Multi-source measurement data fusion method for operation and maintenance management of distribution network equipment

Abstract

The invention relates to a multi-source measurement data fusion method for operation and maintenance management of distribution network equipment, which comprises the following steps of: firstly, establishing a layered data acquisition framework according to operation and maintenance requirements of the distribution network equipment, and establishing a unified data transmission protocol and a standardized data management specification so as to complete construction of a multi-source measurement data acquisition system; the method comprises the following steps of: acquiring equipment operation electrical quantity data, environmental perception data, mechanical state quantity data and historical operation and maintenance record data by utilizing a constructed multi-source measurement data acquisition system, then preprocessing the data, extracting key features capable of reflecting the equipment state in the data, and finally, performing hierarchical fusion to obtain a multi-source measurement result. According to the method, weighted preliminary fusion is firstly carried out on data of the same type, and then deep fusion of cross-latitude data is carried out, so that spatial association and time sequence features of the data are mined, and the problems of non-uniform formats of multi-source data and information redundancy conflicts are effectively solved; the method has the advantages of high adaptability and high fusion precision.

Description

Technical Field

[0001] This invention belongs to the field of distribution network equipment operation and maintenance management technology, and specifically relates to a multi-source measurement data fusion method for distribution network equipment operation and maintenance management. Background Technology

[0002] Currently, with the continuous improvement of the automation level of smart grids and distribution networks, a wide variety and a large number of measurement devices have been deployed in distribution networks. These devices generate massive amounts of heterogeneous, multi-timescale operational data, providing a data foundation for the refined operation and maintenance management of distribution network equipment. However, measurement data from different sources are usually collected and managed by independent systems, resulting in inconsistent formats, large differences in accuracy, and a lack of effective correlation and fusion between the data, forming "data silos." This makes it impossible to form a comprehensive and unified understanding of the equipment status. Furthermore, the massive amount of multi-source measurement data, due to the lack of deep fusion and effective utilization, not only restricts the improvement of the intelligent level of distribution network operation and maintenance, but also makes it impossible to effectively predict and identify the hidden health status and operating trends of the equipment. Therefore, to solve the above problems, it is necessary to develop a multi-source measurement data fusion method for distribution network equipment operation and maintenance management that is highly adaptable and has high fusion accuracy. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-source measurement data fusion method for the operation and maintenance management of distribution network equipment with strong adaptability and high fusion accuracy.

[0004] The objective of this invention is achieved as follows: A method for fusing multi-source measurement data for the operation and maintenance management of distribution network equipment, comprising the following steps: Step 1: Based on the operation and maintenance requirements of the distribution network equipment, build a hierarchical data acquisition architecture, and establish a unified data transmission protocol and standardized data management specifications to complete the construction of a multi-source measurement data acquisition system; Step 2: Determine the type of data source for the distribution network equipment measurement data, which includes equipment operation electrical quantity data, environmental sensing data, mechanical status quantity data, and historical operation and maintenance record data; Step 3: Based on the type of data source for the distribution network equipment measurement, and utilizing sensors, smart meters, and fault recording devices deployed inside and around transformers, switchgear, and cable lines in the hierarchical data acquisition architecture, real-time data acquisition is performed; subsequently, the data is uniformly accessed through the multi-source measurement data access module. Step 4: Preprocess the incoming data, which involves data cleaning, data standardization and alignment, data normalization, and data completion operations in sequence. Step 5: Perform feature extraction on the multi-source measurement data after preprocessing in Step 4, that is, extract key features that can reflect the equipment status from various types of data, such as: extracting the current RMS trend, harmonic distortion rate, and voltage sag / boost number from electrical quantity data; extracting the temperature rise rate and temperature difference with ambient temperature from temperature data; and extracting the number of recent faults and fault current amplitude from fault data. Step 6: Establish a hierarchical data fusion model. The first layer uses a weighted average algorithm to perform preliminary fusion of measurement data of the same type and dynamically allocates weights according to the data credibility. The second layer uses a deep learning network to perform deep fusion of different types of preliminary fused data through an improved gated recurrent unit network and outputs a comprehensive fusion result. Step 7: Verify the validity of the fusion results, and based on the verification results, fine-tune the parameters of the hierarchical data fusion model to ensure the accuracy of the fusion results.

[0005] Furthermore, in step 4, data cleaning involves removing outliers and null values; data standardization and alignment involves unifying timestamps and interpolation; data normalization involves eliminating incommensurability between data features due to differences in units and dimensions; and data completion involves using known information to reasonably estimate missing values ​​in order to construct a complete and consistent dataset.

[0006] Furthermore, the deep learning network in step 6 is an improved convolutional neural network combined with a long short-term memory network model. The convolutional neural network extracts spatial features of the data, while the long short-term memory network captures time series features of the data.

[0007] Furthermore, in step 6, the improved gated recurrent unit network can enhance the extraction weight of key data features by introducing an attention mechanism.

[0008] Furthermore, the validity verification in step 7 uses mean squared error and mean absolute error as evaluation indicators. When the evaluation indicators exceed the preset threshold, the weighted weights and network parameters of the deep learning model are re-optimized.

[0009] Furthermore, the electrical quantity data in step 2 includes current, voltage, active power, reactive power, power factor, and three-phase imbalance; environmental sensing data includes ambient temperature, ambient humidity, and meteorological information; mechanical status data includes oil temperature, winding temperature, opening and closing positions, and energy storage status; historical operation and maintenance record data includes operation and maintenance records, historical fault information, and emergency handling plans corresponding to historical fault information.

[0010] The beneficial effects of this invention are as follows: This invention builds a hierarchical data acquisition architecture based on the operation and maintenance needs of distribution network equipment, and establishes a unified data transmission protocol and standardized data management specifications, thereby completing the construction of a multi-source measurement data acquisition system. Subsequently, the constructed multi-source measurement data acquisition system is used to collect electrical quantity data of equipment operation, environmental perception data, mechanical status data, and historical operation and maintenance record data, breaking the information limitations of a single data source and realizing comprehensive coverage of equipment operation-related data, providing a rich data foundation for accurate assessment of equipment status. By extracting key features that reflect equipment status from various types of data and adopting a hierarchical fusion strategy, the system first performs weighted preliminary fusion of data of the same type, and then uses an improved gated cyclic unit network to strengthen the extraction weight of key data features. This allows for the mining of spatial correlation and time series features of the data, effectively solving the problems of inconsistent formats and information redundancy conflicts among multiple data sources. It significantly improves the accuracy and comprehensiveness of the fusion results and reduces the bias in operation and maintenance judgments caused by incomplete data. In summary, this invention has the advantages of strong adaptability and high fusion accuracy. Detailed Implementation

[0011] The present invention will now be further described.

[0012] Example: A method for fusing multi-source measurement data for the operation and maintenance management of distribution network equipment, comprising the following steps: Step 1: Based on the operation and maintenance requirements of the distribution network equipment, build a hierarchical data acquisition architecture, and establish a unified data transmission protocol and standardized data management specifications to complete the construction of a multi-source measurement data acquisition system; Step 2: Determine the type of data source for the distribution network equipment measurement data. This data source includes electrical quantity data, environmental sensing data, mechanical condition data, and historical maintenance records. Electrical quantity data includes current, voltage, active power, reactive power, power factor, and three-phase imbalance. Environmental sensing data includes ambient temperature, humidity, and meteorological information. Mechanical condition data includes oil temperature, winding temperature, switching position, and energy storage status. Historical maintenance records include maintenance operation records, historical fault information, and corresponding emergency response plans. Step 3: Based on the type of data source for the distribution network equipment measurement, and utilizing sensors, smart meters, and fault recording devices deployed inside and around transformers, switchgear, and cable lines in the hierarchical data acquisition architecture, real-time data acquisition is performed; subsequently, the data is uniformly accessed through the multi-source measurement data access module. Step 4: Preprocess the incoming data, which involves data cleaning, data standardization and alignment, data normalization, and data completion. Data cleaning removes outliers and null values; data standardization and alignment unifies timestamps and performs interpolation; data normalization eliminates incommensurability caused by differences in units and dimensions between data features; and data completion uses known information to reasonably estimate missing values ​​to build a complete and consistent dataset. Step 5: Perform feature extraction on the multi-source measurement data after preprocessing in Step 4, that is, extract key features that can reflect the equipment status from various types of data, such as: extracting the current RMS trend, harmonic distortion rate, and voltage sag / boost number from electrical quantity data; extracting the temperature rise rate and temperature difference with ambient temperature from temperature data; and extracting the number of recent faults and fault current amplitude from fault data. Step 6: Establish a hierarchical data fusion model. The first layer uses a weighted average algorithm to initially fuse similar measurement data and dynamically assigns weights based on data credibility. The second layer first fuses different types of data using a deep learning network, which is an improved convolutional neural network combined with a long short-term memory network. The convolutional neural network is used to extract spatial features of the data, and the long short-term memory network is used to capture time series features of the data. Then, a deep fusion of cross-dimensional data is performed through an improved gated recurrent unit network, and a comprehensive fusion result is output. In this process, the improved gated recurrent unit network can strengthen the extraction weight of key data features by introducing an attention mechanism. Step 7: Verify the effectiveness of the fusion results, and based on the verification results, fine-tune the parameters of the hierarchical data fusion model to ensure the accuracy of the fusion results. The effectiveness verification uses mean squared error and mean absolute error as evaluation indicators. When the evaluation indicators exceed the preset threshold, the weighted weights and network parameters of the deep learning model are re-optimized.

[0013] In use, this invention first establishes a hierarchical data acquisition architecture based on the operation and maintenance requirements of distribution network equipment, and establishes a unified data transmission protocol and standardized data management specifications to complete the construction of a multi-source measurement data acquisition system. Then, it determines the types of measurement data sources for the distribution network equipment (including electrical quantity data, environmental sensing data, mechanical condition data, and historical operation and maintenance record data). Next, based on the types of measurement data sources, it utilizes sensors, smart meters, and fault recording devices deployed within and around transformers, switchgear, and cable lines in the hierarchical data acquisition architecture to collect data in real time. This invention, employing this structure, breaks through the information limitations of a single data source and realizes real-time data acquisition of equipment operation... The comprehensive coverage of relevant data provides a rich data foundation for accurately assessing equipment status. After the multi-source measurement data is collected, the data is uniformly accessed through the multi-source measurement data access module. The accessed data undergoes data cleaning, data standardization and alignment, data normalization, and data completion operations in sequence. Data cleaning removes outliers and null values; data standardization and alignment unifies timestamps and interpolation; data normalization eliminates incommensurability caused by differences in units and dimensions between data features; and data completion uses known information to reasonably estimate missing values ​​to construct a complete and consistent dataset. Then, feature extraction is performed on the processed multi-source measurement data (i.e., extracting key features that reflect the equipment status from various types of data). The model extracts various data types, including: current RMS trend, harmonic distortion rate, and voltage sag / boost frequency from electrical quantity data; temperature rise rate and temperature difference with ambient temperature from temperature data; and recent fault frequency and fault current amplitude from fault data. Finally, a hierarchical data fusion model is established. The first layer uses a weighted average algorithm to initially fuse similar measurement data, dynamically allocating weights based on data reliability. The second layer first uses a deep learning network to initially fuse different types of data. This deep learning network is a combination of an improved convolutional neural network and a long short-term memory network, using the convolutional neural network to extract spatial features and the long short-term memory network to capture time-series features. Finally, an improved gated recurrent unit is used to further fuse the data. The network performs deep fusion of cross-dimensional data and outputs a comprehensive fusion result. During this process, an improved gated recurrent unit network can be used to enhance the extraction weight of key data features by introducing an attention mechanism. After completing the above operations, the effectiveness of the fusion result is verified, and based on the verification result, the parameters of the hierarchical data fusion model are finely adjusted to ensure the accuracy of the fusion result. This invention adopts this structure, which breaks down data silos by fusing multi-dimensional data such as electrical, physical, and environmental data. It effectively solves the problems of inconsistent formats and information redundancy and conflict among multiple data sources, forming a panoramic perception of the operating status of distribution network equipment and reducing the deviation in operation and maintenance judgment caused by incomplete data. In summary, this invention has the advantages of strong adaptability and high fusion accuracy.

[0014] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for fusing multi-source measurement data for the operation and maintenance management of distribution network equipment, characterized in that, Includes the following steps: Step 1: Based on the operation and maintenance requirements of the distribution network equipment, build a hierarchical data acquisition architecture, and establish a unified data transmission protocol and standardized data management specifications to complete the construction of a multi-source measurement data acquisition system; Step 2: Determine the type of data source for the distribution network equipment measurement data, which includes equipment operation electrical quantity data, environmental sensing data, mechanical status quantity data, and historical operation and maintenance record data; Step 3: Based on the type of data source for the distribution network equipment measurement, and utilizing sensors, smart meters, and fault recording devices deployed inside and around transformers, switchgear, and cable lines in the hierarchical data acquisition architecture, real-time data acquisition is performed; subsequently, the data is uniformly accessed through the multi-source measurement data access module. Step 4: Preprocess the incoming data, which involves data cleaning, data standardization and alignment, data normalization, and data completion operations in sequence. Step 5: Perform feature extraction on the multi-source measurement data after preprocessing in Step 4, that is, extract key features that can reflect the equipment status from various types of data, such as: extracting the current RMS trend, harmonic distortion rate, and voltage sag / boost number from electrical quantity data; extracting the temperature rise rate and temperature difference with ambient temperature from temperature data; and extracting the number of recent faults and fault current amplitude from fault data. Step 6: Establish a hierarchical data fusion model. The first layer uses a weighted average algorithm to perform preliminary fusion of measurement data of the same type and dynamically allocates weights according to the data credibility. The second layer uses a deep learning network to perform deep fusion of different types of preliminary fused data through an improved gated recurrent unit network and outputs a comprehensive fusion result. Step 7: Verify the validity of the fusion results, and based on the verification results, fine-tune the parameters of the hierarchical data fusion model to ensure the accuracy of the fusion results.

2. The method for multi-source measurement data fusion for operation and maintenance management of distribution network equipment as described in claim 1, characterized in that: The data cleaning in step 4 is to remove outliers and null values; data standardization and alignment is to unify timestamps and interpolation processing; data normalization is to eliminate the incommensurability between data features caused by different units and sizes; and data completion is to use known information to reasonably estimate missing values ​​in order to build a complete and consistent dataset.

3. The method for multi-source measurement data fusion for operation and maintenance management of distribution network equipment as described in claim 1, characterized in that: The deep learning network in step 6 is an improved model combining a convolutional neural network and a long short-term memory network. The convolutional neural network extracts spatial features of the data, while the long short-term memory network captures time-series features of the data.

4. The multi-source measurement data fusion method for distribution network equipment operation and maintenance management as described in claim 1, characterized in that: In step 6, the improved gated recurrent unit network can enhance the extraction weight of key data features by introducing an attention mechanism.

5. The method for multi-source measurement data fusion for operation and maintenance management of distribution network equipment as described in claim 1, characterized in that: The validity verification in step 7 uses mean squared error and mean absolute error as evaluation indicators. When the evaluation indicators exceed the preset threshold, the weighted weights and network parameters of the deep learning model are re-optimized.

6. The method for multi-source measurement data fusion for operation and maintenance management of distribution network equipment as described in claim 1, characterized in that: The electrical quantity data in step 2 includes current, voltage, active power, reactive power, power factor, and three-phase imbalance; the environmental sensing data includes ambient temperature, ambient humidity, and meteorological information; the mechanical status data includes oil temperature, winding temperature, opening and closing positions, and energy storage status; and the historical operation and maintenance record data includes operation and maintenance records, historical fault information, and emergency handling plans corresponding to historical fault information.