Online monitoring method of rectifier cabinet based on machine learning

Abstract

The invention discloses an online monitoring method of a rectifier cabinet based on machine learning, relates to the technical field of rectifier cabinet monitoring, and solves the technical problem that when an existing rectifier monitoring method is suitable for monitoring a rectifier cabinet unit, the monitoring result is not meticulous enough, the timeliness of data processing is poor, and the monitoring efficiency is low. Comprising the following steps: 1, acquiring historical use data of each rectifier cabinet in a rectifier cabinet group and element characteristic data of each element; 2, generating a plurality of sensitivity scores of each element according to the historical use data and the element characteristic data; 3, acquiring a plurality of pieces of environmental monitoring data; generating a priority processing coefficient according to the environmental monitoring data and the sensitivity score of each element; 4, detecting each element according to the priority processing coefficient to obtain a corresponding element state; 5, sorting and displaying the element states; and the overall detection efficiency of the rectifier cabinet is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of rectifier cabinet monitoring, and relates to rectifier cabinet monitoring technology, and specifically to an online monitoring method of a rectifier cabinet based on machine learning. Background Art

[0002] A rectifier is a rectifying device that can convert alternating current (AC) into direct current (DC). This device has two main functions: one is to convert alternating current (AC) into direct current (DC) and supply it to the load after filtering; the other is to supply power to the energy storage system. As part of the power supply system, the operating status of the rectifier is directly related to the stability of the power supply system.

[0003] The prior art (announcement number CN109443419B) discloses a rectifier online monitoring method based on machine learning, including the steps of: using rectifier historical data for modeling, constructing a temperature model of the rectifier under normal operating conditions, including "data processing" and "machine learning" links; rectifier status monitoring and prediction: first use "data processing" to process new data and construct the same characteristic variables, and then pass the processed data to the rectifier temperature model, and the temperature model calculates the predicted value of the rectifier temperature; rectifier status / warning level calculation: compare the rectifier temperature prediction value with the actual value in real time, use interval judgment and dynamic monitoring to judge the rectifier operating status, and give rectifier warning information and status information in real time. The warning information is divided into three intervals: the warning level of the first interval is 0, indicating that the equipment is operating normally; the warning level of the second interval is 1-11, indicating that the equipment status needs "attention"; the warning level of the second interval is 12, indicating that the equipment status is "abnormal".

[0004] The above-mentioned rectifier online monitoring method uses historical operation data to build a temperature model between the current, ambient temperature and humidity, and rectifier temperature during normal operation of the rectifier, and then judges whether the rectifier is abnormal based on the real-time temperature; the rectifier as a whole is monitored for abnormalities, and the rectifier abnormality is generally caused by abnormalities in its internal components. The abnormal components cannot be accurately determined, making the overall monitoring efficiency low; at the same time, when monitoring the rectifier cabinet group, since the amount of data that the rectifier cabinet group needs to process is huge, and the internal components of the rectifier cabinet group are complex and diverse, different components have different characteristics, resulting in different degrees of influence of environmental changes on different components. When using the above method to monitor the rectifier cabinet group, it may cause the status of some components or rectifier cabinets to be delayed in monitoring, thereby reducing the overall monitoring efficiency. Summary of the invention

[0005] The present application aims to solve at least one of the technical problems existing in the prior art; to this end, the present application proposes an online monitoring method for a rectifier cabinet based on machine learning, which is used to solve the technical problem that when the existing rectifier monitoring method is applied to rectifier cabinet unit monitoring, the monitoring results are not detailed enough and the timeliness of data processing is poor, resulting in low monitoring efficiency.

[0006] To achieve the above objectives, the first aspect of the present application provides an online monitoring method for a rectifier cabinet based on machine learning, comprising the following steps: Step 1: Obtain historical usage data of each rectifier cabinet in the rectifier cabinet group and component characteristic data of each component; Step 2: Generate several sensitivity scores for each component based on historical usage data and component characteristic data; Step 3: Obtain some environmental monitoring data; generate priority processing coefficients based on the environmental monitoring data and the sensitivity scores of each component; Step 4: Detect each component according to the priority processing coefficient to obtain the corresponding component status; Step 5: Arrange and display the component status.

[0007] The present application generates a sensitivity score of each component to each monitoring item based on the historical usage data of the rectifier cabinet and the component characteristic data of each component inside the rectifier cabinet; then generates a priority processing coefficient of the corresponding component based on the current environmental monitoring data and several sensitivity scores of each component, analyzes the monitoring data of each component according to the priority processing coefficient to obtain the component status of each component, and finally displays the status of each component to realize online monitoring of the rectifier cabinet; the monitoring results are made more detailed by monitoring the component level, and at the same time, the priority processing coefficient of each component is set to ensure that the detection data of the corresponding component can be processed in time, thereby improving the timeliness of data processing and thus improving the overall detection efficiency of the rectifier cabinet.

[0008] Preferably, generating a plurality of sensitivity scores for each component based on the historical usage data and the component characteristic data comprises: Obtain historical usage data of the rectifier cabinet, and generate a rectifier cabinet impact score corresponding to each rectifier cabinet based on the historical usage data; Acquire component characteristic data, and generate a number of parameter sensitivity scores corresponding to each component according to the component characteristic data; Several sensitivity scores corresponding to various components are generated according to the rectifier cabinet impact score and the parameter sensitivity score.

[0009] Preferably, generating a rectifier cabinet impact score corresponding to each rectifier cabinet according to the historical usage data includes: Extract the usage environment data and usage duration corresponding to each recorded time in the historical usage data; extract each environmental parameter in the usage environment data and mark it as ; Mark the usage time as ; Where i is the number of the environmental parameter and t is the number of the recording time; By formula Calculate and obtain the rectifier cabinet impact score ZMP of the corresponding rectifier cabinet; in, is the optimal parameter value corresponding to the environmental parameter numbered i; is the unit change value corresponding to the environmental parameter numbered i; DT is the unit time; is the proportional coefficient corresponding to the recording time numbered t, which is used to adjust the influence of each use of the rectifier cabinet on the sensitivity of the rectifier cabinet. The specific value is set according to expert experience; It is the adjustment coefficient, which is used to adjust the influence of environmental parameters on the sensitivity of the rectifier cabinet. The specific value is set according to expert experience. is the proportional coefficient 2 of the environmental parameter numbered i, which is used to adjust the influence of different environmental parameters on the sensitivity of the rectifier cabinet during the same use. The specific value is set according to expert experience; i is a positive integer, and i∈[1,I]; t is a positive integer, and t∈[1,T].

[0010] Preferably, the generating of a plurality of parameter sensitivity scores corresponding to each component according to the component characteristic data includes: Extract the electrical characteristic data and physical characteristic data from the component characteristic data, input the electrical characteristic data and physical characteristic data into the parameter sensitivity scoring model to obtain the initial parameter sensitivity score under the corresponding monitoring item; obtain the initial parameter sensitivity score corresponding to each monitoring item, number it, and mark it as ; n is the number of the monitoring project; Obtain component usage data corresponding to each component; extract the recording time in the component usage data, as well as several corresponding working parameters and usage duration; the working parameters include current parameters, voltage parameters and other parameters related to component operation; generate the influence coefficient of the working parameter corresponding to each recording time according to the recording time, working parameters and usage duration; By formula Calculate the parameter sensitivity score corresponding to the monitoring item numbered n ;in, For the number The proportional coefficient corresponding to the recording time is three; is a positive integer, and ; is the total number of operating parameters of the corresponding component; is a positive integer, and , The total number of times the corresponding component usage data is recorded.

[0011] Preferably, the parameter sensitivity scoring model is obtained by training an artificial intelligence model, including: Acquire electrical characteristic data and physical characteristic data of a plurality of components, and their corresponding initial parameter sensitivity scores, and integrate the electrical characteristic data, physical characteristic data, and the initial parameter sensitivity scores into a plurality of sets of training data and test data; The artificial intelligence model is trained using training data, and the trained artificial intelligence model is tested using test data; a parameter sensitivity scoring model is obtained whose input is electrical characteristic data and physical characteristic data, and whose output is an initial parameter sensitivity score; wherein the artificial intelligence model includes a BP neural network model and an RBF neural network model.

[0012] It can be understood that in this embodiment, initial parameter sensitivity scores corresponding to multiple monitoring items can be generated by training one model; it can also be achieved by training multiple models, with one model being used to generate an initial sensitivity score corresponding to one monitoring item; it can be understood that since the initial sensitivity scores of each component to each monitoring item are generated by the electrical and physical properties of the component itself, the initial sensitivity scores of the same component to the same monitoring item are the same, so this work can be performed before the implementation of the online monitoring method for the rectifier cabinet to reduce the amount of data processing during the implementation of the online monitoring method for the rectifier cabinet, thereby ensuring the efficiency of the implementation of the online monitoring method for the rectifier cabinet.

[0013] Preferably, generating the influence coefficient of the working parameter corresponding to each recording time according to the recording time, the working parameter and the usage duration includes: Mark the working parameters as ; Use duration is marked as ;in, is the number of the working parameter corresponding to the component numbered j; It is the number of each recorded time in the component usage data; By formula The calculated number is Number of records within the time period The influence coefficient corresponding to the working parameters ;in, For the component numbered j, The set upper limit threshold value corresponding to the working parameter; For the component numbered j, The set lower limit threshold value corresponding to the working parameter; For the component numbered j, The set unit change value corresponding to the working parameter; GY is the set inherent influence coefficient; , and It is the parameter adjustment coefficient, which is used to adjust the influence of different working parameters on component sensitivity under different conditions. The specific value is set according to expert experience.

[0014] Preferably, the generating of a plurality of sensitivity scores corresponding to each component according to the rectifier cabinet impact score and the parameter sensitivity score includes: Obtain the rectifier cabinet impact score of each rectifier cabinet, and several parameter sensitivity scores corresponding to each component in the rectifier cabinet; obtain the product of the rectifier cabinet impact score and the parameter sensitivity score corresponding to any monitoring item of any component inside it, and use the product as the sensitivity score of the component under the monitoring item, and obtain the sensitivity score corresponding to each component under each monitoring item in turn.

[0015] Preferably, generating a priority processing coefficient according to the environmental monitoring data and the sensitivity score of each component includes: Extract the parameter values ​​corresponding to each monitoring item in the environmental monitoring data; calculate the difference between the parameter values ​​of each monitoring item and the parameter values ​​in the previous monitoring data, and record it as the change amount of the corresponding monitoring item, and mark it as ; Get the sensitivity score of any component under each monitoring item and mark it as ; By formula The priority processing coefficient YC of the element is calculated; wherein, It is the previous parameter value corresponding to the monitoring item numbered n; n is a positive integer, and n∈[1,N], N is the total number of monitoring items.

[0016] Preferably, the detecting each component according to the priority processing coefficient to obtain the corresponding component state includes: The working parameters of each component are obtained in order from large to small according to the priority processing coefficient, and it is judged whether each working parameter is within the corresponding set lower limit threshold and set upper limit threshold; if yes, the working parameter state corresponding to the working parameter is marked as a normal parameter state; if no, the working parameter state corresponding to the working parameter is marked as an abnormal parameter state; Obtain the working parameter status corresponding to each working parameter corresponding to the component in sequence; determine whether each working parameter status has a parameter abnormal state; if yes, mark the component status corresponding to the component as an abnormal state; if no, mark the component status corresponding to the component as a normal state.

[0017] Preferably, the displaying of the component status includes: Determine whether there is an abnormal state in the current component state. If yes, set the rectifier cabinet state to an abnormal state; if no, set the rectifier cabinet state to a normal state; organize the states of each rectifier cabinet in the rectifier cabinet group into the corresponding rectifier cabinet state sub-interface, and organize each rectifier cabinet interface into a rectifier cabinet monitoring interface; there are several rectifier cabinet state sub-interfaces in the rectifier cabinet monitoring interface, and the rectifier cabinet state sub-interface displays the rectifier cabinet state; The component states belonging to a rectifier cabinet are sorted into the component state interface of the corresponding rectifier cabinet in real time, and the rectifier cabinet state sub-interface can be switched to the corresponding component state sub-interface; the component state interface includes a number of components and their corresponding component states.

[0018] Compared with the prior art, the beneficial effects of this application are: 1. The present application generates a sensitivity score of each component to each monitoring item based on the historical usage data of the rectifier cabinet and the component characteristic data of each component inside the rectifier cabinet; then generates a priority processing coefficient of the corresponding component based on the current environmental monitoring data and several sensitivity scores of each component, analyzes the monitoring data of each component according to the priority processing coefficient to obtain the component status of each component, and finally displays the status of each component to realize online monitoring of the rectifier cabinet; the monitoring results are made more detailed by monitoring the component level, and at the same time, the priority processing coefficient of each component is set to ensure that the detection data of the corresponding component can be processed in time, thereby improving the timeliness of data processing and thereby improving the overall detection efficiency of the rectifier cabinet.

[0019] 2. The present application considers the impact of the historical usage of a component on the component's ability to cope with certain environmental factors into the sensitivity score corresponding to the component, making the component sensitivity score more consistent with the specific usage of the component. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 Schematic diagram of the overall steps of the rectifier cabinet online monitoring method in this application; Figure 2 Flowchart for generating sensitivity scores for components in this application. DETAILED DESCRIPTION

[0022] The technical solution of the present application will be described clearly and completely in conjunction with the embodiments below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0023] See also Figure 1-Figure 2 The first aspect of the present application provides an online monitoring method for a rectifier cabinet based on machine learning, comprising the following steps: Step 1: Obtain the historical usage data of each rectifier cabinet in the rectifier cabinet group and the component characteristic data of each component; the historical usage data of the rectifier cabinet is the environmental data recorded by the rectifier cabinet during the previous use process and the corresponding recording time, recording duration and other data. The environmental data includes the overall temperature and overall humidity of the rectifier cabinet monitored by the rectifier cabinet during operation, and other factors affecting the operation of the rectifier cabinet; the component characteristic data is the data related to the corresponding component to various usage environments, including electrical characteristic data and physical characteristic data. The electrical characteristic data is the electrical characteristic parameters related to the sensitivity of the component to the corresponding environmental factors, including the rated current parameters, rated voltage parameters and resistance parameters of the component; the physical characteristic data is the physical characteristic parameters such as the constituent materials and the structure of the component itself related to the sensitivity of the component to the corresponding environmental factors; Step 2: Generate several sensitivity scores for each component based on historical usage data and component characteristic data. The sensitivity score can be understood as the ability of the component to cope with certain environmental factors, including temperature, humidity, and dust, generated based on the overall characteristics of the rectifier cabinet of the component and the characteristics of the component itself. For example, the ability of the component to cope with temperature changes. The smaller the ability of the component to cope with temperature changes, the more sensitive the component is to temperature changes, and the larger the value of the corresponding sensitivity score. Specifically, it includes: obtaining the historical usage data of the rectifier cabinet, and generating the rectifier cabinet impact score corresponding to each rectifier cabinet according to the historical usage data; the rectifier cabinet impact score is the evaluation of the current state of the rectifier cabinet according to the historical operating environment of the rectifier cabinet; obtaining the component characteristic data, and generating a number of parameter sensitivity scores corresponding to each component according to the component characteristic data; the parameter sensitivity score is the score of the component's ability to cope with changes in certain environmental factors; and generating a number of sensitivity scores corresponding to each component according to the rectifier cabinet impact score and the parameter sensitivity score; Step 3: Obtain some environmental monitoring data; generate a priority processing coefficient based on the environmental monitoring data and the sensitivity scores of each component; the priority processing coefficient is the priority of each component detection data processing generated based on the current environmental monitoring data and the current response of each component to the changes in environmental factors in each environmental monitoring data, that is, the response to the sensitivity scores of each monitoring item. The higher the priority processing coefficient, the greater the impact of the current environmental detection data on the component, the greater the probability of abnormality of the component, and the corresponding detection data needs to be processed with priority; Step 4: Detect each component according to the priority processing coefficient to obtain the corresponding component state; the component state is the current state of the component, including normal state and abnormal state; Step 5: Arrange and display the component status.

[0024] This embodiment generates a sensitivity score of each component to each monitoring item based on the historical usage data of the rectifier cabinet and the component characteristic data of each component inside the rectifier cabinet; then generates a priority processing coefficient of the corresponding component based on the current environmental monitoring data and several sensitivity scores of each component, analyzes the monitoring data of each component according to the priority processing coefficient to obtain the component status of each component, and finally displays the status of each component to achieve online monitoring of the rectifier cabinet; the monitoring results are made more detailed by monitoring the component level, and at the same time, the priority processing coefficient of each component is set to ensure that the detection data of the corresponding component can be processed in time, thereby improving the timeliness of data processing and further improving the overall detection efficiency of the rectifier cabinet.

[0025] Generate the rectifier impact score corresponding to each rectifier cabinet based on the historical usage data, including: extracting the usage environment data and usage duration corresponding to each recording time in the historical usage data; extracting each environmental parameter in the usage environment data and marking it as ; Mark the usage time as , the usage time is the usage time of the rectifier cabinet under the environmental parameters; where i is the number of the environmental parameter, t is the number of the recording time, and the earlier the recording time, the smaller the corresponding number t; By formula Calculate and obtain the rectifier cabinet impact score ZMP of the corresponding rectifier cabinet; in, is the optimal parameter value corresponding to the environmental parameter numbered i; is the unit change value corresponding to the environmental parameter numbered i. The specific value is set according to expert experience. Since the units and change degrees of different environmental parameters are different, the unit change values ​​corresponding to different environmental parameters should be different. It is mainly used to eliminate some dimensions in the formula and adjust the influence of different environmental parameters. DT is the unit time. The specific value is set according to experience. is the proportional coefficient 1 corresponding to the recording time numbered t, which is used to adjust the influence of each use of the rectifier cabinet on the sensitivity of the rectifier cabinet. The specific value is set according to expert experience. In this embodiment, the proportional coefficient 1 is set according to the use time. The longer the use time, the larger the corresponding proportional coefficient 1. When a certain environmental parameter appears for a short time, its influence on the rectifier cabinet can be ignored, that is, when the use time is less than the set minimum time threshold, the corresponding proportional coefficient 1 is set to 0; It is the adjustment coefficient, which is used to adjust the influence of environmental parameters on the sensitivity of the rectifier cabinet. The specific value is set according to expert experience. is the proportional coefficient 2 of the environmental parameter numbered i, which is used to adjust the influence of different environmental parameters on the sensitivity of the rectifier cabinet during the same use. The specific value is set according to expert experience; i is a positive integer, and i∈[1,I], I is the total number of environmental parameters; t is a positive integer, and t∈[1,T], T is the total number of total recording time of historical usage data.

[0026] This embodiment calculates the rectifier impact score of the environment on each rectifier cabinet during the use of the rectifier cabinet through the above formula. If the rectifier cabinet has been used in an inappropriate environment before, the overall state of the rectifier cabinet at this time, including the line state inside the rectifier cabinet, will be reduced, resulting in its ability to cope with the environment becoming weaker, that is, its sensitivity to the environment becomes stronger; the corresponding rectifier cabinet impact score is greater.

[0027] Generate several parameter sensitivity scores corresponding to each component based on the component characteristic data, including: extracting electrical characteristic data and physical characteristic data from the component characteristic data, inputting the electrical characteristic data and physical characteristic data into the parameter sensitivity scoring model to obtain the initial parameter sensitivity score under the corresponding monitoring project; the monitoring project includes several environment-related projects including ambient temperature and humidity; the initial parameter sensitivity score can be understood as the ability of the component's own characteristics to cope with changes in a certain environmental factor. For example, if the component itself has a heat dissipation structure, the component has a stronger ability to cope with temperature and is less sensitive to changes in ambient temperature, and the initial parameter sensitivity score corresponding to the temperature is lower; for example, the voltage and current that a component can withstand are small, and the voltage and current that it can withstand are small. When the voltage or current is large, the temperature will change, causing its performance to deteriorate. In this case, the component will have a weaker ability to cope with temperature changes, and will be more sensitive to changes in ambient temperature. Its initial parameter sensitivity score corresponding to temperature will be higher. For example, if a component itself has a waterproof peritoneal structure, then the component will have a stronger ability to cope with changes in humidity, and will be less sensitive to changes in ambient humidity. Its initial parameter sensitivity score corresponding to humidity will be lower. The higher the initial parameter sensitivity score, the higher the sensitivity of the component to changes in the corresponding parameter. That is, the higher the initial parameter sensitivity score, the higher the probability of damage to the component will be when the corresponding parameter changes slightly. Obtain the initial parameter sensitivity score corresponding to each monitoring item, number it, and mark it as ; n is the number of the monitoring project; Obtain component usage data corresponding to each component; extract the recording time in the component usage data, as well as several corresponding working parameters and usage duration; the working parameters include parameters related to component operation such as current parameters and voltage parameters; generate the influence coefficient of the working parameter corresponding to each recording time according to the recording time, working parameters and usage duration, and the influence coefficient is the degree of influence of the historical usage of the corresponding component on the ability of the component to cope with certain environmental factors; specifically: mark the working parameter as ; Use duration is marked as ;in, is the number of the working parameter corresponding to the component numbered j; It is the number of each recorded time in the component usage data; By formula The calculated number is Number of records within the time period The influence coefficient corresponding to the working parameters ;in, For the component numbered j, The set upper limit threshold value corresponding to the working parameter; For the component numbered j, The set lower limit threshold value corresponds to the working parameter; the set upper limit threshold value and the set lower limit threshold value are the normal working range of the corresponding working parameter; For the component numbered j, The set unit change value corresponding to the working parameter. The specific value is set according to expert experience. Since the units and change degrees of different working parameters are different, the unit change values ​​corresponding to different working parameters should be different. It is mainly used to eliminate some dimensions in the formula and adjust the influence of different working parameters. GY is the set inherent influence coefficient, which can be understood as the aging of the component within the unit time of operation within the normal parameter range. The specific value is obtained by experience or by experiment. , and It is the parameter adjustment coefficient, which is used to adjust the influence of different working parameters on component sensitivity under different conditions. The specific value is set according to expert experience.

[0028] This embodiment calculates the influence coefficient of the historical usage of the component on the component's ability to cope with certain environmental factors through the above formula; when the difference between the working parameters of the component and the normal working parameter range is greater, the influence on the aging and damage of the component is greater, and at the same time, the longer the working time in this state, the more obvious the effect on the aging and damage of the component is, and the larger the value of the corresponding influence coefficient is set at this time; so that the influence coefficient can be taken into account in the component sensitivity score later, so that the component sensitivity score is more in line with the specific usage of the component.

[0029] By formula Calculate the parameter sensitivity score corresponding to the monitoring item numbered n ;in, For the number The proportional coefficient corresponding to the recording time is three; is a positive integer, and , is the total number of working parameters; is a positive integer, and , The total number of times the corresponding component usage data is recorded.

[0030] This embodiment takes into account the historical usage of the component into account in the component's sensitivity to the environment through the above formula. When a component is often used in an abnormal working state, the degree of aging and damage of the component will most likely increase. At this time, the component's ability to cope with environmental changes will be weaker than that of a new component. That is, its sensitivity to the environment will increase. Therefore, this embodiment adjusts the initial parameter sensitivity score of the component according to the usage of the component through the above formula to obtain the parameter sensitivity score of the current component. The parameter sensitivity score of the component is adjusted to match the current state of the component.

[0031] It is understandable that when a component is replaced; if the corresponding component is a new component or an updated component, the corresponding component usage data will be cleared, and if the replaced component is an old component, the corresponding component usage data will become the usage data of the updated component.

[0032] The parameter sensitivity scoring model is obtained through artificial intelligence model training, including: obtaining electrical characteristic data and physical characteristic data of a number of components, and their corresponding initial parameter sensitivity scores, wherein the initial parameter sensitivity scores are experts' evaluations of the components' ability to cope with changes in corresponding monitoring items based on the components' electrical and physical characteristics. The worse the component's ability to cope with changes in a certain monitoring item, the more sensitive the component is to the parameter, and the greater the value of the corresponding initial parameter sensitivity score; The initial parameter sensitivity score in this embodiment includes a temperature sensitivity score and a humidity sensitivity score; the temperature sensitivity score is an expert's evaluation of the component's ability to cope with temperature changes based on the component's electrical and physical properties. The worse the component's ability to cope with temperature changes, the more sensitive the component is to temperature, and the higher the corresponding temperature sensitivity score. When evaluating the temperature sensitivity score, the electrical characteristic data at this time are the electrical characteristic parameters of the component related to temperature sensitivity, including the rated current parameter, rated voltage parameter, and resistance parameter of the component, etc. The physical characteristic data are the physical characteristic parameters of the component's constituent materials and the component's own structure related to temperature sensitivity. ; The humidity sensitivity score is an expert's evaluation of the component's ability to cope with humidity changes based on the component's electrical and physical characteristics. The worse the component's ability to cope with humidity changes, the more sensitive the component is to humidity, and the higher the corresponding humidity sensitivity score. When evaluating the humidity sensitivity score, the electrical characteristic data at this time are the electrical characteristic parameters related to the humidity sensitivity of the component, and the physical characteristic data are the physical characteristic parameters such as the constituent materials and the structure of the component itself related to the humidity sensitivity of the component, such as whether it contains a structure such as an isolation layer. Integrate a number of electrical characteristic data, physical characteristic data and a number of initial parameter sensitivity scores into a number of training data and test data. The artificial intelligence model is trained using training data, and the trained artificial intelligence model is tested using test data; specifically, the electrical characteristic data and physical characteristic data in the test data are input into the trained artificial intelligence model to obtain corresponding outputs, and it is determined whether the difference between the output initial parameter sensitivity score and the initial parameter sensitivity score recorded in the corresponding test data is within an acceptable range. If so, it means that the test data has passed the test and the next set of test data is tested; if not, the relevant parameters of the artificial intelligence are adjusted, and the test is continued using the current set of test data until a set proportion of the test data passes the test, and finally a parameter sensitivity scoring model is obtained with the input of electrical characteristic data and physical characteristic data and the output of the initial parameter sensitivity score; wherein the artificial intelligence model includes a BP neural network model and an RBF neural network model.

[0033] It can be understood that in this embodiment, initial parameter sensitivity scores corresponding to multiple monitoring items can be generated by training one model; it can also be achieved by training multiple models, with one model being used to generate an initial sensitivity score corresponding to one monitoring item; it can be understood that since the initial sensitivity scores of each component to each monitoring item are generated by the electrical and physical properties of the component itself, the initial sensitivity scores of the same component to the same monitoring item are the same, so this work can be performed before the implementation of the online monitoring method for the rectifier cabinet to reduce the amount of data processing during the implementation of the online monitoring method for the rectifier cabinet, thereby ensuring the efficiency of the implementation of the online monitoring method for the rectifier cabinet.

[0034] According to the rectifier cabinet impact score and the parameter sensitivity score, several sensitivity scores corresponding to each component are generated, including: obtaining the rectifier cabinet impact score of each rectifier cabinet, and several parameter sensitivity scores corresponding to each component in the rectifier cabinet; obtaining the product of the rectifier cabinet impact score and the parameter sensitivity score corresponding to any monitoring item of any component inside it, and using the product as the sensitivity score of the component under the monitoring item, and obtaining the sensitivity score corresponding to each component under each monitoring item in turn.

[0035] Since the operating environment of the rectifier cabinet is consistent with the operating environment of its internal components, that is, the impact of the operating environment on the rectifier cabinet as a whole, that is, the impact of the operating environment on the various components and lines inside the rectifier cabinet, these impacts will also cause the components' ability to cope with various monitoring items to change. Therefore, this embodiment obtains the sensitivity score of the component under each monitoring item by comprehensively considering the impact score of the rectifier cabinet and the parameter sensitivity score of the component itself; when the sensitivity score corresponding to a certain monitoring item is larger, the component is more sensitive to the monitoring item; that is, the component has less resistance to changes in the monitoring item.

[0036] Generate a priority processing coefficient based on the environmental monitoring data and the sensitivity scores of each component, including: extracting the parameter values ​​corresponding to each monitoring item in the environmental monitoring data; calculating the difference between the parameter values ​​of each monitoring item and the parameter values ​​in the previous monitoring data, and recording it as the change amount of the corresponding monitoring item, and marking it as ; It is understandable that the types and number of monitoring items in the environmental monitoring data are the same as the types and number of the initial parameter sensitivity scores, or match each other. For example, if the monitoring item is temperature, it has a corresponding temperature sensitivity score; Get the sensitivity score of any component under each monitoring item and mark it as ; By formula The priority processing coefficient YC of the element is calculated; wherein, It is the previous parameter value corresponding to the monitoring item numbered n; n is a positive integer, and n∈[1,N], N is the total number of monitoring items.

[0037] This embodiment calculates the priority processing coefficient of each component through the above formula. When the sensitivity score of a monitoring item corresponding to the component is higher and the change of the monitoring item is larger, the probability of abnormality of the component is greater; or when the sensitivity scores of multiple monitoring items corresponding to the component are higher and the change of these multiple monitoring items are large, the probability of abnormality of the component is greater; at this time, this embodiment sets the priority processing coefficient corresponding to the component to be larger; subsequently, the corresponding components are given priority to ensure that the abnormality of the component can be detected in time; thereby avoiding the situation where the abnormality of the component cannot be detected in time due to the delay in processing the detection data, and the impact on the operation of the rectifier cabinet is expanded.

[0038] It is understandable that, since various data of the same component are related to each other, this embodiment processes various detection data corresponding to the component in a unified manner when analyzing the component.

[0039] It is worth noting that in this embodiment, the operation of detecting an abnormal component each time is set according to the changes in the various environmental monitoring items at that time; specifically, the change value of each monitoring item is obtained in real time. It can be understood that the change value here is the difference between the value of the currently monitored monitoring item and the value of the corresponding monitoring item when the component was detected abnormal last time; when the change value corresponding to each monitoring item is greater than the corresponding set item threshold; yes, then the priority processing coefficient of each component is generated according to the values ​​of a series of monitoring items obtained this time, and subsequent monitoring is performed to obtain the status of each component; the set item difference threshold is set according to the distribution of the sensitivity scores of each component in the rectifier cabinet group to the monitoring item. Specifically, cluster analysis is performed on each sensitivity score of the same monitoring item to obtain a number of cluster clusters, and the cluster center corresponding to the cluster cluster with the largest density in each cluster cluster is obtained, and the sensitivity score corresponding to the cluster center is used as the sensitivity feature score of the monitoring item, and marked as ; Through the formula Calculate the set item difference threshold corresponding to the monitoring item numbered n ;in, is the maximum value of the initial parameter sensitivity scores of the components corresponding to the monitoring item; is a difference threshold corresponding to the initial parameter sensitivity of the component corresponding to the maximum value; is the parameter adjustment coefficient, and its specific value is set according to experience. In the above formula, when the current sensitivity score of each component to certain monitoring items is higher than the original corresponding initial sensitivity score, the present embodiment sets the setting item difference threshold of the corresponding monitoring item to be smaller, so as to ensure the timeliness of component detection.

[0040] Detecting each component according to the priority processing coefficient to obtain the corresponding component state, including: obtaining the working parameters of each component in order from large to small according to the priority processing coefficient, and judging whether each working parameter is within its corresponding set lower limit threshold and set upper limit threshold; if yes, marking the working parameter state corresponding to the working parameter as a normal parameter state; if no, marking the working parameter state corresponding to the working parameter as an abnormal parameter state; Obtain the working parameter status corresponding to each working parameter corresponding to the component in sequence; determine whether each working parameter status has a parameter abnormal state; if yes, mark the component status corresponding to the component as an abnormal state; if no, mark the component status corresponding to the component as a normal state.

[0041] Displaying the component status includes: arranging the component status of a rectifier cabinet in real time into the component status interface of the corresponding rectifier cabinet, judging whether there is an abnormal state in the current component status, if yes, setting the rectifier cabinet status to the abnormal state; if not, setting the rectifier cabinet status to the normal state; arranging the status of each rectifier cabinet in the rectifier cabinet group into the corresponding rectifier cabinet status sub-interface, and arranging each rectifier cabinet interface into a rectifier cabinet monitoring interface; there are several rectifier cabinet status sub-interfaces in the rectifier cabinet monitoring interface, and the rectifier cabinet status sub-interface displays the rectifier cabinet status; arranging the component status of a rectifier cabinet in real time into the component status interface of the corresponding rectifier cabinet, and the rectifier cabinet status sub-interface can be switched to the corresponding component status sub-interface, that is, clicking on the rectifier cabinet status sub-interface will enter the component status interface, and the component status interface includes several components and their corresponding component status, and clicking on the component status will enter the detection parameter interface corresponding to the component; the detection parameter interface displays the values ​​of several detection parameters of the corresponding component this time.

[0042] It can be understood that the component status and rectifier cabinet status are updated. When a component status is obtained during the component status detection, the corresponding rectifier cabinet status will be updated. At the same time, the component status and rectifier cabinet status in each display interface sub-interface will also be updated. It can be understood that when the component status is not detected, the component status corresponding to the component is displayed as empty or the component status of the last detection. It can be understood that the corresponding detection time is marked next to the component status.

[0043] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0044] How this application works: The present application generates a sensitivity score of each component to each monitoring item based on the historical usage data of the rectifier cabinet and the component characteristic data of each component inside the rectifier cabinet; then generates a priority processing coefficient of the corresponding component based on the current environmental monitoring data and several sensitivity scores of each component, analyzes the monitoring data of each component according to the priority processing coefficient to obtain the component status of each component, and finally displays the status of each component to realize online monitoring of the rectifier cabinet; the monitoring results are made more detailed by monitoring the component level, and at the same time, the priority processing coefficient of each component is set to ensure that the detection data of the corresponding component can be processed in time, thereby improving the timeliness of data processing and thus improving the overall detection efficiency of the rectifier cabinet.

[0045] The above embodiments are only used to illustrate the technical method of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical method of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for online monitoring of rectifier cabinets based on machine learning, characterized in that: The following steps are involved: Step 1: Obtain historical usage data of each rectifier cabinet in the rectifier cabinet group and component characteristic data of each component; Step 2: Generate a rectifier cabinet impact score corresponding to each rectifier cabinet based on historical usage data; generate several parameter sensitivity scores corresponding to each component based on component characteristic data; generate several sensitivity scores corresponding to each component based on the rectifier cabinet impact score and the parameter sensitivity score; Step 3: Obtain some environmental monitoring data; generate priority processing coefficients based on the environmental monitoring data and the sensitivity scores of each component; Step 4: Detect each component according to the priority processing coefficient to obtain the corresponding component status; Step 5: Get the component status and display it.

2. According to the online monitoring method of the rectifier cabinet based on machine learning according to claim 1, it is characterized in that: The generating of the rectifier cabinet impact score corresponding to each rectifier cabinet according to the historical usage data includes: Extract the usage environment data and usage duration corresponding to each recorded time in the historical usage data; extract each environmental parameter in the usage environment data and mark it as ; Mark the usage time as ; Where i is the number of the environmental parameter and t is the number of the recording time; By formula Calculate and obtain the rectifier cabinet impact score ZMP of the corresponding rectifier cabinet; in, is the optimal parameter value corresponding to the environmental parameter numbered i; is the unit change value corresponding to the environmental parameter numbered i; DT is the unit time; is the proportional coefficient corresponding to the recording time numbered t; is the adjustment coefficient; is the proportional coefficient of the environmental parameter numbered i; i is a positive integer, and i∈[1,I], I is the total number of environmental parameters; t is a positive integer, and t∈[1,T], T is the total number of total recording time of historical usage data.

3. The online monitoring method of a rectifier cabinet based on machine learning according to claim 1 is characterized in that: The generating of a plurality of parameter sensitivity scores corresponding to each component according to the component characteristic data includes: Extract the electrical characteristic data and physical characteristic data from the component characteristic data, input the electrical characteristic data and physical characteristic data into the parameter sensitivity scoring model to obtain the initial parameter sensitivity score under the corresponding monitoring item; obtain the initial parameter sensitivity score corresponding to each monitoring item, number it, and mark it as ; Obtain component usage data corresponding to each component; extract the recording time in the component usage data, as well as several corresponding working parameters and usage duration; generate the influence coefficient of the working parameter corresponding to each recording time according to the recording time, working parameters and usage duration; By formula Calculate the parameter sensitivity score corresponding to the monitoring item numbered n ;in, For the number The proportional coefficient corresponding to the recording time is three; is a positive integer, and ; is the total number of operating parameters of the corresponding component; is a positive integer, and , The total number of times the corresponding component usage data is recorded.

4. The online monitoring method of a rectifier cabinet based on machine learning according to claim 3 is characterized in that: The parameter sensitivity scoring model is obtained through artificial intelligence model training, including: Acquire electrical characteristic data and physical characteristic data of a plurality of components, and their corresponding initial parameter sensitivity scores, and integrate the electrical characteristic data, physical characteristic data, and the initial parameter sensitivity scores into a plurality of sets of training data and test data; The artificial intelligence model is trained using the training data, and the trained artificial intelligence model is tested using the test data; a parameter sensitivity scoring model is obtained whose input is electrical characteristic data and physical characteristic data, and whose output is an initial parameter sensitivity score.

5. The online monitoring method of a rectifier cabinet based on machine learning according to claim 3 is characterized in that: The generating of the influence coefficient of the working parameter corresponding to each recording time according to the recording time, the working parameter and the usage duration includes: Mark the working parameters as ; Use duration is marked as ;in, is the number of the working parameter corresponding to the component numbered j; It is the number of each recorded time in the component usage data; By formula The calculated number is Number of records within the time period The influence coefficient corresponding to the working parameters ;in, For the component numbered j, The set upper limit threshold value corresponding to the working parameter; For the component numbered j, The set lower limit threshold value corresponding to the working parameter; For the component numbered j, The set unit change value corresponding to the working parameter; GY is the set inherent influence coefficient; , and is the parameter adjustment coefficient.

6. The online monitoring method of a rectifier cabinet based on machine learning according to claim 1 is characterized in that: The method of generating a plurality of sensitivity scores corresponding to each component according to the rectifier cabinet impact score and the parameter sensitivity score includes: Obtain the rectifier cabinet impact score of each rectifier cabinet, and several parameter sensitivity scores corresponding to each component in the rectifier cabinet; obtain the product of the rectifier cabinet impact score and the parameter sensitivity score corresponding to any monitoring item of any component inside it, and use the product as the sensitivity score of the component under the monitoring item, and obtain the sensitivity score corresponding to each component under each monitoring item in turn.

7. The online monitoring method of a rectifier cabinet based on machine learning according to claim 1 is characterized in that: The priority processing coefficient is generated according to the environmental monitoring data and the sensitivity score of each component, including: Extract the parameter values ​​corresponding to each monitoring item in the environmental monitoring data; calculate the difference between the parameter values ​​of each monitoring item and the parameter values ​​in the previous monitoring data, and record it as the change amount of the corresponding monitoring item, and mark it as ; Get the sensitivity score of any component under each monitoring item and mark it as ; By formula The priority processing coefficient YC of the element is calculated; wherein, It is the previous parameter value corresponding to the monitoring item numbered n; n is a positive integer, and n∈[1,N], N is the total number of monitoring items.

8. The online monitoring method of a rectifier cabinet based on machine learning according to claim 1 is characterized in that: The detecting each component according to the priority processing coefficient to obtain the corresponding component state includes: The working parameters of each component are obtained in order from large to small according to the priority processing coefficient, and it is judged whether each working parameter is within the corresponding set lower limit threshold and set upper limit threshold; if yes, the working parameter state corresponding to the working parameter is marked as a normal parameter state; if no, the working parameter state corresponding to the working parameter is marked as an abnormal parameter state; Obtain the working parameter status corresponding to each working parameter corresponding to the component in sequence; determine whether each working parameter status has a parameter abnormal state; if yes, mark the component status corresponding to the component as an abnormal state; if no, mark the component status corresponding to the component as a normal state.

9. The online monitoring method of a rectifier cabinet based on machine learning according to claim 1 is characterized in that: The displaying of the component status includes: Determine whether there is an abnormal state in the current component state. If yes, set the rectifier cabinet state to an abnormal state; if no, set the rectifier cabinet state to a normal state; organize the states of each rectifier cabinet in the rectifier cabinet group into the corresponding rectifier cabinet state sub-interface, and organize each rectifier cabinet interface into a rectifier cabinet monitoring interface; there are several rectifier cabinet state sub-interfaces in the rectifier cabinet monitoring interface, and the rectifier cabinet state sub-interface displays the rectifier cabinet state; The component states belonging to a rectifier cabinet are sorted into the component state interface of the corresponding rectifier cabinet in real time, and the rectifier cabinet state sub-interface can be switched to the corresponding component state sub-interface; the component state interface includes a number of components and their corresponding component states.

Images