Transformer status assessment method and system based on multi-parameter data

Abstract

The present invention discloses a transformer state assessment method and system based on multi-parameter data, and specifically relates to the technical field of transformer state assessment; through multi-parameter data collection and analysis, the operating status of the transformer is comprehensively assessed and abnormal monitoring is performed; first, transformer parameter data is acquired in real time, and a key parameter data set is established; the response capability of the key parameters and the degree of abnormality are assessed according to different load states within the observation time; by analyzing the load state's response capability and the degree of abnormality to the key parameters, it is determined whether the transformer state is abnormal; a benchmark data set is established and a fault identification model is constructed to perform fault judgment on the abnormal state and distinguish between normal and potential fault states; finally, the judgment results are visualized, including the proportion of accurately identifying potential faults and the accuracy division, thereby improving the system's ability to identify key parameters of the transformer and the accuracy of judging potential faults, and increasing the stability of transformer operation.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer state assessment, and in particular to a transformer state assessment method and system based on multi-parameter data. Background Art

[0002] Transformer status assessment with multi-parameter data refers to the process of comprehensively evaluating and analyzing the operating status of the transformer using multiple parameter data. These parameter data can include information on voltage, current, temperature, humidity, vibration, etc. By monitoring and analyzing these data, possible problems or abnormal conditions of the transformer can be discovered in a timely manner, and corresponding measures can be taken for maintenance or repair to ensure the normal operation and safety of the transformer.

[0003] First, the transformer status assessment with multi-parameter data collects and analyzes various parameter data during the operation of the transformer, such as current, voltage, temperature, etc., to fully understand the operating status of the transformer. Secondly, based on the monitoring and analysis of these parameter data, possible problems or abnormal conditions of the transformer, such as overload, overheating, insulation aging, etc., can be identified, and corresponding measures can be taken in time for maintenance or repair to ensure the safe operation of the transformer and extend its service life.

[0004] The prior art has the following disadvantages:

[0005] In the prior art, various parameter data during the operation of the transformer, such as current, voltage, temperature, etc., are collected and analyzed to fully understand the operating status of the transformer. However, if the response speed of certain key parameters is slow, it may lead to inaccurate judgment of potential faults. For example, current abnormalities may reflect fault conditions faster than temperature changes. If the monitored current data lags behind the temperature data, it may lead to missed timely diagnosis of the fault, thereby delaying repair and maintenance. At the same time, as the accuracy of potential fault judgment continues to decrease, it means that the time delay for monitoring potential problems is longer, which may increase the risk of safety accidents during transformer operation. For example, if the temperature rises rapidly but the temperature sensor responds slowly, the discovery of overheating may be delayed, and the accuracy of potential fault judgment will also decrease, which may increase the risk of damage to the transformer or even fire. Summary of the invention

[0006] The purpose of the present invention is to provide a transformer state assessment method and system based on multi-parameter data to address the deficiencies in the background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a transformer state assessment method based on multi-parameter data, comprising the following steps;

[0008] S1: Real-time acquisition of various parameter data during transformer operation, screening and cleaning of the acquired parameter data to extract key parameter information, and establish a key parameter data set;

[0009] S2: Set the observation time, and judge the influence of the load state of the transformer in different modes during the observation time on the response capability of the key parameters. According to the actual response time changes of the key parameters during the observation period, evaluate the abnormality of the response speed of the key parameters.

[0010] S3: Comprehensively analyze the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluate whether the response state of the transformer key parameters is abnormal, and distinguish between abnormal response state and normal response state according to the evaluation results;

[0011] S4: For transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set;

[0012] S5: Perform fault judgment on the transformer in abnormal response state according to the matching result to distinguish between normal state and potential fault state;

[0013] S6: Based on the results of fault judgment, calculate the proportion of potential faults accurately identified by the model, divide the accuracy of fault judgment according to the proportion of potential faults, and divide it into high accuracy judgment, medium accuracy judgment and low accuracy judgment, and visualize the division results.

[0014] In a preferred embodiment, in S1, various parameter data during the operation of the transformer are acquired in real time; the acquired parameter data are screened, and key parameters related to the transformer status assessment are selected, including current, voltage, temperature, humidity, and pressure; the screened parameter data are cleaned to remove abnormal values ​​and erroneous data; the cleaning process includes identifying and processing missing values, abnormal values, and duplicate data; key parameter information is extracted from the cleaned data; and the extracted key parameter information is merged to form a complete data set.

[0015] In a preferred embodiment, in S2, the load data of the transformer in different modes are collected respectively, and the load data abnormality index of the transformer is obtained from the collected load data of the transformer. The method for obtaining the load data abnormality index is:

[0016] Collect the load data of the transformer in different modes respectively, and calculate the moving average of the load data. The specific calculation expression is: In the formula, MA i is the moving average of the i-th time point, LD jis the transformer load data at the jth time point, n is the length of the moving average window; determine the fluctuation degree of the transformer load data, calculate the standard deviation of the load data, and the specific calculation expression is: In the formula, dg r is the standard deviation of the load data, LD k is the mean value of transformer load data, N is the total number of transformer load data;

[0017] The load data anomaly index is calculated by calculating the ratio of the deviation between the transformer load data and the moving average to the standard deviation. The specific calculation expression is: In the formula, wg f is the load data abnormality index, dg r is the standard deviation of the load data, LD j is the transformer load data at the jth time point, MA i is the moving average at the i-th time point.

[0018] In a preferred embodiment, in S2, actual response time variation data of key parameters are respectively obtained, and the response speed drift index of the key parameters is obtained according to the actual response time variation data of the key parameters. The method for obtaining the response speed drift index is:

[0019] The acquired actual response time variation data of key parameters are expressed as a matrix, where each row represents an observation sample and each column represents the actual response time data of a key parameter;

[0020] Calculate the similarity between each pair of samples in the data, use the local weighted matrix W to linearly reconstruct each sample, calculate the linear relationship between each sample and its neighbors by minimizing the reconstruction error, and obtain a local linear reconstruction coefficient matrix C;

[0021] Perform eigenvalue decomposition on the local linear reconstruction coefficient matrix C, extract the first gd eigenvectors as new low-dimensional embedding coordinates, and calculate the average value of the embedding coordinates of all samples. The specific calculation expression is: gd e =∑gd*C; where gd e is the average value of the embedding coordinates of all samples;

[0022] Perform statistical analysis on the embedded coordinates and calculate the response speed drift index. The specific calculation expression is: Where m is the number of samples, that is, the number of data points during the observation period, gd q is the standard value of the embedding coordinates of all samples under the preset standard state, rk e is the response speed drift index.

[0023] In a preferred embodiment, in S3, the influence of the load state of the transformer on the response capability of the key parameters and the abnormality of the response speed of the key parameters are comprehensively analyzed, specifically:

[0024] The load data anomaly index and the response speed drift index are normalized, and the anomaly assessment coefficient of the response state of the key parameters of the transformer is calculated by the normalized load data anomaly index and the response speed drift index.

[0025] In a preferred embodiment, in S3, whether the response state of the transformer key parameters is abnormal is evaluated, and the abnormal response state and the normal response state are distinguished according to the evaluation result, specifically:

[0026] The abnormal assessment coefficient of the transformer key parameter response state is compared with the abnormal threshold. If the abnormal assessment coefficient of the transformer key parameter response state is greater than or equal to the abnormal threshold, the transformer key parameter response state is judged to be an abnormal response state, and an early warning signal is issued; if the abnormal assessment coefficient of the transformer key parameter response state is less than the abnormal threshold, the transformer key parameter response state is judged to be a normal response state, and no early warning signal is issued.

[0027] In a preferred embodiment, in S4, when a decision tree is used as a fault identification model, historical transformer operation data is collected, and a benchmark data set is constructed according to data characteristics, and a decision tree model is trained using the benchmark data set. When real-time transformer parameter data is collected, it is input into the trained decision tree model for matching; specifically:

[0028] Based on the splitting conditions of each internal node, the decision tree model will select a path until it reaches a leaf node;

[0029] The leaf node will give a classification label, indicating whether the data belongs to normal or abnormal state;

[0030] Each step is judged based on the value of the parameter data and the conditions of the decision tree node until the final classification result is obtained;

[0031] Using the trained decision tree model, the transformer parameter data collected in real time is input into the model for classification.

[0032] In a preferred embodiment, in S5, a fault judgment is performed on the transformer in the abnormal response state according to the matching result to distinguish between a normal state and a potential fault state;

[0033] The decision tree makes decisions step by step according to the characteristics of the input data; according to the classification results of the decision tree, the transformer fault is judged. If the classification result indicates that the data is in an abnormal state, the transformer has a potential fault and a transformer fault signal is issued; if the classification result indicates that the data is in a normal state, the transformer is in good operating condition and there is no potential fault in the transformer. At this time, a transformer normal signal is issued and no processing is required.

[0034] In a preferred embodiment, in S6, based on the result of the fault judgment, the proportion of potential faults accurately identified by the calculation model is specifically:

[0035] Obtain the total number of samples H and label them, where H = [1, 2, 3...n]; n is a positive integer greater than 0; obtain the number of potential fault samples M that the model successfully identifies; calculate the proportion of potential faults that the model accurately identifies, and the specific calculation expression is: Among them, P is the proportion of potential faults accurately identified by the model, M is the number of potential fault samples successfully identified by the model, and H is the total number of samples.

[0036] In a preferred embodiment, in S6, fuzzy logic is used to classify the accuracy of fault judgment according to the potential fault ratio, and it is divided into high accuracy judgment, medium accuracy judgment and low accuracy judgment, specifically:

[0037] The proportion P of potential faults accurately identified by the model is converted into a fuzzy set; P is divided into three fuzzy sets: low L, medium Q and high D;

[0038] The Logistic function is used to define the membership function, and its function form is: Low accuracy judgment L: Moderate accuracy judgment Q: High accuracy judgment D: Among them, a i and b i is the parameter of the fuzzy membership function, u L (p),u Q (p),u D (p) represents the fuzzy membership of low, medium, and high accuracy judgments, respectively;

[0039] The fuzzy output is converted into a specific accuracy level, and the accuracy level with the maximum membership is selected as the final accuracy judgment; the specific calculation expression is: Accuracy judgment = argmax(u L (p),u Q (p),u D (p)); where u L (p),u Q (p),u D(p) represents the fuzzy membership of low, medium and high accuracy judgments respectively. The function argmax returns the parameter with the maximum value, and the accuracy level with the maximum membership is selected as the final judgment result.

[0040] In a preferred embodiment, the acquired proportion of potential faults is compared with a gradient threshold, the gradient threshold includes a first threshold and a second threshold, and the first threshold is less than the second threshold, and the proportion of potential faults is compared with the first threshold and the second threshold respectively; if the proportion of potential faults is greater than the second threshold, the output is a high accuracy judgment; if the proportion of potential faults is greater than or equal to the first threshold and less than or equal to the second threshold, the output is a medium accuracy judgment; if the proportion of potential faults is less than the first threshold, the output is an accuracy judgment.

[0041] The present invention also provides a transformer state assessment system based on multi-parameter data, including a data acquisition module, an abnormality monitoring module, an assessment module, a matching module, a state division module and a visualization module;

[0042] Data acquisition module: acquires various parameter data during transformer operation in real time, screens and cleans the acquired parameter data to extract key parameter information, and establishes a key parameter data set;

[0043] Abnormal monitoring module: Set the observation time, and judge the influence of the load state of the transformer in different modes during the observation time on the response ability of key parameters. According to the actual response time changes of key parameters during the observation period, evaluate the abnormality of the response speed of key parameters.

[0044] Evaluation module: comprehensively analyzes the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluates whether the transformer key parameter response state is abnormal, and distinguishes abnormal response state from normal response state based on the evaluation results;

[0045] Matching module: For transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set;

[0046] State classification module: fault judgment is performed on the transformer in abnormal response state according to the matching results to distinguish between normal state and potential fault state;

[0047] Visualization module: Based on the results of fault judgment, the calculation model accurately identifies the proportion of potential faults, and divides the accuracy of fault judgment according to the proportion of potential faults, dividing it into high accuracy judgment, medium accuracy judgment and low accuracy judgment, and visualizes the division results.

[0048] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0049] 1. The present invention can more accurately identify potential fault conditions by acquiring key parameter data during transformer operation in real time and performing comprehensive analysis and abnormal monitoring on these data. Establishing a benchmark data set and building a fault identification model can help compare the actual operation data of the transformer with historical data and identify abnormal response states. It helps to promptly discover abnormal conditions in transformer operation, thereby reducing the situation of missed faults and false alarms and improving the accuracy of fault diagnosis.

[0050] 2. The present invention can help prevent serious safety accidents by timely identifying and accurately judging the potential fault status of the transformer. By judging the fault of the transformer in the abnormal response state, necessary repair and maintenance measures can be taken quickly to avoid the fault from further deteriorating or evolving into a serious problem. This timely fault diagnosis and handling method effectively reduces the risk of safety accidents during transformer operation and ensures the safety and stability of equipment and operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] 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 will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0052] Figure 1 The present invention is a flow chart of the method.

[0053] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0055] Example 1

[0056] See also Figure 1 As shown, the transformer state assessment method based on multi-parameter data described in this embodiment includes the following steps:

[0057] S1: Real-time acquisition of various parameter data during transformer operation, screening and cleaning of the acquired parameter data to extract key parameter information, and establish a key parameter data set;

[0058] S2: Set the observation time, and judge the influence of the load state of the transformer in different modes during the observation time on the response capability of the key parameters. According to the actual response time changes of the key parameters during the observation period, evaluate the abnormality of the response speed of the key parameters.

[0059] S3: Comprehensively analyze the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluate whether the response state of the transformer key parameters is abnormal, and distinguish between abnormal response state and normal response state according to the evaluation results;

[0060] S4: For transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set;

[0061] S5: Perform fault judgment on the transformer in abnormal response state according to the matching result to distinguish between normal state and potential fault state;

[0062] S6: Based on the results of fault judgment, calculate the proportion of potential faults accurately identified by the model, divide the accuracy of fault judgment according to the proportion of potential faults, and divide it into high accuracy judgment, medium accuracy judgment and low accuracy judgment, and visualize the division results.

[0063] Among them, in S1, various parameter data during the operation of the transformer are obtained in real time, the obtained parameter data are screened and cleaned to extract key parameter information, and a key parameter data set is established, specifically:

[0064] According to the requirements, data related to the transformer operating status are selected from the parameter data obtained in real time, and irrelevant or redundant data are excluded.

[0065] To obtain various parameter data of transformer operation in real time and extract key parameter information, various means and technologies are usually required. The following are some common means of extracting key data:

[0066] Sensor technology: Various types of sensors are installed on the transformer to monitor various parameter data in real time, such as current, voltage, temperature, humidity, pressure, etc. The sensor can be a temperature sensor, current sensor, voltage sensor, humidity sensor, etc. Various parameter data can be directly obtained through sensor technology.

[0067] Monitoring equipment: Use special monitoring equipment or monitoring systems to monitor and record various parameter data during transformer operation. These monitoring devices usually have functions such as data acquisition, storage, analysis and alarm, and can obtain and process various parameter data in real time.

[0068] Smart meters: Smart meters are installed on the input and output sides of the transformer to monitor electrical parameters such as current, voltage, power factor, etc. in real time, and transmit the data to the monitoring system for processing and analysis.

[0069] The selected parameter data is cleaned to remove outliers, erroneous data or noise data to ensure the quality and accuracy of the data.

[0070] The specific steps to extract key data include:

[0071] Real-time acquisition of various parameter data during transformer operation, including current, voltage, temperature, humidity, etc. Data collection can be achieved through sensors, monitoring equipment, smart meters, etc.

[0072] The acquired parameter data is screened to select key parameters related to transformer status assessment, which may include current, voltage, temperature, humidity, pressure, etc.

[0073] Clean the filtered parameter data to remove outliers, erroneous data or noise data. The cleaning process includes identifying and processing missing values, outliers and duplicate data.

[0074] Extract key parameter information from the cleaned data. This may involve calculating parameter statistics (such as mean, standard deviation, peak value, etc.), time domain features (such as waveform shape, periodicity, etc.), frequency domain features (such as spectrum analysis, frequency components, etc.), etc.

[0075] The extracted key parameter information is merged to form a complete data set or data record. This can be a time series data set, where each row represents a time point and each column represents a key parameter.

[0076] The merged data is stored in a database or file for subsequent analysis and processing. The data storage can be in the form of a structured database table, a time series database, a text file, etc.

[0077] Analyze the stored data to explore the relationships and patterns between the data. This may involve statistical analysis, machine learning, model building and other methods.

[0078] The analysis results are presented in a visual way.

[0079] S2: Set the observation time, and judge the influence of the load status of the transformer in different modes during the observation time on the response capability of the key parameters. According to the actual response time changes of the key parameters during the observation period, evaluate the abnormality of the response speed of the key parameters.

[0080] Select a suitable time period to observe the operating status of the transformer. This time period can be several hours, a day, a week or even longer, depending on the operating mode and periodic changes of the transformer.

[0081] During the observation time, the load data of the transformer in different modes are collected respectively, and the load data abnormality index of the transformer is obtained from the collected load data of the transformer. The method for obtaining the load data abnormality index is:

[0082] Collect the load data of the transformer in different modes respectively, and calculate the moving average of the load data. The moving average is to calculate the average value of the transformer load data in a window of fixed length. This window will move over time, one time step at a time. The specific calculation expression is: In the formula, MA i is the moving average of the i-th time point, LD j is the transformer load data at the jth time point, and n is the length of the moving average window.

[0083] Calculate the standard deviation of the load data. The standard deviation of the load data is used to measure the fluctuation degree of the transformer load data. The larger the standard deviation, the greater the fluctuation of the data. The specific calculation expression is: In the formula, dg r is the standard deviation of the load data, LD k is the mean of transformer load data, and N is the total number of transformer load data.

[0084] Calculate the load data anomaly index. The load data anomaly index can be obtained by calculating the ratio of the deviation between the transformer load data and the moving average to the standard deviation. The specific calculation expression is: In the formula, wg f is the load data abnormality index, dg r is the standard deviation of the load data, LD j is the transformer load data at the jth time point, MA i is the moving average at the i-th time point.

[0085] During the observation period, the actual response time change data of the key parameters are obtained respectively, and the response speed drift index of the key parameters is obtained according to the actual response time change data of the key parameters. The method for obtaining the response speed drift index is:

[0086] The actual response time variation data of the key parameters obtained during the observation period are represented as a matrix, where each row represents an observation sample and each column represents the actual response time data of a key parameter.

[0087] Calculate the similarity between each pair of samples in the data. Commonly used similarity metrics include Euclidean distance, Manhattan distance, or Gaussian kernel function. Get a similarity matrix, where the element S ij It represents the similarity between sample i and sample j.

[0088] The similarity matrix is ​​locally weighted, usually using the weighted nearest neighbor method, selecting the k nearest neighbors for each sample and weighting these neighbors to obtain the local weighted matrix W.

[0089] The local weighted matrix W is used to linearly reconstruct each sample, that is, by minimizing the reconstruction error, the linear relationship between each sample and its neighbors is calculated to obtain a local linear reconstruction coefficient matrix C.

[0090] Perform eigenvalue decomposition on the local linear reconstruction coefficient matrix C and extract the first gd eigenvectors as new low-dimensional embedding coordinates, where gd is the dimension of the target low-dimensional space. Calculate the average value of the embedding coordinates of all samples. The specific calculation expression is: gd e =∑gd*C; where gd e is the average of the embedding coordinates of all samples.

[0091] Perform statistical analysis on the embedded coordinates and calculate the response speed drift index. The specific calculation expression is: Where m is the number of samples, that is, the number of data points during the observation period, gd q is the standard value of the embedding coordinates of all samples under the preset standard state, rk e is the response speed drift index.

[0092] It should be noted that the standard values ​​of the embedding coordinates of all samples are set by those skilled in the art according to the actual response time changes of specific key parameters, which will not be elaborated here.

[0093] S3: Comprehensively analyze the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluate whether the transformer key parameter response state is abnormal, and distinguish between abnormal response state and normal response state based on the evaluation results.

[0094] The load data anomaly index and the response speed drift index are normalized, and the anomaly assessment coefficient of the response state of the key parameters of the transformer is calculated by the normalized load data anomaly index and the response speed drift index.

[0095] For example, the present invention can use the following formula to calculate the abnormal evaluation coefficient of the transformer key parameter response state, and the calculation expression is: In the formula, tg v is the abnormal assessment coefficient, wg f is the load data abnormality index, rk e is the response speed drift index, a 1 、a 2 is the load data abnormality index, the proportionality coefficient of the response speed drift index, and a 2 >a 1 >0.

[0096] It can be seen from the calculation expression that the load data abnormality index and the response speed drift index are proportional to the abnormality assessment coefficient, and as the load data abnormality index and the response speed drift index increase, the abnormality assessment coefficient also increases, and the possibility of abnormal response state of key parameters of the transformer becomes greater.

[0097] The abnormal evaluation coefficient of the transformer key parameter response state is compared with the abnormal threshold. If the abnormal evaluation coefficient of the transformer key parameter response state is greater than or equal to the abnormal threshold, the transformer key parameter response state is judged to be an abnormal response state. At this time, an early warning signal is issued to notify relevant personnel to take further inspection, maintenance or adjustment measures to resolve the abnormal situation and ensure the safe operation of the transformer. If the abnormal evaluation coefficient of the transformer key parameter response state is less than the abnormal threshold, the transformer key parameter response state is judged to be a normal response state. At this time, no early warning signal is issued. Relevant personnel can continue to monitor the operation of the transformer in the normal response state without taking additional emergency measures.

[0098] In this embodiment, various parameter data during the operation of the transformer are obtained in real time, the obtained parameter data are screened and cleaned to extract key parameter information and establish a key parameter data set; the observation time is set, and the load state of the transformer in different modes during the observation time is judged to determine the degree of influence on the response capability of the key parameters, and the abnormal degree of the response speed of the key parameters is evaluated according to the actual response time changes of the key parameters during the observation period; the influence of the load state of the transformer on the response capability of the key parameters and the abnormal degree of the response speed of the key parameters are comprehensively analyzed to evaluate whether the response state of the key parameters of the transformer is abnormal, and the abnormal response state and the normal response state are distinguished according to the evaluation results. Not only can the operating status of the transformer be fully understood, but any possible abnormal conditions can also be discovered and handled in a timely manner to ensure the safe and stable operation of the transformer.

[0099] Example 2

[0100] S4: For transformers in abnormal response states, a benchmark data set is established based on the historical operation data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set, specifically:

[0101] When using a decision tree as a fault identification model, the matching steps of the data set include:

[0102] Collect the historical operation data of the transformer and build a benchmark data set based on the data characteristics. The benchmark data set should contain parameter values ​​under various normal operating conditions. These data can include parameters such as current, voltage, and temperature.

[0103] The decision tree model is trained using the benchmark dataset. The decision tree model will learn how to classify based on the input parameter data to distinguish between normal and abnormal states.

[0104] When real-time transformer parameter data is collected, it is input into the trained decision tree model for matching.

[0105] The decision tree model will make decisions step by step according to the characteristics of the parameter data and finally give the corresponding classification results, that is, normal or abnormal status.

[0106] First, the key parameter data collected in real time are used as input and compared according to the internal nodes of the decision tree.

[0107] Based on the splitting conditions of each internal node, the decision tree model chooses a specific path until it reaches a leaf node.

[0108] The leaf node will be given a classification label, indicating whether the data belongs to a normal state or an abnormal state.

[0109] The matching process of the decision tree is similar to a series of logical judgments. Each step is judged based on the value of the parameter data and the conditions of the decision tree nodes until the final classification result is obtained.

[0110] S5: Perform fault judgment on the transformer in abnormal response state according to the matching result to distinguish between normal state and potential fault state.

[0111] According to the matching results of the decision tree on the data set, the transformer in abnormal response state is judged to distinguish between normal state and potential fault state. This means using the decision tree model to classify the transformer parameter data collected in real time to determine whether there is an abnormality or potential fault in the current transformer operation state. Specifically:

[0112] Using the trained decision tree model, the transformer parameter data collected in real time is input into the model for classification.

[0113] The decision tree makes decisions step by step based on the characteristics of the input data, and finally gives the classification label to which the data belongs, that is, normal state or abnormal state.

[0114] According to the classification results of the decision tree, the transformer fault is judged.

[0115] If the classification result indicates that the data is in an abnormal state, there is a potential fault in the transformer. In this case, a transformer fault signal is issued and further inspection and maintenance are required.

[0116] If the classification result indicates that the data is in a normal state, the transformer is in good operating condition and there is no potential fault in the transformer. In this case, a normal transformer signal is issued and no special processing is required.

[0117] If the decision tree model classifies the transformer parameter data as abnormal, there may be potential faults or abnormal conditions. In the abnormal state, further measures can be taken, such as notifying maintenance personnel to conduct inspections, taking preventive maintenance measures, etc., to prevent the fault from further deteriorating or avoid unnecessary losses.

[0118] S6: Based on the results of fault judgment, calculate the proportion of potential faults accurately identified by the model, divide the accuracy of fault judgment according to the proportion of potential faults, and divide it into high accuracy judgment, medium accuracy judgment and low accuracy judgment, and visualize the division results.

[0119] Obtain the fault recognition status of the statistical model in a certain period of time, usually historical data or a validation data set. Obtain the total number of samples H and label them, where H = [1, 2, 3...n]; n is a positive integer greater than 0; obtain the number of potential fault samples M that the model successfully identified. Calculate the proportion of potential faults that the model accurately identifies. The specific calculation expression is: Among them, P is the proportion of potential faults accurately identified by the model, M is the number of potential fault samples successfully identified by the model, and H is the total number of samples.

[0120] Fuzzy logic is used to classify the accuracy of fault judgment according to the potential fault ratio, and it is divided into high accuracy judgment, medium accuracy judgment and low accuracy judgment, specifically:

[0121] In the fuzzification step, the proportion of potential faults that the model accurately identifies, P, is converted into a fuzzy set. The output can be fuzzified using an S-shaped membership function, where P represents the proportion of potential faults that the model accurately identifies, ranging from 0 to 1. P is divided into three fuzzy sets: low L, medium Q, and high D.

[0122] The Logistic function is used to define the membership function, and its function form is:

[0123] Low accuracy judgment L:

[0124] Moderate accuracy judgment Q:

[0125] High accuracy judgment D:

[0126] Among them, a i and b i are the parameters of the fuzzy membership function, which can be adjusted according to the specific situation to ensure that the fuzzification effect conforms to the actual situation. L (p),u Q (p),u D (p) represents the fuzzy membership of low, medium and high accuracy judgments, respectively.

[0127] In the defuzzification process, the fuzzy output needs to be converted into a specific accuracy level. A commonly used method is the maximum membership method, which selects the accuracy level with the maximum membership as the final accuracy judgment. The specific calculation expression is: Accuracy judgment = argmax(u L (p),u Q (p),u D (p)); where u L (p),u Q (p),u D (p) represents the fuzzy membership of low, medium and high accuracy judgments respectively. The function argmax returns the parameter with the maximum value, that is, the accuracy level with the maximum membership is selected as the final judgment result.

[0128] When defining fuzzy rules, the influence of the proportion P of potential faults accurately identified by the judgment model on the accuracy judgment is considered. A simple fuzzy rule can be: compare the proportion of potential faults obtained with the gradient threshold, the gradient threshold includes a first threshold and a second threshold, and the first threshold is less than the second threshold, and compare the proportion of potential faults with the first threshold and the second threshold respectively;

[0129] If the proportion of potential faults is greater than the second threshold, the output is a high accuracy judgment;

[0130] If the proportion of potential faults is greater than or equal to the first threshold and less than or equal to the second threshold, the output is a medium accuracy judgment;

[0131] If the proportion of potential faults is less than the first threshold, the output is an accuracy judgment.

[0132] For example, the fuzzy output is u L (p) = 0.2, u Q (p) = 0.5, u D(p) = 0.7, then the accuracy level with the maximum membership degree needs to be selected as the final accuracy judgment. D (p) has a maximum value of 0.7. If the first threshold is 0.4 and the second threshold is 0.8, a medium accuracy judgment can be obtained as the final result.

[0133] The fault judgment results can be visualized using various charts or graphic display methods, such as line charts, scatter plots, heat maps, etc., to intuitively display the operating status and potential fault conditions of the transformer. The fault status can be visualized in space in combination with geographic information display methods such as maps, so as to better understand and analyze the operation of the transformer. At the same time, the results of the visualization display are interpreted and analyzed to provide intuitive information and suggestions for decision makers, so that corresponding repair or maintenance measures can be taken in time to ensure the safe and stable operation of the transformer.

[0134] In this embodiment, for transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set; fault judgment is performed on the transformer in abnormal response state based on the matching result to distinguish between normal state and potential fault state; based on the result of the fault judgment, the proportion of potential faults accurately identified by the calculation model is calculated, and the accuracy of the fault judgment is divided according to the proportion of potential faults, which is divided into high accuracy judgment, medium accuracy judgment and low accuracy judgment. Fault judgment and accuracy evaluation can be performed on transformers in abnormal response states, and the results can be visualized to provide effective reference information for decision makers.

[0135] Example 3

[0136] See also Figure 2 As shown, the financial risk prediction system based on big data described in this embodiment includes a data acquisition module, an abnormality monitoring module, an evaluation module, a matching module, a state division module and a visualization module;

[0137] Data acquisition module: acquires various parameter data during transformer operation in real time, screens and cleans the acquired parameter data to extract key parameter information, and establishes a key parameter data set;

[0138] Abnormal monitoring module: Set the observation time, and judge the influence of the load state of the transformer in different modes during the observation time on the response ability of key parameters. According to the actual response time changes of key parameters during the observation period, evaluate the abnormality of the response speed of key parameters.

[0139] Evaluation module: comprehensively analyzes the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluates whether the transformer key parameter response state is abnormal, and distinguishes abnormal response state from normal response state based on the evaluation results;

[0140] Matching module: For transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set;

[0141] State classification module: fault judgment is performed on the transformer in abnormal response state according to the matching results to distinguish between normal state and potential fault state;

[0142] Visualization module: Based on the results of fault judgment, the calculation model accurately identifies the proportion of potential faults, and divides the accuracy of fault judgment according to the proportion of potential faults, dividing it into high-accuracy judgment, medium-accuracy judgment and low-accuracy judgment, and visualizes the division results.

[0143] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0144] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0145] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0146] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0147] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0148] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A transformer condition assessment method based on multi-parameter data, characterized in that: The steps include: S1: Real-time acquisition of various parameter data during transformer operation, screening and cleaning of the acquired parameter data to extract key parameter information, and establish a key parameter data set; S2: Set the observation time, and judge the influence of the load state of the transformer in different modes during the observation time on the response capability of the key parameters. According to the actual response time changes of the key parameters during the observation period, evaluate the abnormality of the response speed of the key parameters. S3: Comprehensively analyze the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluate whether the response state of the transformer key parameters is abnormal, and distinguish between abnormal response state and normal response state according to the evaluation results; S4: For transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set; S5: Perform fault judgment on the transformer in abnormal response state according to the matching result to distinguish between normal state and potential fault state; S6: Based on the results of fault judgment, calculate the proportion of potential faults accurately identified by the model, divide the accuracy of fault judgment according to the proportion of potential faults, and divide it into high accuracy judgment, medium accuracy judgment and low accuracy judgment, and visualize the division results.

2. The transformer state assessment method based on multi-parameter data according to claim 1 is characterized in that: In S1, various parameter data during the operation of the transformer are obtained in real time; the obtained parameter data are screened and key parameters related to the transformer status assessment are selected, including current, voltage, temperature, humidity, and pressure; the screened parameter data are cleaned to remove outliers and erroneous data; the cleaning process includes identifying and processing missing values, outliers, and duplicate data; key parameter information is extracted from the cleaned data; the extracted key parameter information is merged to form a complete data set.

3. The transformer state assessment method based on multi-parameter data according to claim 2 is characterized in that: In S2, the load data of the transformer in different modes are collected respectively, and the load data abnormality index of the transformer is obtained from the collected load data of the transformer. The method for obtaining the load data abnormality index is: Collect the load data of the transformer in different modes respectively, and calculate the moving average of the load data. The specific calculation expression is: In the formula, MA i is the moving average of the i-th time point, LD j is the transformer load data at the jth time point, n is the length of the moving average window; determine the fluctuation degree of the transformer load data, calculate the standard deviation of the load data, and the specific calculation expression is: In the formula, dg r is the standard deviation of the load data, LD k is the mean value of transformer load data, N is the total number of transformer load data; The load data anomaly index is calculated by calculating the ratio of the deviation between the transformer load data and the moving average to the standard deviation. The specific calculation expression is: In the formula, wg f is the load data abnormality index, dg r is the standard deviation of the load data, LD j is the transformer load data at the jth time point, MA i is the moving average at the i-th time point.

4. The transformer state assessment method based on multi-parameter data according to claim 3 is characterized in that: In S2, actual response time change data of key parameters are obtained respectively, and the response speed drift index of the key parameters is obtained according to the actual response time change data of the key parameters. The method for obtaining the response speed drift index is: The acquired actual response time variation data of key parameters are expressed as a matrix, where each row represents an observation sample and each column represents the actual response time data of a key parameter; Calculate the similarity between each pair of samples in the data, use the local weighted matrix W to linearly reconstruct each sample, calculate the linear relationship between each sample and its neighbors by minimizing the reconstruction error, and obtain a local linear reconstruction coefficient matrix C; Perform eigenvalue decomposition on the local linear reconstruction coefficient matrix C, extract the first gd eigenvectors as new low-dimensional embedding coordinates, and calculate the average value of the embedding coordinates of all samples. The specific calculation expression is: gd e =∑gd*C; where gd e is the average value of the embedding coordinates of all samples; Perform statistical analysis on the embedded coordinates and calculate the response speed drift index. The specific calculation expression is: Where m is the number of samples, that is, the number of data points during the observation period, gd q is the standard value of the embedding coordinates of all samples under the preset standard state, rk e is the response speed drift index.

5. The transformer state assessment method based on multi-parameter data according to claim 4 is characterized in that: In S3, the influence of the transformer load state on the response capability of the key parameters and the abnormality of the key parameter response speed are comprehensively analyzed, specifically: The load data anomaly index and the response speed drift index are normalized, and the anomaly assessment coefficient of the response state of the key parameters of the transformer is calculated by the normalized load data anomaly index and the response speed drift index.

6. The transformer state assessment method based on multi-parameter data according to claim 5 is characterized in that: In S3, whether the response state of the transformer key parameters is abnormal is evaluated, and the abnormal response state and the normal response state are distinguished according to the evaluation results, specifically: The abnormal assessment coefficient of the transformer key parameter response state is compared with the abnormal threshold. If the abnormal assessment coefficient of the transformer key parameter response state is greater than or equal to the abnormal threshold, the transformer key parameter response state is judged to be an abnormal response state, and an early warning signal is issued; if the abnormal assessment coefficient of the transformer key parameter response state is less than the abnormal threshold, the transformer key parameter response state is judged to be a normal response state, and no early warning signal is issued.

7. The transformer state assessment method based on multi-parameter data according to claim 6 is characterized in that: In S4, when the decision tree is used as the fault identification model, the historical operation data of the transformer is collected, and a benchmark data set is constructed according to the data characteristics. The decision tree model is trained using the benchmark data set. When the real-time transformer parameter data is collected, it is input into the trained decision tree model for matching; specifically: Based on the splitting conditions of each internal node, the decision tree model will select a path until it reaches a leaf node; The leaf node will give a classification label, indicating whether the data belongs to normal or abnormal state; Each step is judged based on the value of the parameter data and the conditions of the decision tree node until the final classification result is obtained; Using the trained decision tree model, the transformer parameter data collected in real time is input into the model for classification.

8. The transformer state assessment method based on multi-parameter data according to claim 7 is characterized in that: In S5, a fault judgment is performed on the transformer in the abnormal response state according to the matching result to distinguish between a normal state and a potential fault state; The decision tree makes decisions step by step according to the characteristics of the input data; according to the classification results of the decision tree, the transformer fault is judged. If the classification result indicates that the data is in an abnormal state, the transformer has a potential fault and a transformer fault signal is issued; if the classification result indicates that the data is in a normal state, the transformer is in good operating condition and there is no potential fault in the transformer. At this time, a transformer normal signal is issued and no processing is required.

9. The transformer state assessment method based on multi-parameter data according to claim 8, characterized in that: In S6, based on the result of the fault judgment, the proportion of potential faults accurately identified by the calculation model is: Obtain the total number of samples H and label them, where H = [1, 2, 3...n]; n is a positive integer greater than 0; obtain the number of potential fault samples M that the model successfully identifies; calculate the proportion of potential faults that the model accurately identifies, and the specific calculation expression is: Among them, P is the proportion of potential faults accurately identified by the model, M is the number of potential fault samples successfully identified by the model, and H is the total number of samples.

10. The transformer state assessment method based on multi-parameter data according to claim 9, characterized in that: In S6, fuzzy logic is used to classify the accuracy of fault judgment according to the potential fault ratio, and it is divided into high accuracy judgment, medium accuracy judgment and low accuracy judgment, specifically: The proportion P of potential faults accurately identified by the model is converted into a fuzzy set; P is divided into three fuzzy sets: low L, medium Q and high D; The Logistic function is used to define the membership function, and its function form is: Low accuracy judgment L: Moderate accuracy judgment Q: High accuracy judgment D: Among them, a i and b i is the parameter of the fuzzy membership function, u L (p),u Q (p),u D (p) represents the fuzzy membership of low, medium, and high accuracy judgments, respectively; The fuzzy output is converted into a specific accuracy level, and the accuracy level with the maximum membership is selected as the final accuracy judgment; the specific calculation expression is: Accuracy judgment = argmax(u L (p),u Q (p),u D (p)); where u L (p),u Q (p),u D (p) represents the fuzzy membership of low, medium and high accuracy judgments respectively. The function argmax returns the parameter with the maximum value, and the accuracy level with the maximum membership is selected as the final judgment result; The acquired ratio of potential faults is compared with a gradient threshold, where the gradient threshold includes a first threshold and a second threshold, and the first threshold is less than the second threshold, and the ratio of potential faults is compared with the first threshold and the second threshold respectively; If the proportion of potential faults is greater than the second threshold, the output is a high accuracy judgment; if the proportion of potential faults is greater than or equal to the first threshold and less than or equal to the second threshold, the output is a medium accuracy judgment; if the proportion of potential faults is less than the first threshold, the output is an accuracy judgment.

11. A transformer state assessment system based on multi-parameter data, used to implement the transformer state assessment method based on multi-parameter data according to any one of claims 1 to 10, characterized in that: It includes data acquisition module, anomaly monitoring module, evaluation module, matching module, state division module and visualization module; Data acquisition module: acquires various parameter data during transformer operation in real time, screens and cleans the acquired parameter data to extract key parameter information, and establishes a key parameter data set; Abnormal monitoring module: Set the observation time, and judge the influence of the load state of the transformer in different modes during the observation time on the response ability of key parameters. According to the actual response time changes of key parameters during the observation period, evaluate the abnormality of the response speed of key parameters. Evaluation module: comprehensively analyzes the influence of the transformer load state on the response capability of key parameters and the abnormality of the key parameter response speed, evaluates whether the transformer key parameter response state is abnormal, and distinguishes abnormal response state from normal response state based on the evaluation results; Matching module: For transformers in abnormal response states, a benchmark data set is established based on the historical operating data of the transformer, and a fault identification model is constructed to match the key parameter data set with the benchmark data set; State classification module: fault judgment is performed on the transformer in abnormal response state according to the matching results to distinguish between normal state and potential fault state; Visualization module: Based on the results of fault judgment, the calculation model accurately identifies the proportion of potential faults, and divides the accuracy of fault judgment according to the proportion of potential faults, dividing it into high accuracy judgment, medium accuracy judgment and low accuracy judgment, and visualizes the division results.

Images