Method and device for analyzing correlation of multi-source time series data of power equipment
Through time lag correction and multi-branch attention mechanism, combined with Gumbel reparameterized sampling and graph structure adjacency matrix, the complex dependence problem of multi-source time series data of power equipment is solved, and the accuracy and reliability of correlation analysis are improved.
Patent Information
- Application Number
- CN202510685033.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-26
- Publication Date
- 2025-09-05
AI Technical Summary
The prior art is difficult to effectively deal with the complex dependence and correlation between multi-source time series data of power equipment, especially considering the time lag effect and long-distance dependencies.
Time lag correction, Gumbel reparameterized sampling and multi-branch attention mechanism are used to analyze the multi-source timing data of power equipment through graph structure adjacency matrix and transformer model to build complex dependencies.
It improves the accuracy of multi-source timing data correlation analysis of power equipment, captures hidden relationships, conforms to actual physical processes, and enhances the reliability of analysis results.
Smart Images
Figure CN120596828A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of health monitoring of electric power equipment, and relates to a method and device for correlation analysis of multi-source time series data of electric power equipment. Background Art
[0002] The safe and stable operation of power equipment is crucial to the power system. Health monitoring of power equipment typically involves a variety of sensors to monitor physical quantities such as voltage, current, power, temperature, transformer oil temperature, partial discharge, vibration, and oil gas content. The data collected by these sensors forms a time series (also known as time series data), reflecting the equipment's operating status.
[0003] Time series represent various physical quantities of the same electrical equipment, so their changing trends often exhibit correlations. However, the correlations between multiple time series fall into various patterns, making it difficult to effectively handle complex dependencies between data. For example, the correlations between electrical quantities change rapidly, while those between temperature and dissolved gas in oil change more slowly and exhibit time lags. When a partial discharge occurs within a transformer, the corresponding partial discharge sensor immediately detects the electromagnetic signal, and the acoustic sensor also detects high-frequency acoustic signals. However, dissolved gas in oil requires a certain lag before being detected by the corresponding sensor. Some time series can be considered free variables, while others depend on multiple other time series. For example, power depends on the external load, while transformer oil temperature depends on both power and ambient temperature. Under normal operating conditions, some time series may not exhibit significant correlations. For example, when the load is within the limit, dissolved gas in oil does not change significantly with load. However, when the load is overloaded, characteristic gases such as hydrogen are generated in the oil. Summary of the Invention
[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method and device for correlation analysis of multi-source time series data of power equipment, aiming to effectively handle the complex dependencies and correlations between multi-source time series data involved in health monitoring of power equipment.
[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:
[0006] In a first aspect, the present invention provides a method for analyzing the correlation of multi-source time series data of power equipment, comprising:
[0007] Collect operating sequence data of power equipment;
[0008] Performing time lag correction on the runtime series data;
[0009] Use Gumbel reparameterized sampling to estimate the connection probability between the corrected runtime series data and generate the adjacency matrix of the graph structure;
[0010] The adjacency matrix is input into a graph encoding model composed of dilated convolution and graph convolution, and the obtained output and the runtime data are input into a transformer model together to obtain an analysis result of the correlation of the runtime data.
[0011] Furthermore, the operating time series data includes: time series data of voltage, current, power, temperature, transformer oil temperature, partial discharge, vibration and dissolved gas content in oil; the operating time series data is collected from the running electrical equipment through multiple sensors.
[0012] Furthermore, the time lag correction is performed on the running time series data, including: setting a time lag parameter for each running time series data according to the physical characteristics and historical data characteristics of each running time series data , to affect the time series data with delay Corrected to .
[0013] Furthermore, the time lag parameter The determination method includes: setting the time lag parameter As a hyperparameter of the transformer model, the time lag parameter is determined by a hyperparameter search method during the training process. The value of .
[0014] Furthermore, before performing time lag correction on the runtime data, the method further includes: preprocessing the collected runtime data and normalizing the preprocessed data;
[0015] The preprocessing includes: data cleaning, missing value filling and outlier detection.
[0016] Furthermore, Gumbel reparameterized sampling is used to estimate the connection probability between the runtime series data to generate an adjacency matrix of a graph structure, including:
[0017] The runtime series data consists of M time series. Gumbel reparameterized sampling is used to generate the correlation dependency probability of any two time series in the runtime series data. The generated M*M correlation dependency probabilities constitute an M×M graph structure adjacency matrix.
[0018] Furthermore, the transformer model includes a multi-branch attention mechanism;
[0019] The multi-branch attention mechanism includes global learning attention and local neighborhood attention, which is expressed as:
[0020] ,
[0021] in, is the local neighborhood attention, Learning attention for global context; Represents a dimension concatenation operation.
[0022] In a second aspect, the present invention further provides a device for analyzing the correlation of multi-source time series data of power equipment, the device comprising:
[0023] An operation timing data acquisition module is used to collect operation timing data of power equipment;
[0024] A time lag correction module, configured to perform time lag correction on the runtime series data;
[0025] The adjacency matrix generation module is used to estimate the connection probability between the corrected runtime series data using Gumbel reparameterized sampling and generate the adjacency matrix of the graph structure;
[0026] The correlation analysis result acquisition module is used to input the adjacency matrix into a graph coding model composed of dilated convolution and graph convolution, and input the obtained output together with the runtime series data into the converter model to obtain the analysis result of the correlation of the runtime series data.
[0027] In a third aspect, the present invention further provides a computer device, comprising:
[0028] memory for storing computer programs;
[0029] The processor is used to execute the computer program to implement the steps of the above-mentioned method for correlation analysis of multi-source time series data of power equipment.
[0030] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the steps of the above-mentioned method for analyzing the correlation of multi-source time series data of power equipment.
[0031] Compared with the prior art, the present invention has the following beneficial effects:
[0032] The method for correlation analysis of multi-source time series data of power equipment provided by the present invention introduces a time lag correction parameter, takes into account the time series lag effect of the physical quantity represented by the time series, and makes the correlation analysis results more consistent with the actual physical process; based on the connection probability between the running time series data, an adjacency matrix of a graph structure is constructed, which can automatically learn the complex dependencies between the running time series data and effectively capture the hidden associations of multi-source time series data; the multi-branch attention mechanism is adopted to better capture long-distance dependencies, while taking into account global and local information, thereby improving the accuracy of correlation analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 A flowchart of a method for analyzing the correlation of multi-source time series data of power equipment provided by an embodiment of the present invention;
[0034] Figure 2 Schematic diagram of the framework of the graph coding model and the converter model in the embodiment of the present invention;
[0035] Figure 3 A schematic diagram of the structure of a device for analyzing the correlation of multi-source time series data of electric power equipment provided by an embodiment of the present invention;
[0036] Figure 4 This is a diagram of the internal structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0037] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. The same reference numerals in the drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. The embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations of the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.
[0038] The term "and / or" in this document simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. Additionally, the character " / " in this document generally indicates that the related objects are in an "or" relationship.
[0039] Example 1:
[0040] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a method for analyzing the correlation of multi-source time series data of power equipment. Figure 1 This is a flow chart of the method for analyzing the correlation of multi-source time series data of power equipment. This flow chart only shows the logical sequence of the method described in this embodiment. Under the premise of no conflict, in other possible embodiments of the present invention, different methods can be used. Figure 1 The steps shown or described are accomplished in the order shown.
[0041] The method for analyzing the correlation between multi-source time series data of power equipment provided in this embodiment can be applied to a terminal and can be executed by a device for analyzing the correlation between multi-source time series data of power equipment. The device can be implemented by software and / or hardware and can be integrated into the terminal.
[0042] See also Figure 1 The method of the embodiment of the present invention specifically includes the following steps:
[0043] Step 1: Collect the operating sequence data of the power equipment.
[0044] This operational time series data is collected from operating electrical equipment using a variety of sensors. For example, during transformer operation, voltage and current data are collected via sensors such as voltage transformers and current transformers installed at the high- and low-voltage terminals. Power sensors collect power data, temperature sensors collect winding temperature and transformer oil temperature data, oil level sensors monitor oil level changes, partial discharge sensors detect partial discharge, sound sensors capture sound signals, and gas sensors analyze the gas content in the oil (methane, ethane, ethylene, acetylene, hydrogen, total hydrocarbons, etc.). These sensors collect data according to their respective sampling intervals (e.g., voltage, current, and power may be sampled every millisecond, temperature may be sampled every second, and oil gas content may be sampled every hour), generating time series data of varying lengths and characteristics.
[0045] The present invention performs preprocessing such as data cleaning, missing value filling, and outlier detection on the collected runtime series data, and normalizes the preprocessed data.
[0046] Specifically, the collected data is first checked to remove obvious errors or unreasonable data points. For occasional missing values, linear interpolation is used to fill them. Outlier detection and correction includes: calculating the mean of each time series and standard deviation , will exceed Data points with a large range are considered as outliers and are corrected based on the outlier neighboring data points.
[0047] The formula for normalization is:
[0048] ,
[0049] in, is the normalized data, are the data before normalization.
[0050] Considering the different sampling periods of different time series, it is necessary to fit and resample each normalized time series. For time series with different change patterns, linear piecewise fitting or polynomial fitting methods can be used. At the same timestamp, the fitting of each time series is resampled. Suppose there are M time series. After resampling, a series of M-dimensional vectors are obtained, and the value of each dimension comes from the corresponding time series.
[0051] Step 2: Perform time lag correction on the runtime series data.
[0052] Performing time lag correction on the running series data includes: setting time lag parameters for each running series data according to the physical characteristics and historical data characteristics of each running series data , to affect the time series data with delay Corrected to .
[0053] use Instead of inputting the original time series into the transformer, the correlation between the two time series data can be analyzed more accurately, resulting in better performance.
[0054] Methods for determining the time lag parameter include: If the correlation between two time series varies monotonically with the time lag parameter, a binary search can be used to quickly find the optimal time lag parameter. For more general correlation relationships, a sliding window search can be used to determine the time lag parameter. Furthermore, the time lag parameter can be used as a hyperparameter of the transformer model and determined during training using a hyperparameter search method (such as a grid search).
[0055] Step 3: Use Gumbel reparameterization sampling to estimate the connection probability between the corrected runtime series data and generate the adjacency matrix of the graph structure.
[0056] Gumbel reparameterization sampling solves the gradient backpropagation problem in deep learning models by converting the discrete sampling process into a continuous and differentiable relaxation form.
[0057] Specifically, the Gumbel reparameterization sampling strategy is used to generate the correlation dependence probability of any two time series in the runtime data, assuming that and The length of the alignment is For two time series, the Gumbel reparameterization method is used to generate the correlation dependency probabilities of the two time series and . A total of M*M correlation dependency probabilities are generated for M time series, forming an M×M graph structure adjacency matrix. The adjacency matrix fully represents the correlation dependency relationship between each time series.
[0058] Step 4: Input the adjacency matrix into a graph encoding model composed of dilated convolution and graph convolution, and input the obtained output together with the runtime data into a transformer model to obtain analysis results of the correlation of the runtime data.
[0059] In order to adjust the parameters of the graph structure during training, the adjacency matrix is input into the graph encoding model. The structure of the graph encoding model is as follows: Figure 2 As shown in Figure 1, it is composed of multiple dilated convolutions and graph convolutions in series, and residual connections are made between the dilated convolutions and graph convolutions.
[0060] At the same time, the original runtime data is input into the graph encoding model to extract the features contained therein.
[0061] like Figure 2 As shown, the transformer model includes encoder 1, encoder 2, encoder 3 and decoder connected in sequence.
[0062] In order to capture long-range dependencies within a time series, the encoder 1 of the present invention includes a multi-branch attention mechanism, which includes global learning attention and local neighborhood attention, expressed as:
[0063] ,
[0064] in, is the local neighborhood attention, Learning attention for global context; Represents a dimension concatenation operation.
[0065] Example 2:
[0066] Based on the same inventive concept as Example 1, this embodiment of the present invention also provides a device for analyzing the correlation between multi-source time series data of power equipment, which is used to implement the above-mentioned method for analyzing the correlation between multi-source time series data of power equipment. The implementation solution provided by this device is similar to the implementation solution described in the above-mentioned method. Therefore, the specific limitations in the embodiment of the device for analyzing the correlation between multi-source time series data of power equipment provided below can be found in the above-mentioned limitations on the federated forgetting learning method, and will not be repeated here.
[0067] like Figure 3 As shown, an embodiment of the present invention provides a device for analyzing the correlation of multi-source time series data of power equipment, comprising:
[0068] An operation timing data acquisition module is used to collect operation timing data of power equipment;
[0069] A time lag correction module, configured to perform time lag correction on the runtime series data;
[0070] The adjacency matrix generation module is used to estimate the connection probability between the corrected runtime series data using Gumbel reparameterized sampling and generate the adjacency matrix of the graph structure;
[0071] The correlation analysis result acquisition module is used to input the adjacency matrix into a graph coding model composed of dilated convolution and graph convolution, and input the obtained output together with the runtime series data into the converter model to obtain the analysis result of the correlation of the runtime series data.
[0072] Example 3:
[0073] The embodiment of the present invention further provides a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements the multi-source time series data correlation analysis method of the power equipment in the aforementioned embodiment.
[0074] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.
[0075] Example 4:
[0076] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the following method:
[0077] Collect operating sequence data of power equipment;
[0078] Performing time lag correction on the runtime series data;
[0079] Use Gumbel reparameterized sampling to estimate the connection probability between the corrected runtime series data and generate the adjacency matrix of the graph structure;
[0080] The adjacency matrix is input into a graph encoding model composed of dilated convolution and graph convolution, and the obtained output and the runtime data are input into a transformer model together to obtain an analysis result of the correlation of the runtime data.
[0081] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, 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 magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0082] 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 flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, 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 flowcharts and / or block diagrams. 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.
[0083] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 1 The function specified in one or more boxes.
[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0085] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all protected by the present invention.
Claims
1. A method for analyzing the correlation of multi-source time series data of power equipment, characterized in that: include: Collect operating sequence data of power equipment; Performing time lag correction on the runtime series data; Use Gumbel reparameterized sampling to estimate the connection probability between the corrected runtime series data and generate the adjacency matrix of the graph structure; The adjacency matrix is input into a graph encoding model composed of dilated convolution and graph convolution, and the obtained output and the runtime data are input into a transformer model together to obtain an analysis result of the correlation of the runtime data.
2. The method for analyzing the correlation of multi-source time series data of power equipment according to claim 1, characterized in that: The operating time series data includes: time series data of voltage, current, power, temperature, transformer oil temperature, partial discharge, vibration and dissolved gas content in oil; the operating time series data is collected from the running electrical equipment through multiple sensors.
3. The method for analyzing the correlation of multi-source time series data of electric power equipment according to claim 2, characterized in that: Performing time lag correction on the running series data includes: setting time lag parameters for each running series data according to the physical characteristics and historical data characteristics of each running series data , to affect the time series data with delay Corrected to .
4. The method for analyzing the correlation of multi-source time series data of electric power equipment according to claim 3, characterized in that: The time lag parameter The determination method includes: setting the time lag parameter As a hyperparameter of the transformer model, the time lag parameter is determined by a hyperparameter search method during the training process. value.
5. The method for analyzing the correlation of multi-source time series data of electric power equipment according to claim 1, characterized in that: Before performing time lag correction on the runtime data, the method further includes: preprocessing the collected runtime data and normalizing the preprocessed data; The preprocessing includes: data cleaning, missing value filling and outlier detection.
6. The method for analyzing the correlation of multi-source time series data of electric power equipment according to claim 1, characterized in that: Gumbel reparameterized sampling is used to estimate the connection probability between the runtime series data, and an adjacency matrix of a graph structure is generated, including: The runtime series data consists of M time series. Gumbel reparameterized sampling is used to generate the correlation dependency probability of any two time series in the runtime series data. The generated M*M correlation dependency probabilities constitute an M×M graph structure adjacency matrix.
7. The method for analyzing the correlation of multi-source time series data of electric power equipment according to claim 1, characterized in that: The transformer model includes a multi-branch attention mechanism; The multi-branch attention mechanism includes global learning attention and local neighborhood attention, which is expressed as: ， in, is the local neighborhood attention, Learning attention for global context; Represents a dimension concatenation operation.
8. A device for analyzing the correlation of multi-source time series data of power equipment, characterized in that: include: An operation timing data acquisition module is used to collect operation timing data of power equipment; A time lag correction module, configured to perform time lag correction on the runtime series data; The adjacency matrix generation module is used to estimate the connection probability between the corrected runtime series data using Gumbel reparameterized sampling and generate the adjacency matrix of the graph structure; The correlation analysis result acquisition module is used to input the adjacency matrix into a graph coding model composed of dilated convolution and graph convolution, and input the obtained output together with the runtime series data into the converter model to obtain the analysis result of the correlation of the runtime series data.
9. A computer device, characterized in that: include: memory for storing computer programs; A processor is used to execute the computer program to implement the steps of the method for correlation analysis of multi-source time series data of electric power equipment according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for correlation analysis of multi-source time series data of electric power equipment according to any one of claims 1 to 7 are implemented.