Flexible DC converter valve submodule temperature anomaly detection method and device

By using sliding window statistics and iterative model updates on the temperature data of the flexible DC converter valve submodule, accurate detection of temperature anomalies in the flexible DC converter valve submodule was achieved, solving the problems of low detection efficiency and poor accuracy in existing technologies, and saving human and material resources.

CN119026020APending Publication Date: 2024-11-26ELECTRIC POWER RESEARCH INSTITUTE OF STATE GRID JIBEI ELECTRIC POWER CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411120879.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing methods for diagnosing voltage anomalies in flexible DC converter valve submodules suffer from low detection efficiency or poor accuracy, resulting in a waste of human and material resources.

Method used

By acquiring temperature data from the flexible DC converter valve submodule, numerical statistics are performed using a sliding window to establish a pre-trained model and statistical value matrix, an anomaly probability matrix is ​​obtained, and the dynamic fine-tuning matrix is ​​iteratively updated based on anomaly warning feedback data to achieve temperature anomaly detection.

Benefits of technology

This improves the accuracy and efficiency of temperature anomaly detection in the flexible DC converter valve submodule, saving human and material resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026020A_ABST
    Figure CN119026020A_ABST
Patent Text Reader

Abstract

The invention provides a flexible DC converter valve submodule temperature anomaly detection method and device, and the method comprises the steps: obtaining the temperature data of a flexible DC converter valve submodule, carrying out the numerical statistics of the temperature data in a sliding window manner, and obtaining a statistical value matrix according to the obtained numerical statistics result; and obtaining an abnormal probability matrix by using a pre-established pre-training model and the statistical value matrix, and performing temperature abnormity early warning according to the abnormal probability matrix, and according to the acquired abnormal early warning feedback data, performing iterative updating on a preset dynamic fine tuning matrix, and updating the pre-training model by using the dynamic fine tuning matrix after iterative updating. According to the method, the discrete data of the temperature of the flexible direct current converter valve sub-module is used, the model is established to perform temperature anomaly detection, and the value obtained by each operation result is utilized to perform model weight self-updating, so that accurate and efficient detection of the temperature anomaly of the flexible direct current converter valve sub-module is realized, and manpower and material resources are saved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of flexible direct current converter valves, and in particular to a method and device for detecting abnormal temperature of a submodule of a flexible direct current converter valve. Background Art

[0002] The abnormal voltage diagnosis technology of the flexible direct current converter valve submodule is an important branch in the field of power electronics, and is mainly used in the flexible direct current transmission (HVDC) system in power electronic devices. The flexible direct current transmission (HVDC) system is a system that uses direct current transmission technology to transmit high-capacity electricity over long distances. One of its key components is the flexible direct current converter, and the flexible direct current converter valve submodule is one of the components of the converter. It is very necessary to ensure the operating safety and stability of the HVDC system and meet the needs of power equipment status monitoring and fault diagnosis.

[0003] With the diversification of electric energy, HVDC systems have unique advantages in long-distance, large-capacity power transmission, so their application in electric energy transmission is becoming more and more extensive. The converter in the HVDC system is the core equipment connecting the AC system and the DC system, and its stability and reliability are crucial to the operation of the entire system. In order to ensure the safe operation of the HVDC system, it is crucial to promptly and accurately detect and diagnose abnormal voltage failures in the flexible direct current converter valve submodule to avoid possible system accidents and downtime losses. At present, existing abnormality diagnosis methods often use statistical methods, sensor technology, machine learning and deep learning algorithms. These methods have problems of low efficiency or poor accuracy in abnormality detection, as well as waste of human and material resources to varying degrees. Summary of the invention

[0004] In view of the problems existing in the prior art, the main purpose of the embodiments of the present invention is to provide a method and device for detecting temperature anomalies of a flexible direct current converter valve submodule, so as to achieve accurate detection of temperature anomalies and improve the efficiency and accuracy of anomaly detection.

[0005] In order to achieve the above object, an embodiment of the present invention provides a method for detecting abnormal temperature of a flexible direct current converter valve submodule, the method comprising:

[0006] Acquire the temperature data of the flexible direct current converter valve submodule, perform numerical statistics on the temperature data using a sliding window method, and obtain a statistical value matrix based on the obtained numerical statistical results;

[0007] Using the pre-established pre-training model and the statistical value matrix, an abnormal probability matrix is ​​obtained, and a temperature abnormality warning is performed according to the abnormal probability matrix;

[0008] According to the acquired abnormal warning feedback data, the preset dynamic fine-tuning matrix is ​​iteratively updated, and the pre-trained model is updated using the iteratively updated dynamic fine-tuning matrix.

[0009] Optionally, in an embodiment of the present invention, numerical statistics are performed on the temperature data by using a sliding window method, and according to the obtained numerical statistical results, a statistical value matrix is ​​obtained, including:

[0010] The temperature data is numerically counted using a sliding window method to determine the maximum temperature occurrence frequency value corresponding to each submodule in different time periods, and the maximum temperature occurrence frequency value corresponding to each submodule in different time periods is used as the numerical statistical result;

[0011] The numerical statistical results are converted into a stage matrix, and the ratio of each value in the stage matrix to its total matrix value is calculated to obtain a statistical value matrix.

[0012] Optionally, in an embodiment of the present invention, iteratively updating a preset dynamic fine-tuning matrix according to the acquired abnormal warning feedback data includes:

[0013] The dynamic fine-tuning matrix is ​​iteratively updated according to the acquired abnormal warning feedback data and the preset loss function.

[0014] Optionally, in one embodiment of the present invention, updating the pre-trained model using the iteratively updated dynamic fine-tuning matrix includes:

[0015] The iteratively updated dynamic fine-tuning matrix is ​​superimposed on the pre-trained model to obtain the target matrix, and the target matrix is ​​used as the pre-trained model for the next temperature anomaly detection.

[0016] The embodiment of the present invention further provides a device for detecting abnormal temperature of a flexible direct current converter valve submodule, the device comprising:

[0017] The statistical value matrix module is used to obtain the temperature data of the flexible direct current converter valve submodule, and perform numerical statistics on the temperature data by means of a sliding window, and obtain a statistical value matrix according to the obtained numerical statistical results;

[0018] The abnormal warning module is used to obtain an abnormal probability matrix using a pre-established pre-training model and a statistical value matrix, and to issue an abnormal temperature warning according to the abnormal probability matrix;

[0019] The model updating module is used to iteratively update the preset dynamic fine-tuning matrix according to the acquired abnormal warning feedback data, and to update the pre-trained model using the iteratively updated dynamic fine-tuning matrix.

[0020] Optionally, in one embodiment of the present invention, the statistical value matrix module includes:

[0021] A numerical statistics unit, used to perform numerical statistics on the temperature data by using a sliding window method, determine the maximum temperature occurrence frequency value corresponding to each submodule in different time periods, and use the maximum temperature occurrence frequency value corresponding to each submodule in different time periods as the numerical statistics result;

[0022] The statistical value matrix unit is used to convert the numerical statistical results into a stage matrix, and calculate the ratio of each value in the stage matrix to its total matrix value to obtain the statistical value matrix.

[0023] Optionally, in one embodiment of the present invention, the model updating module is further used to iteratively update the dynamic fine-tuning matrix according to the acquired abnormal warning feedback data and a preset loss function.

[0024] Optionally, in one embodiment of the present invention, the model updating module is also used to superimpose the iteratively updated dynamic fine-tuning matrix with the pre-trained model to obtain a target matrix, and use the target matrix as a pre-trained model for the next temperature anomaly detection.

[0025] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the above method is implemented when the processor executes the program.

[0026] The present invention also provides a computer-readable storage medium, which stores a computer program for executing the above method by a computer.

[0027] The present invention also provides a computer program product, comprising a computer program / instruction, which implements the steps of the above method when the computer program / instruction is executed by a processor.

[0028] The present invention uses discrete data of the temperature of the flexible direct current converter valve submodule and establishes a model for temperature anomaly detection according to the principle that the temperature position distribution of the submodule in the same valve tower is similar to the historical temperature distribution, and uses the value obtained from each calculation result to self-update the model weight, thereby achieving accurate and efficient detection of temperature anomalies of the flexible direct current converter valve submodule and saving manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 This is a flow chart of a method for detecting abnormal temperature of a flexible direct current converter valve submodule according to an embodiment of the present invention;

[0031] Figure 2 A flowchart of obtaining a statistical value matrix in an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the corresponding positions of the statistics of the number of occurrences of the maximum temperature of each submodule in a certain period of time in an embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of the structure of a device for detecting abnormal temperature of a flexible direct current converter valve submodule according to an embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of the structure of a statistical value matrix module in an embodiment of the present invention;

[0035] Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The embodiments of the present invention provide a method and device for detecting abnormal temperature of a flexible direct current converter valve submodule.

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] The following methods are often used in the prior art for anomaly detection: Sensor technology, using high-precision sensors to monitor the voltage of the flexible direct current converter valve submodule in real time, and realize sensitive detection of voltage anomalies. Data analysis and algorithms. Using advanced data analysis technology and algorithms, the monitored voltage data is processed and analyzed in real time to identify abnormal patterns and fault characteristics. Artificial intelligence and machine learning, using artificial intelligence and machine learning technology to establish a model for voltage anomaly diagnosis, and realize automatic identification and classification of different fault types. Online monitoring system, building an online monitoring system, integrating sensors, data analysis and diagnostic algorithms, to realize real-time monitoring and abnormal diagnosis of the voltage of the flexible direct current converter valve submodule. Machine learning and deep learning algorithms, such as neural networks, support vector machines (SVM), random forests, etc., are used in anomaly detection. These algorithms can be used to analyze historical location distribution data and learn the pattern of temperature anomalies from it. Time series analysis methods, including ARIMA (autoregressive integrated moving average model), exponential smoothing, etc., are used to perform trend analysis and periodicity analysis on historical temperature data to detect anomalies. Statistical methods, such as variance analysis and mean hypothesis test, are used to compare temperature data at different locations and determine whether there are anomalies. Sensor technology, including temperature sensors, location sensors, etc., is used to collect real-time data for anomaly detection. Data mining technology, including association rule mining and cluster analysis, is used to discover the correlation between temperature anomalies and other variables. The Internet of Things (IoT) and cloud computing technologies are used to monitor and analyze large amounts of sensor data in real time to detect anomalies in a timely manner. Model prediction methods predict the normal range of temperature by building a model and comparing it with the actual data to detect anomalies. Expert systems and rule engines use expert knowledge and domain rules to design corresponding anomaly detection rules to identify and alarm temperature anomalies.

[0039] like Figure 1 The flowchart of a method for detecting temperature anomaly of a flexible direct current converter valve submodule according to an embodiment of the present invention is shown. The execution subject of the method for detecting temperature anomaly of a flexible direct current converter valve submodule provided by the embodiment of the present invention includes but is not limited to a computer. The present invention uses discrete data of the temperature of the flexible direct current converter valve submodule, and establishes a model for temperature anomaly detection based on the principle that the temperature position distribution of the submodule in the same valve tower is similar to the historical temperature distribution, and uses the value obtained from each calculation result to self-update the model weight, thereby realizing accurate and efficient detection of temperature anomaly of the flexible direct current converter valve submodule, saving manpower and material resources. The method shown in the figure includes:

[0040] Step S1, obtaining temperature data of the flexible direct current converter valve submodule, and performing numerical statistics on the temperature data by using a sliding window method, and obtaining a statistical value matrix according to the obtained numerical statistical results;

[0041] Step S2, using the pre-established pre-training model and statistical value matrix to obtain an abnormal probability matrix, and perform temperature abnormality warning according to the abnormal probability matrix;

[0042] Step S3, iteratively updating the preset dynamic fine-tuning matrix according to the acquired abnormal warning feedback data, and updating the pre-trained model using the iteratively updated dynamic fine-tuning matrix.

[0043] As an embodiment of the present invention, Figure 2 As shown, the temperature data is numerically counted by using a sliding window, and according to the obtained numerical statistical results, a statistical value matrix is ​​obtained including:

[0044] Step S11, using a sliding window to perform numerical statistics on the temperature data, determine the maximum temperature occurrence frequency value corresponding to each submodule in different time periods, and use the maximum temperature occurrence frequency value corresponding to each submodule in different time periods as the numerical statistical result;

[0045] Step S12, converting the numerical statistical results into a stage matrix, and calculating the ratio of each value in the stage matrix to the total value of the matrix to obtain a statistical value matrix.

[0046] Among them, the temperature data of the flexible direct current converter valve submodule is collected in a conventional way, and the temperature data is discrete data. First, the number of occurrences of all temperature data is counted, and it is determined that the temperature of each valve tower submodule in the data tends to be the maximum or minimum in the long-term data, and the maximum and minimum values ​​will basically not appear at the same time. After meeting this condition, the temperature data is numerically counted by sliding window method, and the frequency of occurrence value in each time period is calculated, that is, the numerical statistical result. Therefore,

[0047] Furthermore, the numerical statistical results are constructed into a matrix to obtain the matrix value calculated in the first stage, that is, the stage matrix. The ratio of each value in the stage matrix to the total value (the total number of statistics in each time period) is calculated to obtain the statistical value matrix.

[0048] Furthermore, the pre-established pre-trained model is obtained by training with the historical temperature data of the flexible direct current converter valve submodule. Specifically, an initial matrix, i.e., the initial model, is defined by the time series weight, the abnormal redundancy degree, and the abnormal importance. The initial matrix is ​​then trained with the historical temperature data, i.e., the weights of the initial matrix are updated, for example, with the use of a preset loss function to update the weights, thereby obtaining a pre-trained model, which can also be called a pre-trained matrix.

[0049] Furthermore, the product of the statistical value matrix and the pre-trained model is obtained, and the resulting matrix is ​​the abnormality probability matrix. Each row in the abnormality probability matrix represents the probability of a certain abnormality in each submodule. Therefore, combined with the warning threshold pre-set according to actual needs, the abnormal probability is compared with the warning threshold. If it exceeds the warning threshold, a temperature abnormality warning is issued, so that the operation and maintenance personnel can be informed of the temperature abnormality in time.

[0050] As an embodiment of the present invention, iteratively updating a preset dynamic fine-tuning matrix according to the acquired abnormal warning feedback data includes: iteratively updating the dynamic fine-tuning matrix according to the acquired abnormal warning feedback data and a preset loss function.

[0051] Among them, the abnormal warning feedback data is obtained. Specifically, the abnormal warning feedback data is the time interval from the warning to the operation and maintenance personnel responding to the warning and handling the abnormality. The abnormal warning feedback data and the preset loss function are used to update the weights of the preset dynamic fine-tuning matrix. Among them, if the abnormal warning is a minor problem, the operation and maintenance personnel may not handle it in time, but wait until the overhaul to handle it together, which means that the abnormality is not important or may be a misjudgment. However, if the abnormality is handled immediately after it occurs, it means that the abnormality is correct and has a high weight. This idea is used to perform self-iterative updates of the dynamic fine-tuning matrix.

[0052] Specifically, the pre-trained model is used for calculation. When the pre-trained model calculated an abnormal situation last time and the current situation is also an abnormal situation, and the difference in the probability of the two abnormal situations is less than a certain value, the value before the iteration is temporarily stored in the dynamic fine-tuning matrix, and the dynamic fine-tuning matrix is ​​iterated according to the continuous time length, and the actual situation is assumed to be a normal situation for iteration until the final pre-trained model (i.e., the fusion model) no longer has the abnormal situation. Among them, the dynamic fine-tuning matrix iterative update process can use a preset loss function to judge the difference between the values ​​in the matrix.

[0053] As an embodiment of the present invention, updating the pre-trained model using the iteratively updated dynamic fine-tuning matrix includes: superimposing the iteratively updated dynamic fine-tuning matrix with the pre-trained model to obtain a target matrix, and using the target matrix as the pre-trained model for the next temperature anomaly detection.

[0054] Among them, the dynamic fine-tuning matrix obtained after the weight update is superimposed on the pre-trained model to obtain the target matrix (i.e., the fusion model / target model), and the target matrix is ​​used again as the pre-trained model for the next temperature anomaly detection, thereby continuously improving the prediction accuracy of the model and improving the efficiency and accuracy of temperature anomaly detection.

[0055] In a specific embodiment of the present invention, the temperature anomaly detection process of the flexible direct current converter valve submodule includes:

[0056] First, for the discrete data of sub-module temperature data collected, the amount that can be collected is usually limited. For example, only five maximum and minimum values ​​of data are collected at each moment. To determine whether the data is usable, it is necessary to conduct preliminary regularity verification of the data. The method is as follows.

[0057] 1) The number of occurrences of all data is counted, and it is determined that the temperature of each submodule of the same valve tower in the data tends to be close to the maximum or minimum in the long-term data, and the maximum and minimum values ​​basically do not appear at the same time. The statistical example is shown in Table 1.

[0058] Table 1

[0059] Device Number Maximum temperature Minimum temperature Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 139 4271 0 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 47 0 404 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 21 0 1059 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 137 2486 0 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 19 0 2856 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 1 0 101 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 2 0 1936 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 17 0 2580 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 8 0 2557 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 3 0 1634 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 27 5045 0 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 129 1923 0 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 96 0 2094 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 138 4052 0 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 20 0 2680 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 5 0 2419 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 9 0 264 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 99 1107 0 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 23 0 2766 Pole 2_Phase A_Tower T1_Bridge Arm 2_Device Number 6 0 147

[0060] According to the above results, if the verification is correct, a numerical statistics is performed once in each time period, such as Figure 3 The figure shows the corresponding position diagram of the maximum temperature (similar to the minimum value) of each submodule in a certain period of time. Each point corresponds to a submodule in the valve tower. The redder the point, the higher the frequency of the maximum value at that position in that period of time. Therefore, according to this calculation method, the sliding window method is used to calculate the frequency of occurrence of each period of time.

[0061] At this point, the matrix value calculated in the first stage, that is, the stage matrix, is obtained. The ratio of each value in the matrix to the total value (the total number of statistics in each time period) is calculated to obtain the statistical value matrix. The matrix shape is N*T, where N is the number of submodules and T is the number of windows. The matrix is ​​denoted as A below N*T Then, an initial matrix P is defined based on the time series weight, abnormal redundancy and abnormal importance. T*M , where T is the importance of each time period, M is the impact of the specific abnormal situation in each time period, and

[0062] Among them, if the matrix defines P T*M Then each child element should be defined as p, not a. i,* , in fact, is p i,* , which means that the dimension of each row vector is [M*1], the formula This means that the vector represents the weights of each anomaly in the time period i. The condition is that the sum of the sub-elements of the vector is 1. This ensures that the weight factor of a certain time period becomes the only determining factor of a certain weight, and that the weights corresponding to each time period do not differ too much from each other to affect the final calculation and judgment.

[0063] It should be noted that the initial matrix P T*M There is no need to define specific values. From a mechanistic point of view, the matrix P T*M It is a mixed weight matrix containing time series weight, abnormal redundancy and abnormal importance. As for the initial matrix P T*M What value should be in the , this can be randomly initialized by yourself (generally speaking, a random number is defined, or it can be composed of vectors based on the weights of the corresponding results of the time series under normal circumstances, such as (0.1, 0.2, 0.3, ..., 1)). However, it must be satisfied that the initialized number must be standardized at the end (that is, satisfy the formula

[0064] Furthermore, matrix initialization also requires some skills. For example, a matrix with all zeros will cause the weights to fail to be updated by gradients. However, since the formula restrictions are set here, there is no need to consider too much. The value of the initial matrix is ​​not very important because the matrix needs to be updated according to the weights after the algorithm is running. In other words, the weight gradient descent is the same as the gradient descent of machine learning and deep learning. The reason why the corresponding model is not used is because of Occam's razor principle. In other words, this problem does not need to be deliberately complicated to implement. Iterating a weight matrix is ​​enough.

[0065] Specifically, the time series weight actually means the initial matrix P T*M , assuming that the column vector p of the jth column *,j The dimension is [1*T], which means the weight corresponding to each time period in the jth anomaly is how much, there are T time periods in total, so the dimension is [1*T]. Since the present invention uses data from multiple time periods as input, the inputs from these multiple time periods will definitely have different weight effects on the required final result. Generally speaking, the farther away the time is, the lower the weight will be, and the closer the time is to the current time, the higher the weight will be.

[0066] Furthermore, the abnormal redundancy and abnormal importance are actually the initial matrix P T*M The row vector in , assuming that the row vector of the i-th column is p i,* , the dimension is [M*1], then this row vector represents the sum of the weights of the anomaly redundancy and anomaly importance of each anomaly corresponding to the i-th time period. There are M anomalies in total, so the dimension is [M*1].

[0067] In addition, the degree of abnormal redundancy means that since each abnormality may have similar implicit representations, there may be redundancy between weights (that is, the matrix may be insufficient in rank). In order to make the abnormal probabilities calculated from a piece of data different, rather than completely the same or too different, some restrictions need to be made intentionally during self-iteration. The importance of abnormality means that the abnormalities detected in different situations must have different importance on the operation of the station, so it is necessary to slowly obtain the required target weight matrix, that is, the target matrix, by using the subsequent pre-training + self-iteration method.

[0068] With the above results, we get the statistical data matrix A N*T , the initial matrix P of time series weights, abnormal redundancy and abnormal importance T*M , perform matrix multiplication on the above matrix to obtain the following matrix, where each row represents the probability of a certain anomaly occurring in each module, that is, the anomaly probability matrix:

[0069] R N*M =A N*T *P T*M (1)

[0070] After obtaining the above calculation formula, it can be seen that in the calculation content, only R T*M is a non-fixed value and already has According to the constraints of R N*M , through the preset loss function:

[0071]

[0072] Make P T*M Each item r i,j Compared with the actual abnormal situation i,j After the algorithm performs preliminary anomaly fitting through historical data, it can dynamically adjust its value through the loss function according to the time and situation of abnormal situation elimination under actual operation. By calculating the actual results with the pre-trained model, a dynamic fine-tuning matrix P′ is set. T*M P with pre-trained model T*M Perform superposition operation, and each time the actual operation is performed, P in the model T*M Instead of iterating, iterate P ′ T*M .

[0073] It should be noted that the definition of P T*M The matrix includes P 目标(T*M) and P 过程(T*M) , and the initial matrix should be P 初始(T*M) , all three are P T*MThe matrix is ​​just what the matrix looks like at different times.

[0074] Among them, the difference between these three is first 初始(T*M) is the P matrix randomly initialized according to the requirements, that is, the initial matrix; P 目标(T*M) is the expected P of the target T*M Matrix, that is, the target matrix; P 过程(T*M) is the intermediate matrix obtained through pre-training, that is, the pre-training matrix (pre-training model). In order to obtain the final P 目标(T*M) , need to pass P 过程(T*M) Plus a P′ obtained by self-iteration T*M Matrix, thus obtaining the final matrix P of the target 目标(T*M) .

[0075] In addition, in actual operation, the process of judging abnormal situations is an iterative process. The initial judgment of abnormal situations is based on the P obtained by the pre-trained model. 过程(T*M) Matrix, calculated by formula (1) to get P N*M Afterwards, according to R N*M The probability of abnormal occurrence in the corresponding method determines whether to issue an abnormal alarm.

[0076] Among them, if it is a correct abnormal alarm, then the operation and maintenance personnel will definitely handle it no matter how long it takes, and the weight corresponding to the abnormality will be iterated according to the length of time between the operation and maintenance personnel's handling time and the alarm time. In fact, if it is a minor problem, the operation and maintenance personnel may not handle it in time, but wait until the overhaul to handle it together, which means that the abnormality is not important or may be a misjudgment. However, if the abnormality is handled immediately after it occurs, it means that the abnormality is correct and has a high weight. This idea is used to dynamically fine-tune the matrix P′ T*M The specific process of judging abnormal situations is to obtain R N*M , check whether each abnormal probability value in this matrix is ​​too high, and if so, issue an alarm.

[0077] The model in the present invention has a self-iterative function, and is not fixed after obtaining the pre-trained model. Instead, it automatically adjusts the dynamic fine-tuning matrix P′ according to the subsequent encounters. T*M , and then according to the dynamic fine-tuning matrix P′ T*M and P in the pre-trained model 过程(T*M) Get the final P 目标(T*M) The matrix is ​​then calculated according to the algorithm to obtain the final result, which can greatly improve the prediction accuracy of the model and improve the efficiency and accuracy of temperature anomaly detection.

[0078] In this embodiment, the iteration method is as follows: first, the pre-trained model is used for calculation. When the pre-trained model calculates an abnormal situation last time and this time is also an abnormal situation, and the difference in the probability of the two abnormal situations is less than a certain value, P′ is temporarily stored. T*M The value before iteration, P′ according to the continuous time length T*M Iterate the actual situation as normal and iterate until the fusion model (i.e., the target matrix) no longer has the anomaly. If the probability of abnormal pre-training model results decreases during this period (or the operation and maintenance personnel record the abnormal situation), then P′ T*M Replace it with the temporary value, and consider the time period as an abnormal situation for model fitting. The actual displayed result is the numerical result obtained after the target matrix. According to the behavior of fine-tuning the self-learning coefficient under the actual operation of the model, the impact of false alarms and non-concern anomalies can be minimized.

[0079] Among them, the specific fitting method is gradient descent. According to the above loss function, P is gradually updated through the gradient descent method. T*M The values ​​in the matrix are the same whether it is the first step of fitting historical data or the subsequent fine-tuning of the matrix.

[0080] The present invention uses discrete data of the temperature of the converter valve submodule, based on the principle that the temperature position distribution of the submodule in the same valve tower is similar to the historical temperature distribution and is related to the direction of the water cooling pipe, and uses the value obtained by each calculation result to self-update the temperature position weight, thereby detecting the temperature anomaly of the converter valve submodule. Using the temperature anomaly detection method of the present invention, it is possible to timely discover submodules with different submodule regularities and submodules with regularities that are inconsistent with normal from discrete data, and issue an early warning.

[0081] Aiming at the problem of early warning of flexible direct current converter valve submodule, the present invention proposes an adaptive flexible direct current converter valve submodule temperature anomaly detection algorithm based on historical position distribution. At the same time, statistical thinking, anomaly detection algorithm and self-learning thinking are used to propose a basis for judging submodule temperature anomaly based on historical converter valve submodule temperature changes, and the above steps are used to realize the fault early warning of flexible direct current submodule.

[0082] The present invention uses a smaller amount of more discrete data. It can be seen that through the method of the present invention, a smaller number of data (maximum and minimum values ​​of 10 measuring points) than the total number of sub-modules of each valve tower (generally more than 100 measuring points) can be used for calculations, and there is no need to use massive data for anomaly detection, which solves the problems of difficult data storage and low algorithm calculation efficiency.

[0083] The present invention performs fault diagnosis and early warning on the flexible direct current converter valve through an algorithm. In view of the fact that most algorithms of the flexible direct current converter valve are currently state monitoring or formula mechanism derivation, there are not many algorithms, and since the flexible direct current technology is a new technology that has only been put into practical use recently, there are very few problems with fault cases. The present invention starts with the relationship that should exist between sub-modules, studies the potential relationship of the data, and evaluates the distribution of the sub-module data through statistical methods, and can simultaneously realize the functions of fault diagnosis and fault early warning.

[0084] The present invention is an effective early warning means for flexible DC converter valves. Currently, flexible DC converter valves are all based on state monitoring and formula mechanism derivation, and effective early warning strategies cannot be obtained. The present invention studies the potential relationship of data, can timely identify the occurrence pattern of potential imbalance of sub-module voltages, and give timely early warnings, which can effectively reduce the damage frequency of sub-module equipment.

[0085] The present invention uses a self-learning method to perform self-iteration of the fault diagnosis model. During actual operation, the present invention can perform self-learning based on the continuity of the model's abnormal judgment of data, and according to the module abnormality repair and other situations, the probability judgment of the original model can be used to continuously enhance the judgment ability of abnormal situations that the operation and maintenance personnel are concerned about, and continuously repair its own handling of misjudgment situations.

[0086] like Figure 4 The figure shows a schematic diagram of the structure of a device for detecting abnormal temperature of a flexible direct current converter valve submodule according to an embodiment of the present invention. The device shown in the figure includes:

[0087] The statistical value matrix module 10 is used to obtain the temperature data of the flexible direct current converter valve submodule, and perform numerical statistics on the temperature data by means of a sliding window, and obtain a statistical value matrix according to the obtained numerical statistical results;

[0088] The abnormal warning module 20 is used to obtain an abnormal probability matrix by using the pre-established pre-training model and the statistical value matrix, and to perform temperature abnormality warning according to the abnormal probability matrix;

[0089] The model updating module 30 is used to iteratively update the preset dynamic fine-tuning matrix according to the acquired abnormal warning feedback data, and to update the pre-trained model using the iteratively updated dynamic fine-tuning matrix.

[0090] As an embodiment of the present invention, Figure 5 As shown, the statistical value matrix module 10 includes:

[0091] The numerical statistics unit 11 is used to perform numerical statistics on the temperature data by using a sliding window method, determine the maximum temperature occurrence frequency value corresponding to each submodule in different time periods, and use the maximum temperature occurrence frequency value corresponding to each submodule in different time periods as the numerical statistics result;

[0092] The statistical value matrix unit 12 is used to convert the numerical statistical results into a stage matrix, and calculate the ratio of each value in the stage matrix to its total matrix value to obtain a statistical value matrix.

[0093] As an embodiment of the present invention, the model updating module is also used to iteratively update the dynamic fine-tuning matrix according to the acquired abnormal warning feedback data and the preset loss function.

[0094] As an embodiment of the present invention, the model updating module is also used to superimpose the iteratively updated dynamic fine-tuning matrix with the pre-trained model to obtain a target matrix, and use the target matrix as a pre-trained model for the next temperature anomaly detection.

[0095] Based on the same application concept as the above-mentioned method for detecting abnormal temperature of a flexible direct current converter valve submodule, the present invention also provides the above-mentioned device for detecting abnormal temperature of a flexible direct current converter valve submodule. Since the principle of solving the problem by the device for detecting abnormal temperature of a flexible direct current converter valve submodule is similar to that of the method for detecting abnormal temperature of a flexible direct current converter valve submodule, the implementation of the device for detecting abnormal temperature of a flexible direct current converter valve submodule can refer to the implementation of the method for detecting abnormal temperature of a flexible direct current converter valve submodule, and the repeated parts will not be repeated.

[0096] The present invention uses discrete data of the temperature of the flexible direct current converter valve submodule and establishes a model for temperature anomaly detection according to the principle that the temperature position distribution of the submodule in the same valve tower is similar to the historical temperature distribution, and uses the value obtained from each calculation result to self-update the model weight, thereby achieving accurate and efficient detection of temperature anomalies of the flexible direct current converter valve submodule and saving manpower and material resources.

[0097] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the above method is implemented when the processor executes the program.

[0098] The present invention also provides a computer program product, comprising a computer program / instruction, which implements the steps of the above method when the computer program / instruction is executed by a processor.

[0099] The present invention also provides a computer-readable storage medium, which stores a computer program for executing the above method by a computer.

[0100] like Figure 6 As shown, the electronic device 600 may further include: a communication module 110, an input unit 120, an audio processor 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily have to include Figure 6In addition, the electronic device 600 may also include Figure 6 For components not shown, reference may be made to the prior art.

[0101] like Figure 6 As shown, the central processor 100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor devices and / or logic devices. The central processor 100 receives inputs and controls the operations of various components of the electronic device 600.

[0102] The memory 140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 100 may execute the program stored in the memory 140 to implement information storage or processing.

[0103] The input unit 120 provides input to the CPU 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to provide power to the electronic device 600. The display 160 is used to display display objects such as images and text. The display may be, for example, an LCD display, but is not limited thereto.

[0104] The memory 140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 140 may also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application / function storage unit 142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 600 through the central processor 100.

[0105] The memory 140 may also include a data storage unit 143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 144 of the memory 140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0106] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processor 100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0107] Based on different communication technologies, multiple communication modules 110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module. The communication module (transmitter / receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide an audio output via the speaker 131 and receive an audio input from the microphone 132, thereby realizing a common telecommunication function. The audio processor 130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 130 is also coupled to the central processor 100, so that the sound can be recorded on the local machine through the microphone 132, and the sound stored on the local machine can be played through the speaker 131.

[0108] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for detecting abnormal temperature of a flexible direct current converter valve submodule, characterized in that: The method comprises: Acquire temperature data of the flexible direct current converter valve submodule, perform numerical statistics on the temperature data by using a sliding window, and obtain a statistical value matrix according to the obtained numerical statistical results; Using the pre-established pre-training model and the statistical value matrix, an abnormal probability matrix is ​​obtained, and a temperature abnormality warning is performed according to the abnormal probability matrix; According to the acquired abnormal warning feedback data, the preset dynamic fine-tuning matrix is ​​iteratively updated, and the pre-trained model is updated using the iteratively updated dynamic fine-tuning matrix.

2. The method according to claim 1, characterized in that The temperature data is numerically counted by using a sliding window, and according to the obtained numerical statistical results, a statistical value matrix is ​​obtained, including: Perform numerical statistics on the temperature data by using a sliding window method, determine the maximum temperature occurrence frequency value corresponding to each submodule in different time periods, and use the maximum temperature occurrence frequency value corresponding to each submodule in different time periods as the numerical statistical result; The numerical statistical results are converted into a stage matrix, and the ratio of each value in the stage matrix to the total value of the matrix is ​​calculated to obtain the statistical value matrix.

3. The method according to claim 1, characterized in that According to the acquired abnormal warning feedback data, the preset dynamic fine-tuning matrix is ​​iteratively updated including: The dynamic fine-tuning matrix is ​​iteratively updated according to the acquired abnormal warning feedback data and the preset loss function.

4. The method according to claim 1, characterized in that: Updating the pre-trained model using the iteratively updated dynamic fine-tuning matrix includes: The iteratively updated dynamic fine-tuning matrix is ​​superimposed on the pre-trained model to obtain a target matrix, and the target matrix is ​​used as a pre-trained model for the next temperature anomaly detection.

5. A device for detecting abnormal temperature of a flexible direct current converter valve submodule, characterized in that: The device comprises: A statistical value matrix module is used to obtain temperature data of the flexible direct current converter valve submodule, and to perform numerical statistics on the temperature data by means of a sliding window, and to obtain a statistical value matrix according to the obtained numerical statistical results; An abnormal warning module is used to obtain an abnormal probability matrix by using a pre-established pre-training model and the statistical value matrix, and to perform temperature abnormality warning according to the abnormal probability matrix; The model updating module is used to iteratively update the preset dynamic fine-tuning matrix according to the acquired abnormal warning feedback data, and to update the pre-trained model using the iteratively updated dynamic fine-tuning matrix.

6. The device according to claim 5, characterized in that The statistical value matrix module includes: A numerical statistics unit, used to perform numerical statistics on the temperature data by using a sliding window method, determine the maximum temperature occurrence frequency value corresponding to each submodule in different time periods, and use the maximum temperature occurrence frequency value corresponding to each submodule in different time periods as the numerical statistics result; The statistical value matrix unit is used to convert the numerical statistical results into a stage matrix, and calculate the ratio of each value in the stage matrix to its total matrix value to obtain the statistical value matrix.

7. The device according to claim 5, characterized in that The model updating module is also used to iteratively update the dynamic fine-tuning matrix according to the acquired abnormal warning feedback data and a preset loss function.

8. The device according to claim 5, characterized in that The model updating module is also used to superimpose the iteratively updated dynamic fine-tuning matrix with the pre-trained model to obtain a target matrix, and use the target matrix as a pre-trained model for the next temperature anomaly detection.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for causing a computer to execute the method according to any one of claims 1 to 4.

11. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.