Abnormal state monitoring and early warning method and system for measurement switch

Abstract

The invention relates to the technical field of electrical variable measurement, in particular to an abnormal state monitoring and early warning method and system for a measurement switch, and the method comprises the steps: obtaining historical measurement data, such as mechanical states, electrical characteristics and environmental conditions, and carrying out the preprocessing of the historical measurement data, so as to construct a measurement switch data set; key features of the measurement switch are extracted, a monitoring rule and an abnormal state monitoring model are constructed based on the features and expert experience, and the monitoring rule and the abnormal state monitoring model are imported into an expert knowledge base after being packaged; next, future measurement parameters are predicted through a machine learning model, and a first-level judgment result and a second-level judgment result are generated by combining the measurement parameters obtained in real time and utilizing an expert knowledge base; the judgment result is compared with the actual operation condition of the measurement switch, and the comprehensive false alarm rate and the missing report rate are calculated; and finally, on the basis of the false alarm rate and the missing report rate, the expert knowledge base is optimized by utilizing a genetic algorithm, the intelligence and the accuracy of monitoring the abnormal state of the measurement switch are improved, and a powerful guarantee is provided for the safety and the stability of power grid operation.

Description

Technical Field

[0001] The present invention relates to the technical field of measuring electrical variables, and in particular to a method and system for monitoring and warning abnormal status of a measuring switch. Background Art

[0002] The measuring switch is a low-voltage switch with a high-precision current sensor and a measuring unit. It is widely used to monitor and manage the power consumption of the power grid, and supports remote communication and data interaction to improve the management level of the power grid. However, the abnormal state of the measuring switch may cause the power grid monitoring system to fail or generate false alarms, thereby affecting the stable operation and maintenance efficiency of the power grid. Therefore, it is very important to effectively monitor and warn the abnormal state of the measuring switch.

[0003] The current measurement switch monitoring technology collects electrical parameters in real time through sensors, and monitors common faults based on a simple threshold judgment mechanism. If a fault occurs, the alarm device will notify the operation and maintenance personnel to perform maintenance. Although most existing measurement switch solutions have basic monitoring and protection functions, they are still insufficient in terms of intelligence and fault judgment accuracy. On the one hand, the existing solutions mainly rely on fixed threshold rules, which cannot adapt to grid parameters with large variations and are prone to false alarms or missed alarms. On the other hand, the existing solutions mostly process data at the collection and transmission level, lack in-depth analysis of historical data and trend prediction functions, and cannot provide forward-looking guidance for grid operation. Therefore, it is necessary to further improve the intelligence level of measurement switches to improve the accuracy of abnormal status monitoring and early warning of measurement switches.

[0004] To this end, a method and system for monitoring and warning abnormal status of a measuring switch are proposed. Summary of the invention

[0005] The purpose of the present invention is to provide an abnormal state monitoring and early warning method and system for a measuring switch, by acquiring historical measurement data such as mechanical state, electrical characteristics and environmental conditions, and preprocessing them to construct a measuring switch data set; extracting key features of the measuring switch, and constructing monitoring rules and abnormal state monitoring models based on these features and expert experience, and encapsulating them and importing them into an expert knowledge base; then, predicting future measurement parameters through a machine learning model, combining the measurement parameters obtained in real time, and using the expert knowledge base to generate primary and secondary judgment results; comparing the judgment results with the actual operating conditions of the measuring switch, and calculating the comprehensive false alarm rate and missed alarm rate; finally, based on the false alarm rate and missed alarm rate, using a genetic algorithm to optimize the expert knowledge base, thereby improving the intelligence and accuracy of abnormal state monitoring of the measuring switch, and providing a strong guarantee for the safety and stability of power grid operation.

[0006] To achieve the above object, the present invention provides the following technical solutions: The first part is a method for monitoring and warning abnormal status of a measuring switch, comprising: Acquire historical measurement switch parameters, wherein the historical measurement switch parameters include mechanical state parameters, electrical characteristic parameters, environmental parameters, and data integrity parameters; preprocess the historical measurement switch parameters to obtain a measurement switch data set; Extracting features from the measurement switch data set to obtain measurement switch features by calculation; Constructing a monitoring rule according to the measurement switch feature and expert experience, wherein the monitoring rule is used to set a corresponding feature threshold for each measurement switch feature; Establishing an abnormal state monitoring model, wherein the abnormal state monitoring model is used to determine whether the measuring switch is in an abnormal state; Encapsulating the characteristic threshold and the abnormal state monitoring model, and importing them into an expert knowledge base; Acquire real-time measurement switch parameters, and use a machine learning model to predict future measurement parameters based on the historical measurement switch parameters and the real-time measurement switch parameters; Use the expert knowledge base to predict the real-time measurement switch parameters and the future measurement parameters, and output a primary judgment result and a secondary judgment result; Acquire the actual operation status of the measuring switch at a predetermined time interval, and compare the actual operation status with the primary judgment result and the secondary judgment result to obtain a comprehensive false alarm rate and a comprehensive missed alarm rate; According to the comprehensive false alarm rate and the comprehensive omission rate, the expert knowledge base is optimized using a genetic algorithm to obtain an updated expert knowledge base.

[0007] Furthermore, the historically measured switch parameters specifically include: The mechanical state parameter, the electrical characteristic parameter, the environmental parameter and the data integrity parameter; The mechanical state parameters include switch action time, contact vibration frequency and switch positioning state; The electrical characteristic parameters include current, voltage, active power, reactive power and contact resistance; The environmental parameters include temperature, humidity, dust concentration and air pressure; The data integrity parameters include signal power, signal acquisition time, data loss amount and noise power.

[0008] Furthermore, the feature extraction process includes: Setting a timestamp, wherein the timestamp is used to obtain the historical measurement switch parameters; According to the historical measurement switch parameters, the measurement switch characteristics are calculated; the measurement switch characteristics include mechanical state characteristics, electrical characteristic characteristics, line loss characteristics, comprehensive environment characteristics and data signal characteristics; the mechanical state characteristics include abnormal action time characteristics, abnormal contact vibration characteristics and abnormal positioning state characteristics; the electrical characteristic characteristics include pressure loss characteristics, current loss characteristics and current imbalance characteristics; the data signal characteristics include abnormal time synchronization characteristics, abnormal data loss rate characteristics and abnormal signal quality characteristics; The calculation process of measuring the switch characteristics includes: Calculate the mean, maximum and standard deviation of the switch action time to obtain the time anomaly characteristics; Calculate the frequency spectrum characteristics of the contact vibration frequency using Fourier transform, and extract the main frequency from the frequency spectrum characteristics to obtain the abnormal contact vibration characteristics; Calculating the proportion of the incompletely closed state in the switch positioning state to obtain the abnormal characteristics of the positioning state; Extracting the minimum values ​​of the current and the voltage respectively to obtain the current loss characteristic and the voltage loss characteristic; calculating the current unbalance rate according to the current to obtain the current unbalance characteristic; Calculating a root mean square current according to the current, and obtaining the line loss characteristic according to the root mean square current and contact resistance; Calculating the mean values ​​of temperature, humidity, dust concentration and air pressure, and weighted summing the mean values ​​to obtain the comprehensive environmental characteristics; According to the signal acquisition time, the time synchronization anomaly feature is calculated; According to the data loss amount, signal power and noise power, the data loss rate abnormality feature and the signal quality abnormality feature are respectively calculated and obtained.

[0009] Furthermore, the predetermined time interval increases as the number of optimization times of the expert knowledge base increases, which is expressed as: ; in, For the suboptimal predetermined time interval, is the initial predetermined time interval, is an exponential function, is the attenuation coefficient, is the number of optimizations.

[0010] Furthermore, the optimization step of the expert knowledge base includes: S61: Setting an initial population, wherein the initial population includes the characteristic threshold and randomly generated threshold individuals; S62: Setting a fitness function, wherein the fitness function is expressed as: ; in, is the fitness value, , and is the weight coefficient, is the weight coefficient, is the false alarm rate, is the underreporting rate, is an exponential function, is the current threshold obtained during the optimization process, is the feature threshold, is the smoothing factor; S63: Evaluate the initial population according to the fitness function to obtain the fitness value; S64: According to the fitness value, select individuals whose fitness value is greater than a preset threshold value for crossover and mutation to obtain new threshold individuals; S65: updating the initial population according to the new threshold individuals; S66: If the number of optimizations reaches the predetermined number of iterations, the optimization is terminated and a new feature threshold is output; otherwise, S63 to S66 are repeated.

[0011] The second part is a monitoring and early warning system for abnormal state of a measuring switch, comprising: A parameter preprocessing module, used to obtain historical measurement switch parameters, preprocess the historical measurement switch parameters, and obtain a measurement switch data set; A feature extraction module, used to extract features from the measurement switch data set and calculate measurement switch features; An expert knowledge base establishment module is used to construct monitoring rules based on the measurement switch characteristics and expert experience, wherein the monitoring rules are used to set corresponding feature thresholds for each measurement switch characteristic; establish an abnormal state monitoring model, wherein the abnormal state monitoring model is used to determine whether the measurement switch is in an abnormal state; encapsulate the feature thresholds and the abnormal state monitoring model, and import them into the expert knowledge base; The abnormal monitoring module obtains the real-time measurement switch parameters, and uses the machine learning model to predict the future measurement parameters according to the historical measurement switch parameters and the real-time measurement switch parameters; uses the expert knowledge base to predict the real-time measurement switch parameters and the future measurement parameters, and outputs the primary judgment result and the secondary judgment result; The expert knowledge base optimization module obtains the actual operation status of the measuring switch at predetermined time intervals, and compares the actual operation status with the primary judgment result and the secondary judgment result to obtain a comprehensive false alarm rate and a comprehensive missed alarm rate; based on the comprehensive false alarm rate and the comprehensive missed alarm rate, a genetic algorithm is used to optimize the expert knowledge base to obtain an updated expert knowledge base.

[0012] Furthermore, the historically measured switch parameters specifically include: Mechanical state parameters, electrical characteristic parameters, environmental parameters and data integrity parameters; The mechanical state parameter, the electrical characteristic parameter, the environmental parameter and the data integrity parameter; The mechanical state parameters include switch action time, contact vibration frequency and switch positioning state; The electrical characteristic parameters include current, voltage, active power, reactive power and contact resistance; The environmental parameters include temperature, humidity, dust concentration and air pressure; The data integrity parameters include signal power, signal acquisition time, data loss amount and noise power.

[0013] Furthermore, the feature extraction process includes: Setting a timestamp, wherein the timestamp is used to obtain the historical measurement switch parameters; According to the historical measurement switch parameters, the measurement switch characteristics are calculated; the measurement switch characteristics include mechanical state characteristics, electrical characteristic characteristics, line loss characteristics, comprehensive environment characteristics and data signal characteristics; the mechanical state characteristics include abnormal action time characteristics, abnormal contact vibration characteristics and abnormal positioning state characteristics; the electrical characteristic characteristics include pressure loss characteristics, current loss characteristics and current imbalance characteristics; the data signal characteristics include abnormal time synchronization characteristics, abnormal data loss rate characteristics and abnormal signal quality characteristics; The calculation process of measuring the switch characteristics includes: Calculate the mean, maximum and standard deviation of the switch action time to obtain the time anomaly characteristics; Calculate the frequency spectrum characteristics of the contact vibration frequency using Fourier transform, and extract the main frequency from the frequency spectrum characteristics to obtain the abnormal contact vibration characteristics; Calculating the proportion of the incompletely closed state in the switch positioning state to obtain the abnormal characteristics of the positioning state; Extracting the minimum values ​​of the current and the voltage respectively to obtain the current loss characteristic and the voltage loss characteristic; calculating the current unbalance rate according to the current to obtain the current unbalance characteristic; Calculating a root mean square current according to the current, and obtaining the line loss characteristic according to the root mean square current and contact resistance; Calculating the mean values ​​of temperature, humidity, dust concentration and air pressure, and weighted summing the mean values ​​to obtain the comprehensive environmental characteristics; According to the signal acquisition time, the time synchronization anomaly feature is calculated; According to the data loss amount, signal power and noise power, the data loss rate abnormality feature and the signal quality abnormality feature are respectively calculated and obtained.

[0014] Furthermore, the predetermined time interval increases as the number of optimization times of the expert knowledge base increases, which is expressed as: ; in, For the suboptimal predetermined time interval, is the initial predetermined time interval, is an exponential function, is the attenuation coefficient, is the number of optimizations.

[0015] Furthermore, the optimization step of the expert knowledge base includes: S61: Setting an initial population, wherein the initial population includes the characteristic threshold and randomly generated threshold individuals; S62: Setting a fitness function, wherein the fitness function is expressed as: ; in, is the fitness value, , and is the weight coefficient, is the weight coefficient, is the false alarm rate, is the underreporting rate, is an exponential function, is the current threshold obtained during the optimization process, is the feature threshold, is the smoothing factor; S63: Evaluate the initial population according to the fitness function to obtain the fitness value; S64: According to the fitness value, select individuals whose fitness value is greater than a preset threshold value for crossover and mutation to obtain new threshold individuals; S65: updating the initial population according to the new threshold individuals; S66: If the number of optimizations reaches the predetermined number of iterations, the optimization is terminated and a new feature threshold is output; otherwise, S63 to S66 are repeated.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention refines the key characteristic indicators of the measuring switch through multi-dimensional feature extraction of parameters such as mechanical state, electrical characteristics, environmental parameters and data integrity, improves the ability to capture abnormal states, and makes the monitoring process more comprehensive. In addition, the monitoring rules are constructed by combining expert experience with data-driven methods, and feature thresholds are set according to the calculation results of historical parameters, which enhances the adaptability of the rules and reduces the subjective interference of single threshold setting in traditional methods, thereby improving the accuracy of abnormal state monitoring of the measuring switch.

[0017] 2. The present invention uses monitoring rules to establish an abnormal state monitoring model, encapsulates the characteristic threshold and the abnormal state monitoring model, and imports the expert knowledge base, which can automatically apply this knowledge to judge abnormalities, and realize real-time monitoring based on historical data and experience. In addition, the introduction of a machine learning prediction model can predict future monitoring parameters based on real-time parameters, and combine real-time parameters with future monitoring parameters to output judgment results respectively, which can timely identify potential abnormalities and quickly respond to different types of abnormal situations, thereby improving the accuracy of abnormal state monitoring of the measuring switch.

[0018] 3. The present invention constructs a closed-loop optimization process by introducing a genetic algorithm to optimize the expert knowledge base, thereby reducing the problems of false alarms and missed alarms caused by deviations in feature threshold settings. At the same time, the introduction of false alarm rate and missed alarm rate in the fitness function can continuously improve the optimization strategy according to the actual operating conditions, so that the optimization target is more in line with the actual application requirements, thereby improving the accuracy of abnormal state monitoring. In addition, the dynamic time interval mechanism is introduced during the optimization process, so that the expert knowledge base can adapt to changes in the actual operating environment, which not only ensures the monitoring effect, but also effectively reduces the resource consumption caused by frequent calculations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram of a flow chart of a method for monitoring and warning abnormal status of a measuring switch proposed in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a measuring switch feature proposed in an embodiment of the present invention; Figure 3 The present invention is a schematic diagram of the structure of an abnormal state monitoring and early warning system for a measuring switch proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figures 1 to 3 The present invention provides a method and system for monitoring and warning the abnormal state of a measuring switch, and the technical solution is as follows: Embodiment 1 is as follows: As the load on the power grid gradually increases, the power system's requirements for monitoring accuracy and real-time response capabilities are increasing. As an important part of power grid monitoring and management, measuring switches help dispatch centers and managers to keep abreast of the power grid status in a timely manner. However, with the increase in power grid load and the aging of measuring switch equipment, measuring switches may face a series of potential abnormal risks. For example, the communication module may lose signals due to electromagnetic interference or connection problems, or even the aging of the equipment may cause the switch's response to be delayed or fail. If these abnormal conditions are not discovered in time, the monitoring system will not be able to accurately reflect the true situation of the power grid, and may even cause false alarms or missed alarms, ultimately affecting the stability and operating efficiency of the power grid.

[0022] Some existing solutions can achieve basic monitoring of measuring switches, but they are still insufficient in terms of intelligence and fault judgment accuracy. For example, in the case of some environmental interference, the use of solutions based on preset rules is prone to misjudgment, making it difficult for power grid managers to detect potential faults or hidden dangers in a timely manner. Therefore, further improving the accuracy of abnormal monitoring and early warning of measuring switches has become an urgent problem to be solved to ensure the reliable operation of the power grid.

[0023] Figure 1 The present invention is a flowchart of a method for monitoring and warning abnormal status of a measuring switch according to an embodiment of the present invention.

[0024] like Figure 1 As shown, a method for monitoring and warning abnormal state of a measuring switch includes: Step S10: acquiring historical measurement switch parameters, the historical measurement switch parameters including mechanical state parameters, electrical characteristic parameters, environmental parameters and data integrity parameters; preprocessing the historical measurement switch parameters to obtain a measurement switch data set; In this embodiment, according to the normal operation of the measurement switch, the operation status of the measurement switch is recorded in real time, and different types of parameter data are obtained from multiple sensors to form the historical measurement switch parameters.

[0025] Furthermore, the historically measured switch parameters specifically include: The mechanical state parameter, the electrical characteristic parameter, the environmental parameter and the data integrity parameter; The mechanical state parameters include switch action time, contact vibration frequency and switch positioning state; The electrical characteristic parameters include current, voltage, active power, reactive power and contact resistance; The environmental parameters include temperature, humidity, dust concentration and air pressure; The data integrity parameters include signal power, signal acquisition time, data loss amount and noise power.

[0026] Specifically, in this embodiment, the switch action time is obtained by collecting control signal feedback, and the vibration condition of the contact can be monitored in real time by a vibration sensor installed at the contact position. The open or closed state of the switch can be detected by a position sensor, so that poor contact problems caused by wear or aging can be discovered in time, thereby avoiding the occurrence of electrical failures.

[0027] The electrical characteristic parameters are obtained by measuring the current and voltage sensors equipped on the switch. Electrical parameters such as current, voltage, active power and reactive power directly reflect the operating status of the power system and are also an important basis for judging whether there is an abnormality in the measuring switch. Based on these parameters, the operating efficiency of the measuring switch and the health status of the internal components can be further inferred.

[0028] Environmental parameters are obtained by sensors such as ambient temperature and humidity installed around the measuring switch. Environmental factors such as temperature, humidity, dust concentration and air pressure directly affect the performance of the switch. For example, a sharp drop in temperature may cause the performance of the measuring switch to deteriorate. Monitoring these environmental parameters can help evaluate the operating status of the measuring switch and provide early warning.

[0029] By analyzing the data transmission records between the measurement switch and the monitoring system, data integrity parameters can be obtained. The monitoring data of the measurement switch needs to ensure high integrity and stability, especially indicators such as signal power, data loss and noise power, which can help evaluate problems in the data transmission process and help to detect abnormalities such as communication in a timely manner.

[0030] By acquiring multi-dimensional data, the operating status of the measuring switch is comprehensively analyzed from multiple angles such as mechanical status, electrical characteristics, environmental impact, and data integrity. This can not only accurately determine whether there is an abnormality in the measuring switch, but also locate the specific cause of the abnormality. For example, mechanical status parameters can reflect changes in the physical properties of the measuring switch, helping to achieve comprehensive abnormality monitoring, thereby improving the accuracy of abnormal status monitoring of the measuring switch.

[0031] Furthermore, the historical measured switch parameters are de-noised and missing values ​​are filled to obtain a measured switch data set, which can remove noise data caused by electromagnetic interference or hardware failure, ensure data integrity, and facilitate subsequent analysis.

[0032] Step S20: extracting features from the measurement switch data set, and calculating measurement switch features; Furthermore, the feature extraction process includes: Setting a timestamp, wherein the timestamp is used to obtain the historical measurement switch parameters; Among them, in this embodiment, the generation interval of the timestamp of the mechanical state parameter is 1 second, the generation interval of the timestamp of the electrical characteristic parameter is 10 seconds, the generation interval of the timestamp of the environmental parameter is 30 minutes, and the generation interval of the timestamp of the data integrity parameter is 10 minutes, and a total of 10 days of historical measurement switch parameters are obtained.

[0033] The measurement switch characteristics are calculated based on the historical measurement switch parameters.

[0034] Figure 2 It is a schematic diagram of the structure of measuring switch characteristics proposed in an embodiment of the present invention.

[0035] like Figure 2 As shown, the measurement switch characteristics include mechanical state characteristics, electrical characteristic characteristics, line loss characteristics, comprehensive environment characteristics and data signal characteristics; the mechanical state characteristics include abnormal action time characteristics, abnormal contact vibration characteristics and abnormal positioning state characteristics; the electrical characteristic characteristics include pressure loss characteristics, current loss characteristics and current imbalance characteristics; the data signal characteristics include abnormal time synchronization characteristics, abnormal data loss rate characteristics and abnormal signal quality characteristics; The calculation process of measuring the switch characteristics includes: The mean, maximum and standard deviation of the switch action time are calculated to obtain the time anomaly characteristics, which are expressed as: ; ; ; in, is the mean switching time, is the total number of switching actions, For the The time interval between switching actions, is the standard deviation of the switch action time, is the maximum value of the switch action time, is the maximum value function, It is the serial number of the number of switch actions.

[0036] The frequency spectrum characteristics of the contact vibration frequency are calculated using Fourier transform, and the main frequency is extracted from the frequency spectrum characteristics to obtain the abnormal contact vibration characteristics, which are expressed as: ; in, is the main frequency, is the spectrum value obtained by Fourier transform, is the frequency value, The maximum function is used to return the spectrum value Maximum frequency value .

[0037] The proportion of the incompletely closed state in the switch positioning state is calculated to obtain the abnormal characteristics of the positioning state, which is expressed as: ; in, is the proportion of the incompletely closed state, is the number of incomplete closures, is the total number of switches.

[0038] Extracting the minimum values ​​of current and voltage respectively to obtain the current loss characteristic and the voltage loss characteristic; The current imbalance rate is calculated according to the current to obtain the current imbalance characteristic, which is expressed as: ; in, is the current unbalance rate, is the three-phase current value, is the mean value of the three-phase current, is the maximum value function, is the minimum function.

[0039] Among them, the calculation of three-phase current usually involves the current values ​​of three phases A, B and C. Three-phase imbalance may cause equipment overload, damage or energy waste. By calculating the current imbalance rate, fault problems in the power system can be identified in time, so as to determine whether there is an abnormality in the operating state of the measuring switch. In addition, in this embodiment, the loss of current characteristics and the root mean square current are monitored for a single phase.

[0040] The root mean square current is calculated according to the current, and the line loss characteristic is obtained according to the root mean square current and the contact resistance, which is expressed as: ; ; in, is the root mean square current, reflecting the comprehensive intensity of the current; For the Current value; is the total number of sampling points; is the contact resistance; is the line loss power; It is the serial number of the current sampling point.

[0041] The mean values ​​of temperature, humidity, dust concentration and air pressure are calculated, and the weighted sum of the mean values ​​is performed to obtain the comprehensive environmental characteristics, which is expressed as: ; in, For comprehensive environmental characteristics, are the mean values ​​of temperature, humidity, dust concentration and air pressure, respectively. is the weight of each parameter; Among them, in this embodiment, temperature has the greatest impact on the measurement switch, so the weights of various parameters are set to 0.4, 0.3, 0.2 and 0.1 respectively, and can be adjusted according to the actual operating environment.

[0042] According to the signal acquisition time, the time synchronization anomaly feature is calculated and expressed as: ; in, It is the abnormal characteristic of time synchronization. is the number of signals, For the The timestamp of signal acquisition, For the The actual acquisition time of the signal, is the sequence number of the signal sampling point.

[0043] According to the data loss amount, signal power and noise power, the data loss rate abnormality feature and the signal quality abnormality feature are calculated respectively and expressed as: ; ; in, is the abnormal characteristic of data loss rate, is the number of lost packets, is the total number of packages, Abnormal signal quality characteristics, is the signal power, is the noise power, This is the logarithm with base 10.

[0044] The mechanical state characteristics calculated based on historical measured switch parameters can detect abnormalities in the switch mechanical parts in advance and reduce the downtime of the measured switch. The electrical characteristics and line loss characteristics can evaluate circuit abnormalities and provide data support for optimization such as load distribution. Comprehensive environmental characteristics and data signal characteristics can be used to identify problems in data collection and ensure the timeliness and accuracy of monitoring data. At the same time, the calculation results of these characteristics can help set monitoring rules. For example, the threshold of contact vibration frequency or current imbalance rate is set according to historical data to form targeted abnormality identification standards, thereby improving the accuracy of abnormal state monitoring of the measured switch.

[0045] Step S30: constructing a monitoring rule according to the measurement switch feature and expert experience, wherein the monitoring rule is used to set a corresponding feature threshold for each measurement switch feature; Establishing an abnormal state monitoring model, wherein the abnormal state monitoring model is used to determine whether the measuring switch is in an abnormal state; The characteristic threshold and the abnormal state monitoring model are encapsulated and imported into an expert knowledge base.

[0046] Specifically, taking the current imbalance rate as an example, the average current imbalance characteristic is calculated to be 20.63% based on the historical measurement of switch parameters. Combined with expert experience, the characteristic threshold of the current imbalance rate is set as: Abnormal state: greater than 20%; Warning status: greater than 10% and less than or equal to 20%; Normal state: less than or equal to 10%.

[0047] Among them, the warning state is to identify potential fault risks in advance so that necessary measures can be taken for inspection or prevention, while the abnormal state means that the operation of the measuring switch has deviated from the normal range, which may have a serious impact on the operating efficiency or safety of the power grid. Therefore, the threshold of the abnormal state will be further optimized in the later stage to improve the accuracy of monitoring the abnormal state of the measuring switch.

[0048] The current imbalance rate is used as an input feature and compared with the characteristic threshold. If the calculated current imbalance rate is greater than 20%, the output state is "abnormal", if the calculated current imbalance rate is greater than 10% and less than or equal to 20%, the output state is "warning", and if the calculated current imbalance rate is less than or equal to 10%, the output state is "normal". The calculation formula of the current imbalance rate, the characteristic threshold and the abnormal state monitoring model are encapsulated and imported into the expert knowledge base for real-time monitoring of the measurement switch.

[0049] Step S40: obtaining real-time measurement switch parameters, and using a machine learning model to predict future measurement parameters based on the historical measurement switch parameters and the real-time measurement switch parameters; Among them, the machine learning model can be a random forest model, a decision tree model, a support vector machine model, and a long short-term memory network (LSTM) model, etc.

[0050] The expert knowledge base is used to predict the real-time measurement switch parameters and the future measurement parameters, and a primary judgment result and a secondary judgment result are output.

[0051] Specifically, the LSTM model is good at processing time series data and can learn the dynamic patterns of historical data. Therefore, in this embodiment, the LSTM model is selected to predict the measured parameters. Taking the current unbalance rate as an example, the three-phase current (phase A current) within 10 days is collected. , B phase current , C phase current ), which is input into the LSTM model. The LSTM model outputs the three-phase current at the next moment (the future A phase current , B phase current , C phase current ). The current three-phase current ( =100, =95, =80), the three-phase current at the next moment ( =105, =85, =90) is input into the expert knowledge base, and the first-level judgment result is "abnormal" and the second-level judgment result is "warning". The alarm is triggered based on the second-level judgment result, and maintenance personnel are advised to take measures before more serious problems may occur in the future.

[0052] Step S50: obtaining the actual operation status of the measuring switch at a predetermined time interval, and comparing the actual operation status with the primary judgment result and the secondary judgment result to obtain a comprehensive false alarm rate and a comprehensive missed alarm rate; The actual operation status refers to the actual state of the measured switch during operation. The manual inspection and self-test functions record abnormal points, such as the switch not closed and the contact stuck. By collecting the system judgment results, including the first-level judgment results and the second-level judgment results, and comparing the actual operation status, a confusion matrix is ​​formed. According to the data of the confusion matrix, the false alarm rate and the missed alarm rate are calculated, which is expressed as: ; ; in, is the false alarm rate; is the underreporting rate; For the expert knowledge base to determine anomalies, the situation is actually abnormal; The expert knowledge base determines that the abnormal situation is actually normal; This is a situation where the expert knowledge base determines that it is normal, but it is actually abnormal.

[0053] Furthermore, the predetermined time interval increases as the number of optimization times of the expert knowledge base increases, which is expressed as: ; in, For the suboptimal predetermined time interval, is the initial predetermined time interval, is an exponential function, is the attenuation coefficient, is the number of optimizations.

[0054] Specifically, as the number of knowledge base optimizations increases, the monitoring rules become more complete, the model prediction accuracy improves, and the false positives and false negatives of abnormal switch features are reduced. The collection time interval can be appropriately extended to reduce system resource consumption. Set to 5 minutes, decay factor 0.2, after the first optimization, the scheduled time interval For about 6 minutes, after the second optimization, the scheduled time interval About 7 minutes and 30 seconds, after the fifth optimization, the scheduled time interval It takes about 12 minutes and 15 seconds to avoid redundant calculations caused by oversampling. It not only improves the intelligence level and accuracy of measurement switch monitoring, but also achieves a balance between optimizing performance and reducing resource consumption.

[0055] Step S60: optimizing the expert knowledge base using a genetic algorithm according to the comprehensive false alarm rate and the comprehensive omission rate to obtain an updated expert knowledge base.

[0056] Furthermore, the optimization step of the expert knowledge base includes: S61: Setting an initial population, wherein the initial population includes the characteristic threshold and randomly generated threshold individuals; S62: Setting a fitness function, wherein the fitness function is expressed as: ; in, is the fitness value, is the weight coefficient, is the weight coefficient, is the false alarm rate, is the underreporting rate, is an exponential function, is the current threshold obtained during the optimization process, is the feature threshold, is the smoothing factor; S63: Evaluate the initial population according to the fitness function to obtain the fitness value; S64: According to the fitness value, select individuals whose fitness value is greater than a preset threshold value for crossover and mutation to obtain new threshold individuals; S65: updating the initial population according to the new threshold individuals; S66: If the number of optimizations reaches the predetermined number of iterations, the optimization is terminated and a new feature threshold is output; otherwise, S63 to S66 are repeated.

[0057] Specifically, in this embodiment, the initial population size is 50 individuals, the feature threshold in the expert knowledge base is obtained, and the other 49 individuals are randomly generated in the corresponding interval. The weight coefficients are set to 0.4, 0.4 and 0.2 respectively, the smoothing factor is set to 1, and the fitness values ​​of all 50 individuals are calculated respectively, and they are sorted according to the fitness values. The individuals with the top 20% of the fitness values ​​are selected to cross each other to generate new individuals, and 10% of the individuals are mutated to obtain new threshold individuals. The new threshold individuals replace the individuals with low fitness values, and the initial population is updated. When the predetermined number of iterations reaches 100 times, the optimal threshold is output. By optimizing the feature threshold, after the fifth optimization, the false alarm rate is effectively reduced by 3% and the missed alarm rate is 2%, the accuracy of the measurement switch monitoring is improved, and it can adapt to the feature changes under different operating environments to ensure the stable operation of the measurement switch.

[0058] The present invention comprehensively analyzes and extracts features of historical measurement switch parameters, combines expert experience to construct monitoring rules and define feature thresholds, effectively reducing the subjectivity of threshold setting and improving the sensitivity of abnormal state monitoring. The machine learning model is introduced to accurately predict future measurement parameters, and early warning of abnormal states is achieved, thereby reducing safety hazards caused by sudden abnormalities. Based on the dual judgment of real-time measurement parameters and predicted parameters, the comprehensive false alarm rate and missed alarm rate are obtained, further improving the reliability and accuracy of monitoring. The expert knowledge base is optimized through genetic algorithms, and fitness functions are used for evaluation and optimization iterations, so that the monitoring rules and monitoring models are continuously optimized in long-term operation and adapt to the dynamic changes of the actual operating environment, thereby improving the accuracy of measurement switch monitoring and building an efficient and reliable measurement switch abnormal state monitoring system.

[0059] Embodiment 2 is as follows: A power supply company is responsible for managing a large number of measuring switches, which are operated in a complex workshop environment with high humidity and high dust concentration for a long time. The company found that some measuring switches were not completely closed, resulting in reduced grid operation efficiency and potential safety hazards. To solve this problem, the application of the present invention can improve the accuracy of measuring switch monitoring, reduce the missed reporting rate of abnormal monitoring, and realize early warning of abnormal conditions.

[0060] Figure 3 A schematic diagram of the structure of an abnormal state monitoring and early warning system for a measuring switch proposed in an embodiment of the present invention; like Figure 3 As shown, a monitoring and early warning system for abnormal state of a measuring switch comprises: A parameter preprocessing module, used to obtain historical measurement switch parameters, preprocess the historical measurement switch parameters, and obtain a measurement switch data set; A feature extraction module, used to extract features from the measurement switch data set and calculate measurement switch features; An expert knowledge base establishment module is used to construct monitoring rules based on the measurement switch characteristics and expert experience, wherein the monitoring rules are used to set corresponding feature thresholds for each measurement switch characteristic; establish an abnormal state monitoring model, wherein the abnormal state monitoring model is used to determine whether the measurement switch is in an abnormal state; encapsulate the feature thresholds and the abnormal state monitoring model, and import them into the expert knowledge base; The abnormal monitoring module obtains the real-time measurement switch parameters, and uses the machine learning model to predict the future measurement parameters according to the historical measurement switch parameters and the real-time measurement switch parameters; uses the expert knowledge base to predict the real-time measurement switch parameters and the future measurement parameters, and outputs the primary judgment result and the secondary judgment result; Among them, in this embodiment, the machine learning model uses an LSTM model, and the measurement switch data set is divided into 80% and 20% for training and testing of the LSTM model.

[0061] Specifically, the LSTM model has 2 hidden layers, 64 LSTM units in each layer, ReLU activation function, 0.001 learning rate, and 64 batch size. After 100 rounds of training, all parameters are predicted and the accuracy is averaged. The accuracy can meet the requirements. The ARIMA model and SVR model are used as comparison models. The results are shown in Table 1. It can be seen that the LSTM model shows the highest prediction accuracy and the lowest error in both training and testing stages.

[0062] Table 1 Comparison of model results

[0063] The expert knowledge base optimization module obtains the actual operation status of the measuring switch at predetermined time intervals, and compares the actual operation status with the primary judgment result and the secondary judgment result to obtain a comprehensive false alarm rate and a comprehensive missed alarm rate; based on the comprehensive false alarm rate and the comprehensive missed alarm rate, a genetic algorithm is used to optimize the expert knowledge base to obtain an updated expert knowledge base.

[0064] Furthermore, the historically measured switch parameters specifically include: Mechanical state parameters, electrical characteristic parameters, environmental parameters and data integrity parameters; The mechanical state parameter, the electrical characteristic parameter, the environmental parameter and the data integrity parameter; The mechanical state parameters include switch action time, contact vibration frequency and switch positioning state; The electrical characteristic parameters include current, voltage, active power, reactive power and contact resistance; The environmental parameters include temperature, humidity, dust concentration and air pressure; The data integrity parameters include signal power, signal acquisition time, data loss amount and noise power.

[0065] Furthermore, the feature extraction process includes: Setting a timestamp, wherein the timestamp is used to obtain the historical measurement switch parameters; According to the historical measurement switch parameters, the measurement switch characteristics are calculated; the measurement switch characteristics include mechanical state characteristics, electrical characteristic characteristics, line loss characteristics, comprehensive environment characteristics and data signal characteristics; the mechanical state characteristics include abnormal action time characteristics, abnormal contact vibration characteristics and abnormal positioning state characteristics; the electrical characteristic characteristics include pressure loss characteristics, current loss characteristics and current imbalance characteristics; the data signal characteristics include abnormal time synchronization characteristics, abnormal data loss rate characteristics and abnormal signal quality characteristics; Furthermore, the calculation process of measuring the switch characteristics includes: Calculate the mean, maximum and standard deviation of the switch action time to obtain the time anomaly characteristics; Calculate the frequency spectrum characteristics of the contact vibration frequency using Fourier transform, and extract the main frequency from the frequency spectrum characteristics to obtain the abnormal contact vibration characteristics; Calculating the proportion of the incompletely closed state in the switch positioning state to obtain the abnormal characteristics of the positioning state; Extracting the minimum values ​​of the current and the voltage respectively to obtain the current loss characteristic and the voltage loss characteristic; calculating the current unbalance rate according to the current to obtain the current unbalance characteristic; Calculating a root mean square current according to the current, and obtaining the line loss characteristic according to the root mean square current and contact resistance; Calculating the mean values ​​of temperature, humidity, dust concentration and air pressure, and weighted summing the mean values ​​to obtain the comprehensive environmental characteristics; According to the signal acquisition time, the time synchronization anomaly feature is calculated; According to the data loss amount, signal power and noise power, the data loss rate abnormality feature and the signal quality abnormality feature are respectively calculated and obtained.

[0066] Furthermore, the predetermined time interval increases as the number of optimization times of the expert knowledge base increases, which is expressed as: ; in, For the suboptimal predetermined time interval, is the initial predetermined time interval, is an exponential function, is the attenuation coefficient, is the number of optimizations.

[0067] Furthermore, the optimization step of the expert knowledge base includes: S61: Setting an initial population, wherein the initial population includes the characteristic threshold and randomly generated threshold individuals; S62: Setting a fitness function, wherein the fitness function is expressed as: ; in, is the fitness value, is the weight coefficient, is the weight coefficient, is the false alarm rate, is the underreporting rate, is an exponential function, is the current threshold obtained during the optimization process, is the feature threshold, is the smoothing factor; S63: Evaluate the initial population according to the fitness function to obtain the fitness value; S64: According to the fitness value, select individuals whose fitness value is greater than a preset threshold value for crossover and mutation to obtain new threshold individuals; S65: updating the initial population according to the new threshold individuals; S66: If the number of optimizations reaches the predetermined number of iterations, the optimization is terminated and a new feature threshold is output; otherwise, S63 to S66 are repeated.

[0068] Table 2 First optimization results

[0069] Table 3 Optimization result example

[0070] Specifically, in this embodiment, the predetermined number of iterations is set to 40. The first optimization result is shown in Table 2. It can be seen that when the number of iterations is 1 to 10, the maximum fitness value is 0.92, that is, the feature threshold set according to expert experience performs well, but as the number of iterations increases, the fitness value gradually increases, and the genetic algorithm tends to converge, stopping after the 40th optimization, and the final average fitness value reaches 0.95, and the maximum individual fitness value is 0.96, indicating that the threshold optimization effect is significant.

[0071] In this embodiment, the optimization is terminated after 10 optimizations. The results of the 10 optimizations are shown in Table 3. It can be seen that the threshold amplitude of each optimization gradually decreases, and the feature threshold is finally stable and achieves the best effect. At the same time, controlling the number of optimizations within 10 times also avoids unnecessary complexity caused by excessive adjustment.

[0072] In order to prove the effect of the present invention, the optimized feature threshold is compared with the original solution of a company. As shown in Table 4, the optimization comparison results of the missed alarm rate and the false alarm rate are given. It can be seen that the present invention reduces the missed alarm and false alarm rates by optimizing the feature threshold, and also verifies the superiority of the present invention in measuring switch abnormality detection.

[0073] Table 4 Comparison results of optimization of missed alarm rate and false alarm rate

[0074] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and warning abnormal status of a measuring switch, characterized in that: include: Acquiring historical measured switch parameters, wherein the historical measured switch parameters include mechanical state parameters, electrical characteristic parameters, environmental parameters, and data integrity parameters; Preprocessing the historical measurement switch parameters to obtain a measurement switch data set; Extracting features from the measurement switch data set to obtain measurement switch features by calculation; Constructing a monitoring rule according to the measurement switch feature and expert experience, wherein the monitoring rule is used to set a corresponding feature threshold for each measurement switch feature; Establishing an abnormal state monitoring model, wherein the abnormal state monitoring model is used to determine whether the measuring switch is in an abnormal state; Encapsulating the characteristic threshold and the abnormal state monitoring model, and importing them into an expert knowledge base; Acquire real-time measurement switch parameters, and use a machine learning model to predict future measurement parameters based on the historical measurement switch parameters and the real-time measurement switch parameters; Use the expert knowledge base to predict the real-time measurement switch parameters and the future measurement parameters, and output a primary judgment result and a secondary judgment result; Acquire the actual operation status of the measuring switch at a predetermined time interval, and compare the actual operation status with the primary judgment result and the secondary judgment result to obtain a comprehensive false alarm rate and a comprehensive missed alarm rate; According to the comprehensive false alarm rate and the comprehensive omission rate, the expert knowledge base is optimized using a genetic algorithm to obtain an updated expert knowledge base.

2. The abnormal state monitoring and early warning method of a measuring switch according to claim 1 is characterized in that: The historical measurement switch parameters specifically include: The mechanical state parameter, the electrical characteristic parameter, the environmental parameter and the data integrity parameter; The mechanical state parameters include switch action time, contact vibration frequency and switch positioning state; The electrical characteristic parameters include current, voltage, active power, reactive power and contact resistance; The environmental parameters include temperature, humidity, dust concentration and air pressure; The data integrity parameters include signal power, signal acquisition time, data loss amount and noise power.

3. The abnormal state monitoring and early warning method of a measuring switch according to claim 1 is characterized in that: The feature extraction process includes: Setting a timestamp, wherein the timestamp is used to obtain the historical measurement switch parameters; According to the historical measurement switch parameters, the measurement switch characteristics are calculated; the measurement switch characteristics include mechanical state characteristics, electrical characteristic characteristics, line loss characteristics, comprehensive environment characteristics and data signal characteristics; the mechanical state characteristics include abnormal action time characteristics, abnormal contact vibration characteristics and abnormal positioning state characteristics; the electrical characteristic characteristics include pressure loss characteristics, current loss characteristics and current imbalance characteristics; the data signal characteristics include abnormal time synchronization characteristics, abnormal data loss rate characteristics and abnormal signal quality characteristics; The calculation process of measuring the switch characteristics includes: Calculate the mean, maximum and standard deviation of the switch action time to obtain the time anomaly characteristics; Calculate the frequency spectrum characteristics of the contact vibration frequency using Fourier transform, and extract the main frequency from the frequency spectrum characteristics to obtain the abnormal contact vibration characteristics; Calculating the proportion of the incompletely closed state in the switch positioning state to obtain the abnormal characteristics of the positioning state; Extracting the minimum values ​​of the current and the voltage respectively to obtain the current loss characteristic and the voltage loss characteristic; calculating the current unbalance rate according to the current to obtain the current unbalance characteristic; Calculating a root mean square current according to the current, and obtaining the line loss characteristic according to the root mean square current and contact resistance; Calculating the mean values ​​of temperature, humidity, dust concentration and air pressure, and weighted summing the mean values ​​to obtain the comprehensive environmental characteristics; According to the signal acquisition time, the time synchronization anomaly feature is calculated; According to the data loss amount, signal power and noise power, the data loss rate abnormality feature and the signal quality abnormality feature are respectively calculated and obtained.

4. The abnormal state monitoring and early warning method of a measuring switch according to claim 1 is characterized in that: The predetermined time interval is expressed as: ； in, For the suboptimal predetermined time interval, is the initial predetermined time interval, is an exponential function, is the attenuation coefficient, is the number of optimizations.

5. The abnormal state monitoring and early warning method of a measuring switch according to claim 1 is characterized in that: The optimization steps of the expert knowledge base include: S61: Setting an initial population, wherein the initial population includes the characteristic threshold and randomly generated threshold individuals; S62: Setting a fitness function, wherein the fitness function is expressed as: ； in, is the fitness value, , and is the weight coefficient, is the false alarm rate, is the underreporting rate, is an exponential function, is the current threshold obtained during the optimization process, is the feature threshold, is the smoothing factor; S63: Evaluate the initial population according to the fitness function to obtain the fitness value; S64: According to the fitness value, select individuals whose fitness value is greater than a preset threshold value for crossover and mutation to obtain new threshold individuals; S65: updating the initial population according to the new threshold individuals; S66: If the number of optimizations reaches the predetermined number of iterations, the optimization is terminated and a new feature threshold is output; otherwise, S63 to S66 are repeated.

6. A monitoring and early warning system for abnormal status of a measuring switch, characterized in that: include: A parameter preprocessing module, used to obtain historical measurement switch parameters, preprocess the historical measurement switch parameters, and obtain a measurement switch data set; A feature extraction module, used to extract features from the measurement switch data set and calculate measurement switch features; An expert knowledge base establishment module is used to construct monitoring rules based on the measurement switch characteristics and expert experience, wherein the monitoring rules are used to set corresponding feature thresholds for each measurement switch characteristic; establish an abnormal state monitoring model, wherein the abnormal state monitoring model is used to determine whether the measurement switch is in an abnormal state; encapsulate the feature thresholds and the abnormal state monitoring model, and import them into the expert knowledge base; The abnormal monitoring module obtains the real-time measurement switch parameters, and uses the machine learning model to predict the future measurement parameters according to the historical measurement switch parameters and the real-time measurement switch parameters; uses the expert knowledge base to predict the real-time measurement switch parameters and the future measurement parameters, and outputs the primary judgment result and the secondary judgment result; The expert knowledge base optimization module obtains the actual operation status of the measuring switch at a predetermined time interval, and compares the actual operation status with the primary judgment result and the secondary judgment result to obtain a comprehensive false alarm rate and a comprehensive missed alarm rate; According to the comprehensive false alarm rate and the comprehensive omission rate, the expert knowledge base is optimized using a genetic algorithm to obtain an updated expert knowledge base.

7. The abnormal state monitoring and early warning system of a measuring switch according to claim 6, characterized in that: The historical measurement switch parameters specifically include: Mechanical state parameters, electrical characteristic parameters, environmental parameters and data integrity parameters; The mechanical state parameter, the electrical characteristic parameter, the environmental parameter and the data integrity parameter; The mechanical state parameters include switch action time, contact vibration frequency and switch positioning state; The electrical characteristic parameters include current, voltage, active power, reactive power and contact resistance; The environmental parameters include temperature, humidity, dust concentration and air pressure; The data integrity parameters include signal power, signal acquisition time, data loss amount and noise power.

8. The abnormal state monitoring and early warning system of a measuring switch according to claim 6, characterized in that: The feature extraction process includes: Setting a timestamp, wherein the timestamp is used to obtain the historical measurement switch parameters; According to the historical measurement switch parameters, the measurement switch characteristics are calculated; the measurement switch characteristics include mechanical state characteristics, electrical characteristic characteristics, line loss characteristics, comprehensive environment characteristics and data signal characteristics; the mechanical state characteristics include abnormal action time characteristics, abnormal contact vibration characteristics and abnormal positioning state characteristics; the electrical characteristic characteristics include pressure loss characteristics, current loss characteristics and current imbalance characteristics; the data signal characteristics include abnormal time synchronization characteristics, abnormal data loss rate characteristics and abnormal signal quality characteristics; The calculation process of measuring the switch characteristics includes: Calculate the mean, maximum and standard deviation of the switch action time to obtain the time anomaly characteristics; Calculate the frequency spectrum characteristics of the contact vibration frequency using Fourier transform, and extract the main frequency from the frequency spectrum characteristics to obtain the abnormal contact vibration characteristics; Calculating the proportion of the incompletely closed state in the switch positioning state to obtain the abnormal characteristics of the positioning state; Extracting the minimum values ​​of the current and the voltage respectively to obtain the current loss characteristic and the voltage loss characteristic; calculating the current unbalance rate according to the current to obtain the current unbalance characteristic; Calculating a root mean square current according to the current, and obtaining the line loss characteristic according to the root mean square current and contact resistance; Calculating the mean values ​​of temperature, humidity, dust concentration and air pressure, and weighted summing the mean values ​​to obtain the comprehensive environmental characteristics; According to the signal acquisition time, the time synchronization anomaly feature is calculated; According to the data loss amount, signal power and noise power, the data loss rate abnormality feature and the signal quality abnormality feature are respectively calculated and obtained.

9. The abnormal state monitoring and early warning system of a measuring switch according to claim 6, characterized in that: The predetermined time interval is expressed as: ； in, For the suboptimal predetermined time interval, is the initial predetermined time interval, is an exponential function, is the attenuation coefficient, is the number of optimizations.

10. The abnormal state monitoring and early warning system of a measuring switch according to claim 6, characterized in that: The optimization steps of the expert knowledge base include: S61: Setting an initial population, wherein the initial population includes the characteristic threshold and randomly generated threshold individuals; S62: Setting a fitness function, wherein the fitness function is expressed as: ； in, is the fitness value, , and is the weight coefficient, is the false alarm rate, is the underreporting rate, is an exponential function, is the current threshold obtained during the optimization process, is the feature threshold, is the smoothing factor; S63: Evaluate the initial population according to the fitness function to obtain the fitness value; S64: According to the fitness value, select individuals whose fitness value is greater than a preset threshold value for crossover and mutation to obtain new threshold individuals; S65: updating the initial population according to the new threshold individuals; S66: If the number of optimizations reaches the predetermined number of iterations, the optimization is terminated and a new feature threshold is output; otherwise, S63 to S66 are repeated.

Images