A method for detecting power equipment status based on multi-source data fusion

Abstract

The present invention relates to a method for detecting the state of electric power equipment based on multi-source data fusion, comprising the steps of acquiring data, eliminating invalid data, primary fusion and secondary fusion. The present invention adopts multiple data fusion methods to comprehensively analyze the state of electric power equipment, and the obtained state of electric power equipment has a high accuracy rate and a high application value, and can effectively improve the intelligent level of electric power inspection work.

Description

Technical Field

[0001] The invention relates to a method for detecting the state of electric equipment based on multi-source data fusion, and belongs to the technical field of electric equipment. Background Art

[0002] With the continuous expansion of the scale of my country's power system development and the continuous development of new technologies such as communication technology, Internet technology, and sensors, human society is stepping into the era of smart grids. A large number of sensors and computer analysis generate a large amount of data in different forms and structures. These data reflect the operating status and operating environment of smart grids from all aspects. How to accurately use this resource has become a global topic. The short-term goal of the development of modern power system technology is to establish a real-time monitoring system that includes multiple links of the power system to solve the problems of single data acquisition channels, narrow application of monitoring data, and incomplete analysis in the past.

[0003] As the basic component of the power grid, power equipment is the basis for the safe, reliable and stable operation of the power grid. Among them, the status monitoring data of power equipment is an important basis for analyzing and predicting the development of equipment status. With the development of status monitoring technology and the interaction of various automation information application systems in the power grid, the amount of power equipment status data presents big data, which has the characteristics of large scale, fast update speed, multi-source heterogeneity and low value density. Therefore, it has the applicability and feasibility of using data fusion technology for analysis and processing. In the face of the diverse sources of power equipment data, it is necessary to fuse the data of power equipment to improve the accuracy of the basis for judging the status of power equipment and provide the most powerful decision support for the intelligent analysis system. At present, the research on multi-source data fusion of power equipment is mostly aimed at fault location of power equipment and identification and verification of multi-source data of distribution network. How to use effective data fusion technology to combine the results of video surveillance intelligent analysis with control data and sensor data to comprehensively analyze the status of power equipment has become one of the problems that the current power system needs to solve. Summary of the invention

[0004] The present invention proposes a method for detecting the state of electric power equipment based on multi-source data fusion, which solves the problems of single data acquisition channel, narrow application of monitoring data and incomplete analysis.

[0005] In order to solve the above technical problems, the technical solution proposed by the present invention is: a method for detecting the state of power equipment based on multi-source data fusion, comprising the following steps:

[0006] Step S1, obtaining telemetry data, telesignaling data and televiewing data of power equipment;

[0007] Step S2, using a clustering method to remove abnormal values ​​in the telemetry data, telesignaling data, and remote viewing data, reduce the total noise level, and remove invalid abnormal data without data fusion;

[0008] Step S3, first-level fusion: Adopting adaptive weighted average method to fuse telemetry data and telesignaling data, adaptively finding corresponding weights according to minimum mean square error;

[0009] Step S4, using a convolutional neural network intelligent algorithm to extract the features of the remote viewing data, and training a model, using the model to analyze the remote viewing data to achieve remote viewing image recognition;

[0010] Step S5, using the covariance crossover algorithm to perform secondary fusion on the results of step S3 and step S4.

[0011] The above solution is further improved in that: the step S2 specifically includes the following process:

[0012] Use the clustering method to construct the time series at time k, and then obtain the predicted value at time k+1 and the residual between the predicted value and the observed value; judge whether the residual at time k+1 is abnormal according to the Dixon principle; if it is abnormal, remove the current acquired value, otherwise keep it.

[0013] The above solution is further improved in that: step S3 specifically includes the following process:

[0014] Adopting the adaptive weighted average method, the data processed in step S2 is fused, and the fused data satisfies the formula:

[0015] The mean square error is:

[0016] Where X is the true value to be estimated, X i ,X j is the telemetry data and telesignal data processed in step 2, w i is the corresponding weight, and the weight corresponding to each data is adaptively found according to the principle of minimum mean square error

[0017] The above solution is further improved in that: the step S5 specifically includes the following process:

[0018] S51, let the result obtained in step S3 be a, let the result obtained in step S4 be b, then the autocovariance matrix and cross-covariance matrix of a and b are respectively: in

[0019] S52, the fusion method is based on c=W a a+W bb is calculated, where c is the estimated value after fusion, W a , W b is a linear weighted matrix, W a =P bb (P aa +P bb ) -1 , W b =P aa (P aa +P bb ) -1 , the covariance calculation formula of c is: Since a and b are uncorrelated, p ab =0, then the fusion estimate is: c = a + p aa (p aa +p bb ) -1 (ba);

[0020] S53, p aa (p aa +p bb ) -1 (ba) determines the optimality of the fusion estimate, through p cc The minimum determinant criterion is obtained by optimization method.

[0021] The above scheme is further improved in that: in step S1, the telemetry data of the power equipment is obtained from the equipment production management system, the remote signaling data of the power equipment is obtained from the dispatching management system, and the remote viewing data of the power equipment is obtained from the unified video monitoring platform combined with the intelligent analysis module.

[0022] The power equipment status detection method based on multi-source data fusion provided by the present invention integrates the telemetry information of the power equipment production management system, the telesignaling information of the dispatching management system, the remote viewing information of the unified video monitoring platform and the artificial intelligence analysis module, comprehensively analyzes and obtains the current status of the relevant power equipment, and provides a powerful triple judgment basis for the inspection personnel to accurately judge and analyze the current equipment status. The present invention adopts multiple data fusion methods to comprehensively analyze the equipment status, and the obtained equipment status has a high accuracy rate and high application value, which can effectively improve the intelligent level of power inspection work. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described below in conjunction with the accompanying drawings.

[0024] Figure 1 It is a schematic diagram of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0025] Example

[0026] The present embodiment provides a method for detecting the state of an electric power device based on multi-source data fusion, such as Figure 1 As shown, the following steps are included:

[0027] Step S1, obtaining the telemetry data of the equipment from the equipment production management system, obtaining the remote signaling data of the equipment from the dispatch management system, and obtaining the remote viewing data of the equipment from the unified video monitoring platform combined with the intelligent analysis module;

[0028] Step S2: invalid and abnormal data will affect the final fusion result. Therefore, the collected data needs to be analyzed for validity before fusion. The abnormal values ​​in the collected signal are eliminated by clustering method to reduce the total noise level. The invalid abnormal data is eliminated and no data fusion is performed. Data preprocessing: construct the time series at time k by clustering method, and then obtain the predicted value at time k+1 and the residual between the predicted value and the observed value. Determine whether the residual at time k+1 is abnormal according to Dixon's principle. If it is abnormal, remove the current acquired value, and if it is normal, retain it. Specifically, in this embodiment, the observed value is also the actual data.

[0029] Step S3, first-level fusion: Adopt the adaptive weighted average method to fuse the telemetry data and telesignaling data, and adaptively find the corresponding weights according to the minimum mean square error; the specific process is:

[0030] Adopt adaptive weighted average method to fuse the preprocessed data. Let x i is the preprocessed data, w i is the corresponding weight, then the fused data satisfies the formula:

[0031] The mean square error is:

[0032] Where X is the true value to be estimated, X i ,X j is the telemetry data and telesignal data processed in step 2, w i is the corresponding weight, and the weight corresponding to each data is adaptively found according to the principle of minimum mean square error

[0033] Step S4, using a convolutional neural network intelligent algorithm to extract the features of the remote viewing data, and training a model, using the model to analyze the remote viewing data to achieve remote viewing image recognition;

[0034] Step S5, using the covariance crossover algorithm to perform secondary fusion on the results of step S3 and step S4; the specific process is:

[0035] S51, let the result obtained in step S3 be a, let the result obtained in step S4 be b, then the autocovariance matrix and cross-covariance matrix of a and b are respectively: in

[0036] S52, the fusion method is based on c=W a a+W b b is calculated, where c is the estimated value after fusion, W a , W b is a linear weighted matrix, W a =P bb (P aa +P bb ) -1 , W b =P aa (P aa +P bb ) -1 , the covariance calculation formula of c is: Since a and b are unrelated here, p ab =0, then the fusion estimator can be recorded as: c = a + p aa (p aa +p bb ) -1 (ba);

[0037] S53, p aa (p aa +p bb ) -1 (ba) determines the optimality of the fusion estimate, through p cc The minimum determinant criterion is obtained by optimization method.

[0038] Further, data classification refers to the equipment classification in the following table:

[0039] Table 1 Equipment data classification

[0040]

[0041] The present invention is not limited to the above embodiments, and any technical solutions formed by equivalent replacement shall fall within the protection scope required by the present invention.

Claims

1. A method for detecting the state of electric power equipment based on multi-source data fusion, characterized in that: The steps include: Step S1, obtaining telemetry data, telesignaling data and televiewing data of power equipment; Step S2, removing abnormal values ​​in the telemetry data, telesignaling data and remote viewing data by clustering method to reduce the total noise level; Step S3, first-level fusion: Adopting adaptive weighted average method to fuse telemetry data and telesignaling data, adaptively finding corresponding weights according to minimum mean square error; Step S4, using a convolutional neural network algorithm to extract features of the remote viewing data, and training a model, using the model to analyze the remote viewing data to achieve remote viewing image recognition; Step S5, using a covariance crossover algorithm to perform secondary fusion on the results of step S3 and step S4; S51, let the result obtained in step S3 be a, let the result obtained in step S4 be b, then the autocovariance matrix and cross-covariance matrix of a and b are respectively: in ; S52, the fusion method is based on c=W a a+W b b is calculated, where c is the estimated value after fusion, W a , W b is a linear weighted matrix, W a =P bb (P aa +P bb ) -1 , W b =P aa (P aa +P bb ) -1 , the covariance calculation formula of c is: Since a and b are uncorrelated, p ab =0, then the fusion estimate is: c = a + p aa (p aa +p bb ) -1 (ba); S53, p aa (p aa +p bb ) -1 (ba) determines the optimality of the fusion estimate, and the c value is obtained through the optimization method through the minimum determinant criterion of PCC.

2. The method for detecting the state of electric power equipment based on multi-source data fusion according to claim 1 is characterized in that: The step S2 specifically includes the following process: Use clustering method to construct time series at time k, then obtain the predicted value at time k+1 and the residual between the predicted value and the observed value; judge whether the residual at time k+1 is abnormal according to Dixon principle; If it is abnormal, the current value is removed, otherwise it is retained.

3. The method for detecting the state of electric power equipment based on multi-source data fusion according to claim 1 is characterized in that: The step S3 specifically includes the following process: Adopting the adaptive weighted average method, the data processed in step S2 is fused, and the fused data satisfies the formula: The mean square error is: Where X is the true value to be estimated, X i , X j are the telemetry data and telesignaling data processed in step S2, respectively. i 、w j They are the weights of the corresponding telemetry data and telesignaling data respectively. According to the principle of minimum mean square error, the weights corresponding to each data are adaptively found.

4. The method for detecting the state of electric power equipment based on multi-source data fusion according to claim 1 is characterized in that: In the step S1, the telemetry data of the power equipment is obtained from the equipment production management system, the remote signaling data of the power equipment is obtained from the dispatching management system, and the remote viewing data of the power equipment is obtained from the unified video monitoring platform combined with the intelligent analysis module.

Images