. Calculate the waveform randomness factor based on the discrete-time current sequence; S6. Combining high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor, an electrical disturbance feature signature is generated through a nonlinear fusion model. S7. Dynamically generate adaptive tolerance threshold based on discrete-time current sequence; S8. If the duration of the electrical disturbance signature being greater than the adaptive tolerance threshold exceeds the preset confirmation window, then output a synchronous trip command to the MCB and RCD. If the electrical disturbance signature is less than or equal to the adaptive tolerance threshold, the current operating state is maintained. Electrical disturbance feature signatures are generated using a nonlinear fusion model, including: The maximum-minimum normalization function is used to process the high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor respectively to generate normalized feature parameters. The normalized feature parameters are then fused through a nonlinear multiplication gating model: the normalized time-frequency distribution entropy and waveform randomness factor are weighted and summed, and then multiplied by the normalized high-frequency energy integral as a gating factor to generate the final feature signature. The boundary values ​​required for the maximum-minimum normalization function are determined by statistical analysis of a pre-labeled electrical event waveform database; the weight coefficients are determined by optimization training using a supervised machine learning algorithm on the electrical event waveform database.

2. The MCB and RCD coordinated distribution box control method based on intelligent algorithms according to claim 1, characterized in that, The time-frequency distribution entropy is obtained by calculating the Shannon entropy of the time-frequency energy distribution.

3. The MCB and RCD coordinated distribution box control method based on intelligent algorithms according to claim 1, characterized in that, The waveform randomness factor is obtained by calculating the variance of the residual signal formed by the difference between the discrete-time current sequence and the fundamental frequency component.

4. The MCB and RCD coordinated distribution box control method based on intelligent algorithms according to claim 1, characterized in that, Dynamically generate adaptive tolerance thresholds, including: Set a basic protection threshold; calculate the short-term and long-term exponential moving averages of the line current RMS value in parallel; calculate the normalized difference between the short-term and long-term exponential moving averages; combine the basic protection threshold and the normalized difference, and introduce an exponential decay factor to generate an adaptive tolerance threshold.

5. The MCB and RCD coordinated distribution box control method based on intelligent algorithms according to claim 4, characterized in that, An exponential decay factor is introduced to allow the adaptive tolerance threshold to recover from the increased value to the basic protection threshold over time after a load change event is triggered. The load change event is triggered when the normalized difference is greater than the preset event trigger threshold.

6. A control system for a collaborative distribution box of MCB and RCD based on intelligent algorithms, applied to the control method for a collaborative distribution box of MCB and RCD based on intelligent algorithms as described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to acquire discrete-time current sequences; The feature calculation module is used to generate a time-frequency coefficient matrix based on the discrete-time current sequence, and to calculate the high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor based on the time-frequency coefficient matrix and the discrete-time current sequence. The feature signature generation module is used to combine high-frequency energy integral, time-frequency distribution entropy and waveform randomness factor to generate electrical disturbance feature signatures through a nonlinear fusion model. The dynamic threshold generation module is used to dynamically generate adaptive tolerance thresholds based on discrete-time current sequences. The collaborative protection decision module is used to output a synchronous trip command if the duration of the state where the electrical disturbance signature is greater than the adaptive tolerance threshold exceeds a preset confirmation window; if the electrical disturbance signature is less than or equal to the adaptive tolerance threshold, the control system maintains the current operating state.

Images