Method, device and storage medium based on fusion analysis of multi-source heterogeneous data of electric power
Through deep neural networks and fusion networks, the multi-source heterogeneous data of transmission lines are extracted and fused, and fault detection errors and large-scale data processing challenges caused by power data independence are solved, and more accurate fault information acquisition and classification are achieved. .
Patent Information
- Application Number
- CN202310153519.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-02-23
- Publication Date
- 2024-12-13
- Estimated Expiration
- 2043-02-23
AI Technical Summary
In power transmission lines, the source of power data is complex and independent, which leads to the inability to effectively integrate and comprehensive judgment during fault detection, which easily generates judgment errors and large-scale data processing faces challenges.
One-dimensional features are extracted by deep neural network constructed based on stacked autoencoder, multi-dimensional features are extracted by convolutional neural network, and the features are alternately optimized and fused by stacked autoencoder construction, and fault information classification is used by Softmax classifier. , obtain and classify real-time and hidden fault information.
Through data preprocessing and secondary processing, redundant data transmission is reduced, the accuracy of fault information is improved, the error judgment rate is reduced, and the operation status of the transmission line is more accurately obtained.
Smart Images
Figure CN116108402B_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of power data analysis, and specifically relates to a method, device and storage medium based on fusion analysis of multi-source heterogeneous power data. Background Art
[0002] In simple terms, multi-source heterogeneity means that a whole is composed of components from multiple different sources, including both mixed data (including structured and unstructured) and discrete data (data distributed in different systems or platforms). The Internet is a typical heterogeneous network, and the fusion communication matrix is a typical multi-source heterogeneous data network.
[0003] In transmission lines, circuit data information is very complex and comes from a wide range of sources. Fault detection and other tasks rely on this complex and extensive data information. However, these power data information are all in relatively independent systems. During fault detection, judgments can only be made in independent systems and cannot be effectively integrated for comprehensive judgment, which can easily lead to judgment errors. In addition, as the data type and data scale become larger and larger, the density of valuable data in large-scale data is very low. The huge amount of data has posed a challenge to existing data processing methods. Summary of the invention
[0004] The purpose of this application is to provide a method, device and storage medium based on power multi-source heterogeneous data fusion analysis to solve the problems faced by the above-mentioned background technology.
[0005] The purpose of this application can be achieved through the following technical solutions:
[0006] A method for fusion analysis of multi-source heterogeneous data of electric power, comprising the following steps:
[0007] Step 1: Obtain sample data related to power in the transmission line;
[0008] Step 2: Preprocess and reprocess the collected sample data;
[0009] Step 3: Use a deep neural network constructed by stacked autoencoders to extract one-dimensional features of the transmission line, use a convolutional neural network to extract multi-dimensional features of the transmission line, and then use a stacked autoencoder to construct a fusion network to alternately optimize and fuse the one-dimensional features and multi-dimensional features to obtain real-time fault information and hidden fault information;
[0010] Step 4: Use the Softmax classifier to classify fault information.
[0011] As a further solution of the present application, the sample data in step 1 includes structured data and unstructured data related to electricity in the transmission line.
[0012] As a further solution of the present application, the preprocessing in step 2 includes parsing the acquired structured data to obtain corresponding digital information, and dehazing the acquired unstructured data and converting it into corresponding digital information.
[0013] As a further solution of the present application, the secondary processing step in step 2 includes:
[0014] According to the length of the transmission line, several sections are set up at equal distances, and each section has a corresponding collection node;
[0015] All real-time monitoring data collected within the interval are sent to the aggregation node, and processed through the Kalman filter algorithm to obtain the fusion output data of each monitoring data;
[0016] By formula Get the average fusion output data of each monitoring data; where A i is the fusion output data of the i-th interval, α i is the weighting coefficient, i=1, 2, 3…, n.
[0017] As a further solution of the present application, the method for obtaining the real-time fault information in step 3 is:
[0018] Through the formula B=θ0+θ1x1+θ2x2+…+θ n x n The real-time fault parameter B of the transmission line is obtained, where x n is the average fusion output data of the nth measured state quantity, θ n is the compensation coefficient of each measured state quantity.
[0019] The obtained real-time fault parameter B is compared with the preset fault parameter B q Make comparisons;
[0020] If B≥B q , then the data is judged to be abnormal and real-time fault information is output;
[0021] Otherwise, the data is judged to be normal.
[0022] As a further solution of the present application, the method for obtaining the hidden fault information in step 3 is:
[0023] When the data is normal, obtain the curves of each measured state quantity changing with time within h seconds;
[0024] Compare the time-varying curves of each measured state quantity with the respective preset threshold curves;
[0025] The area enclosed by the bottom of each measured state quantity change curve and the top of each threshold curve is recorded as B n
[0026] By formula Calculate the hidden fault parameter D, where B n is the area enclosed by the bottom of the nth measured state quantity change curve and the top of its own threshold curve, η n is the area compensation coefficient of each measured state quantity;
[0027] The obtained hidden fault parameter D is compared with the preset standard fault parameter threshold D q Make comparisons;
[0028] If D ≥ D q , then the data is judged to be abnormal and real-time fault information is output;
[0029] Otherwise, the data is judged to be normal.
[0030] As a further solution of the present application, the fault classification in step 4 includes internal fault causes and external fault causes, and both the internal fault causes and the external fault causes are provided with fault danger level classifications.
[0031] An electronic device comprises a memory and a processor, wherein a computer program is stored in the memory, and wherein when the computer program is executed by the processor, the processor executes the steps of the described method.
[0032] A computer-readable storage medium stores a computer program, which implements the steps of the method when executed by a processor.
[0033] Beneficial effects of this application:
[0034] This application will pre-process and re-process the collected data to remove redundant data and reduce the transmission of redundant data, so as to reduce the pressure on the system for data processing; at the same time, through the fusion processing of one-dimensional features and multi-dimensional features, analysis can be performed to obtain real-time fault information and hidden fault information, so as to more accurately obtain the operating status of the transmission line, thereby greatly reducing the misjudgment of transmission line faults.
[0035] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0037] Figure 1 This is a flowchart of the application process. DETAILED DESCRIPTION
[0038] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0039] Methods based on fusion analysis of multi-source heterogeneous data of power, such as Figure 1 As shown, the method comprises the following steps:
[0040] Step 1: Obtain sample data related to power in the transmission line;
[0041] Step 2: Preprocess and reprocess the collected sample data;
[0042] Step 3: Use a deep neural network constructed by stacked autoencoders to extract one-dimensional features of the transmission line, use a convolutional neural network to extract multi-dimensional features of the transmission line, and then use a stacked autoencoder to construct a fusion network to alternately optimize and fuse the one-dimensional features and multi-dimensional features to obtain real-time fault information and hidden fault information;
[0043] Step 4: Use the Softmax classifier to classify fault information.
[0044] Through the above technical solution, the present application will pre-process and re-process the collected relevant data, remove redundant data and reduce the transmission of redundant data, so as to reduce the pressure of the system on data processing. After the fusion processing of one-dimensional features and multi-dimensional features, real-time fault information and hidden fault information can be obtained by analysis, so as to greatly reduce the misjudgment of transmission lines.
[0045] As a real-time method of the present application, the sample data in step 1 includes structured data and unstructured data related to electricity in the transmission line.
[0046] Through the above technical solution, the structured data mainly includes basic information data of power equipment, power consumption information data, historical fault information data, and various information data of real-time monitoring on transmission lines, while the unstructured data mainly includes image information data of transmission lines collected by cameras.
[0047] As a real-time method of the present application, the preprocessing in step 2 includes parsing the acquired structured data to obtain corresponding digital information, and dehazing the acquired unstructured data and converting it into corresponding digital information.
[0048] Through the above technical solution, since most of the structured data is stored in the database, in order to facilitate query and use, it is uniformly parsed and converted into digital information and stored in the data table. The image information collected by the camera may be blurred, and a clear image is obtained after dehazing processing, and the image is converted into digital information, which facilitates data processing.
[0049] In the above technical solution, the analysis of structured data and the conversion of unstructured data are both implemented by existing technologies and will not be elaborated here.
[0050] As a real-time method of the present application, the secondary processing step in step 2 includes:
[0051] According to the length of the transmission line, several sections are set up at equal distances, and each section has a corresponding collection node;
[0052] All real-time monitoring data collected within the interval are sent to the aggregation node, and processed through the Kalman filter algorithm to obtain the fusion output data of each monitoring data;
[0053] By formula Get the average fusion output data of each monitoring data; where A i is the fusion output data of the i-th interval, α i is the weighting coefficient, i=1, 2, 3…, n.
[0054] Through the above technical solution, in order to better reflect the real state of the transmission line at a certain moment, multiple sensor nodes of the same type will be used to monitor the same target state in real time. For example, the temperature of the conductor, the tension of the conductor, the leakage current of the insulator and other signals are all monitored by multiple sensor nodes; similarly, the photos of the tower and the inclination of the guy wire collected by the camera will also have a large amount of repeated image data, which will cause the data to contain a large amount of redundant data. The transmission of a large amount of redundant data in multi-source data will seriously waste the bandwidth resources of the communication network and may even cause network congestion. In addition, the monitoring data that is interfered by noise is transmitted to the control center, which will also cause the judgment of the state of the transmission line. Therefore, the Kalman filter algorithm is used to process the collected monitoring data to obtain a target fusion output data for each interval. Since there may be differences in each interval, the formula is used. An average fusion output data of each monitoring data is obtained, thereby reducing the transmission of redundant data.
[0055] As a real-time method of the present application, the method for obtaining real-time fault information in step 3 is:
[0056] Through the formula B=θ0+θ1x1+θ2x2+…+θ n x n The real-time fault parameter B of the transmission line is obtained, where x n is the average fusion output data of the nth measured state quantity, θ n is the compensation coefficient of each measured state quantity.
[0057] The obtained real-time fault parameter B is compared with the preset fault parameter B q Make comparisons;
[0058] If B≥B q , then the data is judged to be abnormal and real-time fault information is output;
[0059] Otherwise, the data is judged to be normal.
[0060] Through the above technical solution, the processed data uses a deep neural network constructed by stacked autoencoders to extract the one-dimensional features of the transmission line, and the multi-dimensional features of the transmission line are extracted by a convolutional neural network. Then, a fusion network is constructed by stacked autoencoders to alternately optimize and fuse the one-dimensional features and the multi-dimensional features. When the existing multi-source heterogeneous data features are fused, multiple independently collected data features are directly spliced to judge the fault of the transmission line. However, this direct method ignores the correlation and complementarity between multi-source heterogeneous data, which can easily cause errors in judgment or failure to detect. To this end, a fusion network is constructed by stacked autoencoders to alternately optimize and fuse the one-dimensional features and the multi-dimensional features, so that real-time fault information and hidden fault information can be obtained, and the fault of the transmission line can be found more clearly and accurately to reduce the occurrence of misjudgment. Through the formula B=θ0+θ1x1+θ2x2+…+θ n x n The real-time fault parameter B of the transmission line is obtained, where x n is the average fusion output data of the nth measured state quantity, which can be the measured tower inclination, conductor temperature, conductor tension, etc., θ n is the compensation coefficient of each measured state quantity. These measured state quantities are comprehensively analyzed to obtain a real-time fault parameter B. Then the real-time fault parameter B is compared with the preset fault parameter B. q By comparing the data, it can be more accurately determined whether the transmission line is normal.
[0061] As a real-time method of the present application, the method for obtaining the hidden fault information in step 3 is:
[0062] When the data is normal, obtain the curves of each measured state quantity changing with time within h seconds;
[0063] Compare the time-varying curves of each measured state quantity with the respective preset threshold curves;
[0064] The area enclosed by the bottom of each measured state quantity change curve and the top of each threshold curve is recorded as B n
[0065] By formula Calculate the hidden fault parameter D, where B n is the area enclosed by the bottom of the nth measured state quantity change curve and the top of its own threshold curve, η n is the area compensation coefficient of each measured state quantity;
[0066] The obtained hidden fault parameter D is compared with the preset standard fault parameter threshold D q Make comparisons;
[0067] If D ≥ Dq , then the data is judged to be abnormal and real-time fault information is output;
[0068] Otherwise, the data is judged to be normal.
[0069] Through the above technical scheme, this scheme analyzes the hidden faults of the transmission line when all data are normal, and establishes a curve of each measured state quantity detected over time, such as the conductor temperature change, the conductor tension change over time curve, and then compares it with various preset standard curves. For example, if the conductor temperature curve is above the temperature standard curve for a long time, there may be aging of the conductor or rust of the equipment. If the conductor tension change curve is above the tension standard curve for a long time, there may be corrosion and aging of the conductor. Through the formula The hidden fault parameter D is calculated comprehensively. It can be seen that the larger the enclosed area, the more serious the damage is and the greater the possibility of failure. The obtained hidden fault parameter D is then compared with the preset standard fault parameter threshold D q By making comparisons, hidden faults can be identified, and the operating status of the transmission line can be more clearly understood to discover hidden faults.
[0070] As a real-time method of the present application, the fault classification in step 4 includes internal fault causes and external fault causes, and fault danger level classification is provided in both the internal fault causes and the external fault causes.
[0071] Through the above technical solution, since the failure may be caused by external or internal factors, such as problems with the equipment itself, or it may be caused by external factors such as weather, by classifying the failures, it can be more clear for the staff to deal with them. The fault danger level classification is also set up, which can make it easier for the staff to understand the danger level of the failure, facilitate handling, and better protect the personal safety of the staff.
[0072] An electronic device includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes various steps of the above-mentioned analysis method.
[0073] A computer-readable storage medium stores a computer program, which implements the steps of the above-mentioned analysis method when executed by a processor.
[0074] The above contents are merely examples and explanations of the concept of the present application. The technicians in this technical field may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they should all fall within the scope of protection of the present application.
Claims
1. A method based on fusion analysis of multi-source heterogeneous data of electric power, characterized in that: The method comprises the following steps: Step 1: Obtain sample data related to electricity in the transmission line, where the sample data includes structured data and unstructured data related to electricity in the transmission line; Step 2: Preprocess and reprocess the collected sample data; Step 3: Use a deep neural network constructed by stacked autoencoders to extract one-dimensional features of the transmission line, use a convolutional neural network to extract multi-dimensional features of the transmission line, and then use a stacked autoencoder to construct a fusion network to alternately optimize and fuse the one-dimensional features and multi-dimensional features to obtain real-time fault information and hidden fault information; Step 4: Use Softmax classifier to classify fault information; The preprocessing in step 2 includes parsing the acquired structured data to obtain corresponding digital information, and defogging the acquired unstructured data to convert them into corresponding digital information; The secondary processing step in step 2 includes: according to the length of the transmission line, a number of intervals are set up at equal intervals, and a collection node is set in each interval; All real-time monitoring data collected within the interval are sent to the aggregation node, and processed through the Kalman filter algorithm to obtain the fusion output data of each monitoring data; By formula Get the average fusion output data of each monitoring data; where A i is the fusion output data of the i-th interval, α i is the weighting coefficient, i = 1, 2, 3…, n; The method for obtaining real-time fault information in step 3 is: through the formula B = θ0 + θ1x1 + θ2x2 + ... + θ n x n The real-time fault parameter B of the transmission line is obtained, where x n is the average fusion output data of the nth measured state quantity, θ n is the compensation coefficient of each measured state quantity; The obtained real-time fault parameter B is compared with the preset fault parameter B q Make comparisons; If B≥B q , then the data is judged to be abnormal and real-time fault information is output; Otherwise, the data is judged to be normal; The method for obtaining the hidden fault information in step 3 is: when the data is normal, within h seconds, respectively obtain the curves of each measured state quantity changing with time; Compare the time-varying curves of each measured state quantity with the respective preset threshold curves; The area enclosed by the bottom of each measured state quantity change curve and the top of each threshold curve is recorded as B n ; By formula Calculate the hidden fault parameter D, where B n is the area enclosed by the bottom of the nth measured state quantity change curve and the top of its own threshold curve, η n is the area compensation coefficient of each measured state quantity; The obtained hidden fault parameter D is compared with the preset standard fault parameter threshold D q Make comparisons; If D ≥ D q , then the data is judged to be abnormal and real-time fault information is output; Otherwise, the data is judged to be normal.
2. The method based on fusion analysis of multi-source heterogeneous data of electric power according to claim 1 is characterized in that: The fault classification in step 4 includes internal fault causes and external fault causes, and the internal fault causes and external fault causes are both provided with fault danger level classifications.
3. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 2.
4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.