Standby power supply system self-security fault protection method based on logistic regression model

Abstract

The invention discloses a backup power supply system self-security fault protection method based on a logistic regression model, and the method comprises the following steps: analyzing the operation states of a main power supply system under different faults, and determining the motion logic; analyzing action logic, and judging an error scene; feature mining is carried out on different false rejection situations, and typical features during failure are extracted; calculating a failure probability by adopting a logistic regression algorithm through classification of historical data; according to the fault type and the influence degree, carrying out grade division on the error scene; and aiming at different fault scenes, designing protection strategies. According to the invention, the maloperation and refusal operation risks in a complex fault scene are effectively reduced; and through an error probability model and machine learning analysis, a fault mode is accurately identified and graded protection is realized.

Description

Technical Field

[0001] This invention belongs to the field of power system distribution technology, specifically relating to a method for protecting the self-safety of backup power systems based on a logistic regression model. Background Technology

[0002] In modern power systems, backup / emergency power systems are a crucial component for ensuring uninterrupted power supply to critical loads, widely used in scenarios with extremely high power reliability requirements, such as hospitals, data centers, and industrial production. Automatic transfer switch (ATS) devices, as a core component of backup power systems, are responsible for rapidly switching to backup power when the main power source fails, maintaining power continuity. These devices typically detect electrical parameters such as voltage and current, and automatically switch according to preset logic, with switching times generally ranging from tens to hundreds of milliseconds.

[0003] In existing technologies, the operating logic of automatic transfer switch (ATS) devices is mainly based on simple voltage loss or overcurrent detection, and completes power switching with a fixed time limit. For example, a typical ATS device triggers the switch action by monitoring the bus voltage below a certain threshold or the line current exceeding a set value. However, in actual operation, the power distribution system may face various complex fault types, including single-phase grounding faults, three-phase short-circuit faults, and voltage loss faults. Furthermore, the operating logic of the device may misjudge or fail due to factors such as the operating mode of primary equipment, system power flow distribution, and protection time limit coordination. Summary of the Invention

[0004] This invention is proposed to address the problems existing in the prior art, and its purpose is to provide a method for protecting the self-safety of backup power systems based on a logistic regression model.

[0005] The technical solution of this invention is: a method for self-safety fault protection of backup power systems based on logistic regression models, comprising the following steps: A. Analyze the operating status of the main power system under different faults and determine the action logic; B. Analyze the action logic and determine the error scenario; C. Perform feature mining on different false start scenarios to extract typical features when failure occurs; D. By classifying historical data, the failure probability is calculated using the logistic regression algorithm; E. Classify error scenarios into levels based on the type of failure and the degree of impact; F. Design protection strategies for different fault scenarios.

[0006] Furthermore, step A analyzes the operating status of the main power system under different faults and determines the action logic. The specific process is as follows: First, determine the operating logic of the automatic transfer switch under the fault type; The voltage detection formula is as follows: ; In the formula, It is a three-phase voltage. To detect voltage and determine the state of undervoltage; The formula based on current detection is as follows: ; In the formula, For three-phase current, This is the overcurrent setting value, used to trigger switching; Then, for single-phase grounding faults, the automatic transfer switch determines the fault conditions by detecting zero-sequence voltage and current, and switches to the backup power supply. Subsequently, in response to a three-phase short-circuit fault, the device switches to the backup power supply by detecting a sudden voltage drop. Finally, in response to a voltage loss fault, the device uses multi-point voltage monitoring to determine the duration of the voltage loss and switches to the backup power supply.

[0007] Furthermore, step B analyzes the action logic and identifies the error scenario, including the analysis of influencing factors. The specific process is as follows: First, identify the factors that affect the correct operation of the device, including the impact of the primary equipment's operating mode, the impact of the system power flow distribution, the impact of the coordination between automatic transfer switch and line protection time limits, the impact of voltage loss cause identification, and the impact of the incoming line automatic transfer mode. Then, for each influencing factor, we analyze the possible scenarios that may lead to errors in action logic judgment under different fault types; Finally, the error scenarios are classified into levels based on the type of failure and the degree of impact.

[0008] Further analysis of influencing factors follows as follows: First, the operating mode of the primary equipment has an impact. In open-loop operation, deviations in the power flow direction can lead to misjudgments, namely: ; In the formula, From a trend perspective, These are active and reactive power, respectively. Then, due to the influence of system power flow distribution, power flow overload is misjudged as a short circuit fault, that is: ; In the formula, Overload factor, Rated current; Furthermore, due to the influence of timing coordination, the automatic switching action time is mismatched with the protection, that is: ; In the formula, To allow time for the self-throwing action, For the duration of the protection action; Furthermore, the determination of the cause of pressure loss was affected, and the type of pressure loss was not correctly distinguished, namely: ; In the formula, These are the external and internal pressure loss components, respectively; Finally, the impact of the automatic transfer switch mode: priority errors lead to switching to an unreliable power source, i.e.: ; In the formula, For power priority, As weight, These represent the power supply voltage and current.

[0009] Furthermore, step C performs feature mining on different false start scenarios to extract typical features at the time of failure. The specific process is as follows: First, collect historical operating data of the device, including action records, fault waveforms, and switch status; Then, the fault mechanism of the protection device failing to operate or operating erroneously is analyzed; Finally, feature mining was performed on different false start scenarios to extract typical features of failure.

[0010] Furthermore, step D involves classifying historical data and using a logistic regression algorithm to calculate the failure probability. The specific process is as follows: The mathematical expression of the failure probability model is: ; In the formula, It is a non-linear function, and a logistic regression model is used to fit historical data: ; In the formula, It is a linear combination of influencing factors: ; In the formula, The intercept of the model represents the base error probability; The weights of each influencing factor, reflecting their contribution to the error probability, are obtained through training with historical data.

[0011] Furthermore, step E categorizes the error scenarios into levels based on the type and severity of the fault, as detailed below: The error scenarios are classified into the following levels: Level 1 Fault: Probability of Failure >0.9, causing the device to completely fail to operate or to operate malfunctioning; Level 2 fault: failure probability 0.5< <0.9, causing delays in action or switching to a non-optimal power supply; Level 3 Fault: Probability of Failure <0.5, slight motion deviation.

[0012] Furthermore, step F involves designing protection strategies for different fault scenarios, the specific process of which is as follows: Design protection strategies for different failure scenarios; For level 1 faults, modify the action logic and add redundant judgment conditions: ; In the formula, For redundant discriminant values, As weight; For level 2 faults, add auxiliary devices and optimize switching priorities: ; In the formula, Prioritize each power supply; For level 3 faults, adjust the time limit parameter: ; In the formula, To optimize time limits; Establish a dynamic update mechanism to adjust logical parameters based on real-time data.

[0013] Backup / emergency power systems based on logistic regression models for self-safety fault protection include: Equipped with an automatic transfer device, the action logic switching is executed. The data acquisition module monitors the system status and records the operating status. The fault analysis module analyzes the fault mechanism and identifies fault characteristics. The protection strategy execution module dynamically adjusts logical parameters to implement tiered protection.

[0014] The beneficial effects of this invention are as follows: This invention significantly improves the power supply reliability and adaptability of backup power systems by optimizing the operation logic of automatic transfer switch, quantifying influencing factors, identifying fault characteristics, and formulating graded protection strategies.

[0015] This invention employs multi-dimensional electrical parameter detection and dynamic logic adjustment to effectively reduce the risk of false operation and failure to operate in complex fault scenarios; through error probability models and machine learning analysis, it accurately identifies fault modes and achieves graded protection. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the structure of the present invention; Figure 3 This is a schematic diagram of the automatic transfer switch under the dual busbar connection method in this invention. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: like Figures 1 to 3 As shown, the self-safety fault protection method for backup power systems based on logistic regression models is characterized by the following steps: A. Analyze the operating status of the main power system under different faults and determine the action logic; B. Analyze the action logic and determine the error scenario; C. Perform feature mining on different false start scenarios to extract typical features when failure occurs; D. By classifying historical data, the failure probability is calculated using the logistic regression algorithm; E. Classify error scenarios into levels based on the type of failure and the degree of impact; F. Design protection strategies for different fault scenarios.

[0018] Step A analyzes the operating status of the main power system under different faults and determines the action logic. The specific process is as follows: First, determine the operating logic of the automatic transfer switch under the fault type; The voltage detection formula is as follows: (1); In the formula, It is a three-phase voltage. To detect voltage and determine the state of undervoltage; The formula based on current detection is as follows: (2); In the formula, For three-phase current, This is the overcurrent setting value, used to trigger switching; Then, for single-phase grounding faults, the automatic transfer switch determines the fault conditions by detecting zero-sequence voltage and current, and switches to the backup power supply. Subsequently, in response to a three-phase short-circuit fault, the device switches to the backup power supply by detecting a sudden voltage drop. Finally, in response to a voltage loss fault, the device uses multi-point voltage monitoring to determine the duration of the voltage loss and switches to the backup power supply.

[0019] Specifically, for single-phase ground faults, the automatic transfer switch determines the fault conditions through zero-sequence voltage and current detection, as follows: (3) In the formula, or The device triggers the isolation fault section and switches to the backup power supply; Specifically, for a three-phase short-circuit fault, the device detects a sudden voltage drop, as follows: (4); In the formula, This is normal voltage. The voltage descent threshold. The device first disconnects the faulty line, and then switches to the backup power supply; Specifically, for undervoltage faults, the device uses multi-point voltage monitoring to determine the undervoltage duration: (5); In the formula, The threshold for pressure loss. The switchover will be completed in time.

[0020] Step B involves analyzing the action logic and identifying the error scenario, including an analysis of influencing factors. The specific process is as follows: First, identify the factors that affect the correct operation of the device, including the impact of the primary equipment's operating mode, the impact of the system power flow distribution, the impact of the coordination between automatic transfer switch and line protection time limits, the impact of voltage loss cause identification, and the impact of the incoming line automatic transfer mode. Then, for each influencing factor, we analyze the possible scenarios that may lead to errors in action logic judgment under different fault types; Finally, the error scenarios are classified into levels based on the type of failure and the degree of impact.

[0021] Specifically, we analyze the scenarios that may lead to incorrect action logic judgments under different fault types, based on an action probability model: (6); In the formula, For load current, For system voltage, Due to time limit deviation, This is due to communication delay.

[0022] Specifically, error scenarios are classified into levels, which include: Level 1 Fault: Operation Probability >0.9, causing the device to completely fail to operate or to operate malfunctioning; Level 2 fault: probability of operation 0.5 < <0.9, causing delays in action or switching to a non-optimal power supply; Level 3 Fault: Operation Probability <0.5, slight motion deviation.

[0023] The analysis of influencing factors is conducted as follows: First, the operating mode of the primary equipment has an impact. In open-loop operation, deviations in the power flow direction can lead to misjudgments, namely: (7); In the formula, From a trend perspective, These are active and reactive power, respectively. Then, due to the influence of system power flow distribution, power flow overload is misjudged as a short circuit fault, that is: (8); In the formula, Overload factor, Rated current; Furthermore, due to the influence of timing coordination, the automatic switching action time is mismatched with the protection, that is: (9); In the formula, To allow time for the self-throwing action, For the duration of the protection action; Furthermore, the determination of the cause of pressure loss was affected, and the type of pressure loss was not correctly distinguished, namely: (10); In the formula, These are the external and internal pressure loss components, respectively; Finally, the impact of the automatic transfer switch mode: priority errors lead to switching to an unreliable power source, i.e.: (11); In the formula, For power priority, As weight, These represent the power supply voltage and current.

[0024] Step C involves feature mining for different false start scenarios to extract typical features at the time of failure. The specific process is as follows: First, collect historical operating data of the device, including action records, fault waveforms, and switch status; Then, feature mining is performed on different false start scenarios to extract typical features of failure.

[0025] Specifically, extracting typical characteristics of failure includes: 1. Refusal to move characteristic: signal detection value No switch was triggered; 2. Malfunction characteristics: Malfunction signals Error switching; 3. Delay characteristics: Action time .

[0026] More specifically, in the failure-to-operate characteristic, the device fails to trigger a switch when it should operate, usually due to signal detection failure or logical judgment error; Detection of motion signal value Is it zero? (13); like This indicates that no undervoltage or overcurrent signal was detected, which may be due to sensor malfunction or an excessively high threshold setting.

[0027] More specifically, in the characteristics of malfunction, the device erroneously triggers switching under fault-free or unnecessary conditions, usually due to misjudging normal fluctuations as faults; Detection of malfunction signals Does it exceed the set threshold? (14); like This indicates that a brief fluctuation in voltage or current is misjudged as a fault, possibly due to a sudden change in power flow or electromagnetic interference.

[0028] More specifically, in terms of delay characteristics, the device action time exceeds the maximum allowable time limit, resulting in a handover delay; Calculate actual action time With maximum allowed time Deviation: (15); like This indicates a delay in action, which may be due to communication delays or slow hardware response.

[0029] Step D involves classifying historical data and using a logistic regression algorithm to calculate the failure probability. The specific process is as follows: The mathematical expression of the failure probability model is: (16); In the formula, It is a non-linear function, and a logistic regression model is used to fit historical data: (17); In the formula, It is a linear combination of influencing factors: (18); In the formula, The intercept of the model represents the base error probability; The weights of each influencing factor, reflecting their contribution to the error probability, are obtained through training with historical data.

[0030] Step E involves classifying the error scenarios into levels based on the type and severity of the fault. The specific process is as follows: The error scenarios are classified into the following levels: Level 1 Fault: Probability of Failure >0.9, causing the device to completely fail to operate or to operate malfunctioning; Level 2 fault: failure probability 0.5< <0.9, causing delays in action or switching to a non-optimal power supply; Level 3 Fault: Probability of Failure <0.5, slight motion deviation.

[0031] Step F involves designing protection strategies for different fault scenarios. The specific process is as follows: Design protection strategies for different failure scenarios; For level 1 faults, modify the action logic and add redundant judgment conditions: (19); In the formula, For redundant discriminant values, As weight; For level 2 faults, add auxiliary devices and optimize switching priorities: (20); In the formula, Prioritize each power supply; For level 3 faults, adjust the time limit parameter: (twenty one); In the formula, To optimize time limits.

[0032] Backup / emergency power systems based on logistic regression models for self-safety fault protection include: Equipped with an automatic transfer device, the action logic switching is executed. The data acquisition module monitors the system status and records the operating status. The fault analysis module analyzes the fault mechanism and identifies fault characteristics. The protection strategy execution module dynamically adjusts logical parameters to implement tiered protection.

[0033] Application Example 1 The self-safety fault protection method for backup power systems based on logistic regression models is as follows: The system analyzes the error scenarios in the action logic judgment, performs feature mining on different false action situations, calculates the failure probability by classifying historical data, and designs protection strategies for different fault scenarios based on this.

[0034] Specifically, with Figure 3 Taking the automatic transfer switch (ATS) in the double busbar connection method as an example, the #2 incoming line is fixedly connected to Bus I and the #3 incoming line is fixedly connected to Bus II. The ATS is connected to the voltage of Bus I and Bus II respectively, and the positions of switches 1DL, 2DL, and 4DL are as follows. The relevant circuits for the ATS are designed according to the fixed connection method. The operation of the ATS is analyzed below.

[0035] For single-phase ground faults, the automatic transfer switch determines the fault conditions by detecting zero-sequence voltage and current: (twenty two); In the formula, or The device triggers the isolation fault section and switches to the backup power supply; In response to a three-phase short-circuit fault, the device detects a sudden voltage drop: (twenty three); In the formula, This is normal voltage. The voltage descent threshold. The device first disconnects the faulty line, and then switches to the backup power supply; In response to voltage loss faults, the device uses multi-point voltage monitoring to determine the duration of voltage loss. (twenty four); In the formula, The threshold for pressure loss. The switchover will be completed in time.

[0036] Specifically, the fault scenario design protection strategy involves feature mining of the above-mentioned error scenarios in the action logic judgment under different faults, as follows: By classifying historical data and using a logistic regression algorithm, the failure probability is calculated: (25); In the formula, It is a linear combination of influencing factors: (26); In the formula, The intercept of the model represents the base error probability; This is a refusal to move signal. This is a false alarm signal. Due to time limit deviation, For communication delay; The weights of each influencing factor, reflecting their contribution to the error probability, are obtained through training with historical data; Level 1 Fault: Probability of Failure >0.9, causing the device to completely fail to operate or to operate malfunctioning; Level 2 fault: failure probability 0.5< <0.9, causing delays in action or switching to a non-optimal power supply; Level 3 Fault: Probability of Failure <0.5, slight motion deviation.

[0037] For level 1 faults, modify the action logic and add redundant judgment conditions: (27); In the formula, For redundant discriminant values, As weight; For level 2 faults, add auxiliary devices and optimize switching priorities: (28); In the formula, Prioritize each power supply; For level 3 faults, adjust the time limit parameter: (29); In the formula, To optimize time limits.

[0038] The key parameters in this application example are as follows: The value is 200V. The value is 100A. The value is 100V. The value is 0.3s; The value is 0.05s.

[0039] This invention analyzes the operating status of the main power supply system under different fault types such as single-phase grounding fault, three-phase short-circuit fault, and undervoltage fault; and determines the action logic of the device based on voltage detection formula and current detection formula, covering voltage / current detection, switch status judgment and switching timing control.

[0040] This invention provides a method for analyzing factors affecting operation, which identifies factors such as primary equipment operation mode, system power flow distribution, protection time limit coordination, communication delay, undervoltage cause identification, and incoming line automatic transfer mode. The degree of influence is quantified through an error probability model, and the fault scenarios are divided into three levels: Level 1 (complete failure to operate / false operation), Level 2 (operation delay / non-optimal switching), and Level 3 (minor deviation) fault levels.

[0041] This invention proposes a fault mechanism analysis method based on historical operating data. By collecting action records, fault waveforms and switch states, the fault occurrence rate is calculated, and a logistic regression model is used to mine features such as refusal to operate, false operation and delay, and to identify typical failure modes of the device.

[0042] This invention provides a method for formulating a graded protection strategy, which modifies the action logic or adds auxiliary devices for different fault levels, such as redundant discrimination conditions, optimized switching priorities, and adjusted time limit parameters.

[0043] The present invention also discloses a backup power supply fault protection system, including an automatic transfer switch, a data acquisition module, a fault analysis module, and a protection strategy execution module, which accurately identifies fault modes and achieves graded protection.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for self-protection against failure of a backup power system based on a logistic regression model, characterized in that: Comprising the following steps: A. Analyzing the operating state of the main power supply system under different faults and determining the action logic; B. Analyzing the action logic to determine the error scenario; C. Feature mining for different misoperation scenarios to extract typical features at the time of failure; D. By classifying historical data, using a logistic regression algorithm to calculate the failure probability; E. According to the fault type and the degree of influence, the error scenario is classified; F. Design protection strategies for different fault scenarios.

2. The method of claim 1, wherein the method is based on a logistic regression model of the backup power system self-safety failure protection. Step A analyzes the operating state of the main power supply system under different faults and determines the action logic, the specific process is as follows: First, determine the action logic of the backup power automatic switching device under the fault type; Among them, based on the voltage detection formula, as follows: ； In the formula, is a three-phase voltage, is a detection voltage for determining a voltage loss state; Among them, based on the current detection formula, as follows: ； wherein is the three-phase current, is the overcurrent setting value for triggering the switching; Then, for single-phase ground fault, the backup power automatic switching device detects zero sequence voltage and current to determine the fault condition and switch to the backup power supply; Then, for three-phase short-circuit fault, the device detects voltage drop to switch to the backup power supply; Finally, for loss of voltage fault, the device monitors multiple points of voltage to determine the loss of voltage time limit and switch to the backup power supply.

3. The self-safety fault protection method for backup power systems based on logistic regression models according to claim 1, characterized in that: Step B analyzes the action logic to determine the error scenario, including the analysis of influencing factors, the specific process is as follows: First, identify factors that affect the correct operation of the device, including the influence of primary equipment operation mode, the influence of system power flow distribution, the influence of backup power automatic switching and line protection time limit coordination, the influence of loss of voltage reason discrimination, and the influence of incoming line automatic switching mode; Then, for each influencing factor, analyze the error scenario that may lead to incorrect action logic under different fault types; Finally, according to the fault type and the degree of influence, the error scenario is classified.

4. The method of claim 3, wherein the method further comprises: The analysis of influencing factors is as follows: First, the influence of primary equipment operation mode, when open-loop operation, the power flow direction deviation leads to misjudgment, that is: ； wherein is the power angle, are the active and reactive power, respectively; Then, the influence of system power flow distribution, misjudgment of power flow overload as short-circuit fault, that is: ； wherein is the overload factor, is the rated current; Then, the influence of time limit coordination, mismatch between backup power automatic switching time limit and protection, that is: ； In the formula, is the backup action time, is the protection action time; Then, the influence of loss of voltage reason discrimination, incorrect discrimination of loss of voltage type, that is: ； wherein respectively the external and internal pressure loss components; Finally, the influence of incoming line automatic switching mode: priority error leads to switching to non-reliable power supply, that is: ； wherein is the power priority, is the weight, is the power voltage and current.

5. The method of claim 1, wherein the method further comprises: Step C extracts typical features at the time of failure by feature mining for different misoperation scenarios, the specific process is as follows: First, collect historical operation data of the device, including action records, fault waveforms and switch states; Then, analyze the fault mechanism of the protection device misoperation or misoperation; Finally, feature mining is performed for different misoperation scenarios to extract typical features at the time of failure.

6. The method of claim 1, wherein the method further comprises: Step D calculates the failure probability by classifying historical data and using a logistic regression algorithm, the specific process is as follows: The mathematical expression of the failure probability model is: ； wherein is a non-linear function, which is fitted to historical data using a logistic regression model: ； wherein is a linear combination of influencing factors: ； In the formula, is the intercept of the model, representing the base error probability; is the weight of each influencing factor, reflecting its contribution to the error probability, which is obtained by training historical data.

7. The method of claim 1, wherein the method further comprises: Step E classifies the error scenario according to the fault type and the degree of influence, the specific process is as follows: The error scenario classification is as follows: Primary failure: probability of failure > 0.9, leading to complete denial of operation or malfunction of the device; Secondary failure: 0.5 probability of failure <0.9, resulting in action delay or switch to non-optimal power source; Level 3 fault: failure probability <0.5, slight action deviation.

8. The method of claim 1, wherein the method further comprises: Step F designs protection strategies for different fault scenarios, the specific process is as follows: Design protection strategies for different fault scenarios; For first-level faults, modify the action logic and add redundant discrimination conditions: ； In the formula, is a redundancy decision value, is a weight; For second-level faults, add auxiliary devices and optimize switching priority: ； In the formula, is the priority of each power source; For three-level fault, adjust the time limit parameter: ； In the formula, to optimize the time limit; Establish a dynamic updating mechanism, adjust the logic parameters based on real-time data.

9. The backup / emergency power system of the method for self-protection against failure of a backup power system based on a logistic regression model according to claim 1, characterized in that: Including: The spare power automatic throw-in device executes the action logic switching; The data acquisition module monitors the system state and records the running state; The fault analysis module analyzes the fault mechanism and excavates the fault characteristics; The protection strategy execution module dynamically adjusts the logic parameters and implements hierarchical protection.

Images