A method for fusion of power equipment status data based on big data

Abstract

A big data-based power equipment status data fusion method first constructs a power equipment information model, then extracts and integrates the collected power equipment status data based on the power equipment information model. Simulation data for each logical node type is calculated using an RPN convolutional neural network. The simulated data is then compared with the corresponding standard data to screen out logical node types that share the same simulated and standard data. Feature extraction is then performed on the filtered logical node type instance data. Finally, spatial, temporal, and redundant data association is performed, and the associated data is fused into indicators for predictive analysis. This design effectively improves the efficiency of power equipment data fusion.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a method for fusion of power equipment status data based on big data. Background Art

[0002] As the energy supply system upon which economic development and human life depend, the power system is geographically distributed and has a vast base of power equipment. This places high demands on the stability and safe operation of these equipment, and any failures can result in significant losses. These characteristics determine that the amount of power data generated during power system operation is massive, rapidly growing, and diverse in type, fully meeting all the characteristics of big data and representing a typical example of big data. With the advancement of smart grids, the digitalization, informatization, and intelligentization of power systems are continuously developing. Power big data has acquired a wider range of data sources, such as data generated by smart meters from hundreds of millions of households and businesses, and rapidly growing monitoring data collected by power equipment status monitoring systems from tens of thousands of generators, transformers, switchgear, overhead lines, high-voltage cables, and other equipment. Therefore, amidst the explosive growth of power big data, traditional data processing technologies have encountered bottlenecks and are unable to meet the analytical needs for rapidly acquiring knowledge and information from massive amounts of data. The integration of power big data is an inevitable requirement for the informatization and intelligent development of the power industry.

[0003] As a fundamental component of the power grid, power equipment is the foundation for its safe, reliable, and stable operation. Power equipment data primarily consists of basic data, condition monitoring data, operation and maintenance data, public safety data, and historical data. Condition monitoring data is a crucial basis for analyzing and predicting equipment status development. With the development of condition monitoring technology and the interconnectedness of various automated information application systems in the power grid, power equipment status data exhibits the characteristics of big data: large scale, rapid update speed, multi-source heterogeneity, and low value density. Therefore, it is suitable and feasible to analyze and process data using big data fusion technology. To address the heterogeneity of power equipment data, a unified data exchange standard needs to be established. Furthermore, given the diverse sources and massive volume of power equipment data, data fusion is necessary to reduce data transmission volume, improve data accuracy, and provide the most powerful decision-making support for intelligent systems.

[0004] When fusing multi-source data, only the fusion error is usually considered. However, multi-source data contains a large amount of redundant data. The transmission of redundant data wastes bandwidth resources and even causes network congestion in severe cases, reducing fusion efficiency. The data that needs to be processed in the data fusion process may come from different types of data, and the data is multi-source and heterogeneous. Moreover, these data are separated in different systems, forming highly heterogeneous data, which brings great challenges to data processing. Summary of the Invention

[0005] The purpose of the present invention is to solve the above problems existing in the prior art and to provide a method for fusion of power equipment status data based on big data, which can effectively improve the efficiency of power equipment data fusion.

[0006] To achieve the above objectives, the present invention provides the following technical solutions:

[0007] A method for fusion of power equipment status data based on big data comprises the following steps in sequence:

[0008] Step A: constructing an electric power equipment information model;

[0009] Step B: first extract and integrate the collected power equipment status data based on the power equipment information model, and calculate the simulation data of each logical node type based on the RPN convolutional neural network, then compare the simulation data with the corresponding standard data to screen out the logical node types with the same simulation data as the standard data, wherein the standard data is the power equipment status data after information extraction and integration;

[0010] Step C: first extract features from the filtered logical node type instance data, then associate the extracted feature data in space, time axis, and redundant data, and fuse the associated data into indicators for predictive analysis data, wherein the indicators for predictive analysis data include at least one of economic indicators, safety indicators, and energy-saving indicators of power equipment.

[0011] In step C, the feature extraction is as follows: first determining feature data based on the CCIPCA algorithm, then performing correlation matrix analysis on the obtained feature data in batches, and eliminating feature data in the same batch that is irrelevant to other feature data in the batch.

[0012] The elimination method is as follows: numbering each logical node contained in the power equipment, and using the power parameter expression of the equipment to describe the three-dimensional coordinate value of each logical node, combining the power equipment information model data to generate three-dimensional data, and then each orientation coordinate value is used as a comparison value, and each characteristic data value is compared with the comparison value. If the difference between the two exceeds the threshold range, the characteristic data is eliminated.

[0013] In step C, the fusion is normalized fusion:

[0014]

[0015] In the above formula, Y a 、Y ab are the feature data before and after fusion, Y max 、Y minare the maximum and minimum values ​​of the feature data respectively.

[0016] In step C, the redundant data is data monitored in real time by sensors;

[0017] The method for determining and processing redundant data is as follows: for the same pixel position, starting from the last selected non-redundant data, all data within the time window threshold T are eliminated as redundant data. For the eliminated redundant data, the redundant data of all sensors used to measure the same state quantity are weighted averaged and then output.

[0018] The step C uses the JPDA algorithm to associate spatial, temporal and redundant data.

[0019] In step B, if the simulation data of a certain logical node type is partially identical to the corresponding standard data, it is split and compared with the standard data one by one. If there is data in the standard data that is not defined by the power equipment information model, it is added or modified accordingly in the power equipment information model and then re-simulated.

[0020] The power equipment information model includes a server layer, a logical device layer, a logical node layer, and a data and attribute information layer. The server is composed of at least one logical device, which is used to provide data and allow other functional nodes to access its resources. A single power device contains at least one server. The logical device is composed of at least one logical node and corresponding services. A single logical node is a specific function in a single physical device and is described by specific data and functional attributes. Multiple logical nodes exchange data and complete related functions through logical connections.

[0021] The information extraction and integration means: first decomposing the functions of the power equipment into multiple logical nodes, then identifying the specific data and the corresponding functional attributes and grouping them into the corresponding logical nodes, then combining the related logical nodes into logical devices, and combining the logical devices into servers.

[0022] In step A, the power equipment status data includes static equipment data, spatial information data, and model topology data. The static equipment data includes equipment cloud maintenance, inspection, technical transformation, infrastructure and power grid operation data. The spatial information data includes online data, offline data and video monitoring information data. The model topology data includes functional data for online monitoring of the main network and distribution network of the power equipment.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. The present invention provides a method for fusion of power equipment status data based on big data. The method first constructs a power equipment information model, then extracts and integrates the collected power equipment status data based on the power equipment information model, and calculates each logical node type simulation data based on the RPN convolutional neural network. The simulation data is then compared with the corresponding standard data to screen out the logical node types with the same simulation data as the standard data. Feature extraction is then performed on the screened logical node type instance data. Finally, spatial, temporal, and redundant data association is performed on the data, and the associated data is fused into indicators for pre-judgment analysis data. This method constructs a power equipment information model, extracts and analyzes the data information calculated based on the model, establishes spatial, temporal, and redundant data associations, and fuses the associated data into indicators for pre-judgment analysis data. Based on these indicators, its impact on the power equipment can be judged, realizing multi-dimensional, real-time data perception and fusion sharing, thereby solving the problems of repeated data access and confusing data definitions, and improving the efficiency of power equipment data fusion. Therefore, the present invention improves the efficiency of power equipment data fusion.

[0025] 2. The present invention's big data-based power equipment status data fusion method analyzes and processes redundant data, controlling inter-node communication in an event-driven manner based on a time window threshold T. If threshold T is reached, i.e., an event occurs, communication occurs; otherwise, the nodes do not communicate. This method of controlling the number of inter-node communications effectively reduces the transmission of redundant data, prevents network congestion, and further improves fusion efficiency. Therefore, the present invention further improves the efficiency of power equipment data fusion. DETAILED DESCRIPTION

[0026] The present invention will be further described below in conjunction with specific embodiments.

[0027] A method for fusion of power equipment status data based on big data comprises the following steps in sequence:

[0028] Step A: constructing an electric power equipment information model;

[0029] Step B: first extract and integrate the collected power equipment status data based on the power equipment information model, and calculate the simulation data of each logical node type based on the RPN convolutional neural network, then compare the simulation data with the corresponding standard data to screen out the logical node types with the same simulation data as the standard data, wherein the standard data is the power equipment status data after information extraction and integration;

[0030] Step C: first extract features from the filtered logical node type instance data, then associate the extracted feature data in space, time axis, and redundant data, and fuse the associated data into indicators for predictive analysis data, wherein the indicators for predictive analysis data include at least one of economic indicators, safety indicators, and energy-saving indicators of power equipment.

[0031] In step C, the feature extraction is as follows: first determining feature data based on the CCIPCA algorithm, then performing correlation matrix analysis on the obtained feature data in batches, and eliminating feature data in the same batch that is irrelevant to other feature data in the batch.

[0032] The elimination method is as follows: numbering each logical node contained in the power equipment, and using the power parameter expression of the equipment to describe the three-dimensional coordinate value of each logical node, combining the power equipment information model data to generate three-dimensional data, and then each orientation coordinate value is used as a comparison value, and each characteristic data value is compared with the comparison value. If the difference between the two exceeds the threshold range, the characteristic data is eliminated.

[0033] In step C, the fusion is normalized fusion:

[0034]

[0035] In the above formula, Y a 、Y ab are the feature data before and after fusion, Y max 、Y min are the maximum and minimum values ​​of the feature data respectively.

[0036] In step C, the redundant data is data monitored in real time by sensors;

[0037] The method for determining and processing redundant data is as follows: for the same pixel position, starting from the last selected non-redundant data, all data within the time window threshold T are eliminated as redundant data. For the eliminated redundant data, the redundant data of all sensors used to measure the same state quantity are weighted averaged and then output.

[0038] The step C uses the JPDA algorithm to associate spatial, temporal and redundant data.

[0039] In step B, if the simulation data of a certain logical node type is partially identical to the corresponding standard data, it is split and compared with the standard data one by one. If there is data in the standard data that is not defined by the power equipment information model, it is added or modified accordingly in the power equipment information model and then re-simulated.

[0040] The power equipment information model includes a server layer, a logical device layer, a logical node layer, and a data and attribute information layer. The server is composed of at least one logical device, which is used to provide data and allow other functional nodes to access its resources. A single power device contains at least one server. The logical device is composed of at least one logical node and corresponding services. A single logical node is a specific function in a single physical device and is described by specific data and functional attributes. Multiple logical nodes exchange data and complete related functions through logical connections.

[0041] The information extraction and integration means: first decomposing the functions of the power equipment into multiple logical nodes, then identifying the specific data and the corresponding functional attributes and grouping them into the corresponding logical nodes, then combining the related logical nodes into logical devices, and combining the logical devices into servers.

[0042] In step A, the power equipment status data includes static equipment data, spatial information data, and model topology data. The static equipment data includes equipment cloud maintenance, inspection, technical transformation, infrastructure and power grid operation data. The spatial information data includes online data, offline data and video monitoring information data. The model topology data includes functional data for online monitoring of the main network and distribution network of the power equipment.

[0043] The principle of the present invention is described as follows:

[0044] Creating logical node instances: defining the mapping relationship between the group title of the modeling symbol and the logical node class name, creating the modeling symbol, storing the timestamp of the modeling symbol, and automatically aggregating the modeling symbols by the modeling rules to form different instances of each group to be processed.

[0045] In view of the problem that directly extracting features from the collected data will result in large errors, increase the burden on the system, and affect the data processing efficiency, the present invention proposes to extract and integrate the collected raw data information based on the constructed power equipment information model, and compare the simulated data with the corresponding standard data, and then screen out the logical node types with the same information for subsequent operations, so as to achieve faster data processing efficiency while reducing errors.

[0046] The method for constructing the power equipment information model is as follows:

[0047] Logical nodes and data modeling: Logical nodes are determined by function. First, the functions of the power equipment are determined, then they are decomposed by the functions, and finally the data attributes and attribute values ​​(i.e., data sets) involved in the functions are defined. Logical nodes can be at least one of the following: liquid dielectric insulation SIML, gas dielectric insulation SIMG, arc monitoring and diagnosis SARC, partial discharge monitoring and diagnosis SPDC, power transformer YPTR, voltage transformer TVTR, current transformer TCTR, circuit breaker XCBR, switch CSWI, measurement node MMXU, and metering node MMTR.

[0048] Logical device modeling: Logical devices are used to reflect the functions of power equipment and perform data conversion. Logical device modeling consists of logical nodes and services. A single logical device can only reflect the status monitoring of a single power equipment.

[0049] Server modeling: Classify each instance according to different forms to form a consistent information exchange service model, namely the server model. A single physical device contains one or more servers. A single physical device can be modeled as a server object and contains at least one logical device and common components provided by communication services, namely time synchronization and file transfer. There are two modes of server communication. One is the publisher / subscriber mode, which uses a point-to-point communication method to quickly transmit information between logical devices. The other is the server / client mode, which is used to control parameter settings and data reading and sending.

[0050] The power parameter expression of the device described in the present invention is obtained based on the actual operating data of the device through power engineering calculation formulas.

[0051] Example 1:

[0052] A method for fusion of power equipment status data based on big data is carried out in the following steps:

[0053] 1. Construct an information model for electric power equipment, including a server layer, a logical device layer, a logical node layer, and a data and attribute information layer. The server is composed of at least one logical device, used to provide data and allow other functional nodes to access its resources. A single electric power device contains at least one server. The logical device is composed of at least one logical node and corresponding services. A single logical node is a specific function in a single physical device, described by specific data and functional attributes. Multiple logical nodes exchange data and complete related functions through logical connections.

[0054] 2. Set the model, capacity, load loss, voltage level and input time of the power equipment, and use sensors to collect partial discharge signals, dielectric loss, hot spot temperature, gas content in oil, core magnetic field strength, core grounding current, vibration signals, low-voltage pulse signals, and frequency response signals;

[0055] 3. Input the above data into the constructed power equipment information model for information extraction and integration. Calculate with the RPN convolutional neural network to obtain logical node type simulation data, and then compare the simulation data with the corresponding standard data. If the simulation data is the same as the standard data, filter it out and proceed to step 4. If the simulation data is partially the same as the standard data, split it and compare it with the standard data one by one. If there is data in the standard data that is not defined by the power equipment information model, add or modify it in the power equipment information model and then re-simulate. The information extraction and integration refers to: first decomposing the functions of the power equipment into multiple logical nodes, then identifying the specific data and the corresponding functional attributes and grouping them into the corresponding logical nodes, and then combining the relevant logical nodes into logical devices, and combining the logical devices into servers. The standard data is the power equipment status data after information extraction and integration;

[0056] 4. First, determine the characteristic data of the filtered logical node type instance data based on the CCIPCA algorithm, then perform correlation matrix analysis on the obtained characteristic data in batches, and eliminate characteristic data in the same batch that is irrelevant to other characteristic data in the batch. The elimination method is as follows: number each logical node included in the power equipment, and use the power equipment parameter function expression to describe the three-dimensional coordinate value of each logical node. Combined with the power equipment information model data, generate three-dimensional data, and then use the coordinate value of each direction as a comparison value. Compare each characteristic data value with the comparison value. If the difference between the two exceeds a threshold range, eliminate the characteristic data.

[0057] 5. Use JPDA algorithm to associate the logical node type instance data after feature extraction with space, time axis and redundant data, where:

[0058] The redundant data is data monitored in real time by sensors. The method for determining and processing redundant data is as follows: for the same pixel position, starting from the last selected non-redundant data, all data within the time window threshold T (which can be 40-90 milliseconds) are removed as redundant data. For the removed redundant data, the redundant data of all sensors used to measure the same state quantity are weighted averaged and output:

[0059]

[0060] In the above formula, n is the number of sensors measuring a certain state quantity of the device, w iis the weighting coefficient of the i-th sensor, X i is the output data of the i-th sensor;

[0061] 6. Normalize and fuse the associated logical node type instance feature data:

[0062]

[0063] In the above formula, Y a 、Y ab are the feature data before and after fusion, Y max 、Y min are the maximum and minimum values ​​of the characteristic data respectively;

[0064] 7. Predict the equipment health status indicators based on the fused data and calculate the entropy weight m of each indicator:

[0065]

[0066] C d =Y ab ×InY ab

[0067] In the above formula, m d is the entropy weight of the dth indicator. The smaller the entropy weight, the better the indicator. ab is the logical node type instance data corresponding to the dth indicator, C d is the information entropy of the d-th indicator.

Claims

1. A method for fusion of power equipment status data based on big data, characterized by: The method comprises the following steps in sequence: Step A: constructing an electric power equipment information model; Step B: first extract and integrate the collected power equipment status data based on the power equipment information model, and obtain simulation data of each logical node type based on the RPN convolutional neural network calculation, and then compare the simulation data with the corresponding standard data to screen out the logical node type with the same simulation data as the standard data, wherein the standard data is the power equipment status data after information extraction and integration, and the power equipment information model includes a server layer, a logical device layer, a logical node layer, and a data and attribute information layer. The server is composed of at least one logical device, which is used to provide data and allow other functional nodes to access its resources. A single power device contains at least one server, and the logical device is composed of at least one logical node and corresponding services. A single logical node is a specific function in a single physical device, which is described by specific data and functional attributes. Multiple logical nodes exchange data and complete related functions through logical connections; The method for constructing the power equipment information model is as follows: Logical nodes and data modeling: Logical nodes are determined by function. First, the functions of the power equipment are determined, then they are decomposed by function, and finally the data attributes and attribute values ​​related to the functions are defined. Logical nodes are at least one of liquid dielectric insulation, gas dielectric insulation, arc monitoring and diagnosis, partial discharge monitoring and diagnosis, power transformer, voltage transformer, current transformer, circuit breaker, switch, measurement node, and metering node. Logical device modeling: Logical devices are used to reflect the functions of power equipment and perform data conversion. Logical device modeling consists of logical nodes and services. A single logical device can only reflect the status monitoring of a single power equipment. Server modeling: Classify each instance into different forms to form a consistent information exchange service model, namely the server model. A single physical device can be modeled as a server object and contains at least one logical device and common components provided by communication services, namely time synchronization and file transfer. The information extraction and integration refers to: first decomposing the functions of the power equipment into multiple logical nodes, then identifying the specific data and corresponding functional attributes and grouping them into the corresponding logical nodes, then combining the related logical nodes into logical devices, and combining the logical devices into servers; Step C: first extract features from the filtered logical node type instance data, then associate the extracted feature data in space, time axis, and redundant data, and fuse the associated data into indicators for predictive analysis data, wherein the indicators for predictive analysis data include at least one of economic indicators, safety indicators, and energy-saving indicators of power equipment.

2. The method for fusion of power equipment status data based on big data according to claim 1, characterized in that: In step C, the feature extraction is as follows: first determining feature data based on the CCIPCA algorithm, then performing correlation matrix analysis on the obtained feature data in batches, and eliminating feature data in the same batch that is irrelevant to other feature data in the batch.

3. The method for fusion of power equipment status data based on big data according to claim 2, characterized in that: The elimination method is as follows: numbering each logical node contained in the power equipment, and using the power parameter expression of the equipment to describe the three-dimensional coordinate value of each logical node, combining the power equipment information model data to generate three-dimensional data, and then each orientation coordinate value is used as a comparison value, and each characteristic data value is compared with the comparison value. If the difference between the two exceeds the threshold range, the characteristic data is eliminated.

4. The method for fusion of power equipment status data based on big data according to claim 1, characterized in that: In step C, the fusion is normalized fusion: ； In the above formula, 、 are the feature data before and after fusion, 、 are the maximum and minimum values ​​of the feature data respectively.

5. The method for fusion of power equipment status data based on big data according to claim 1, characterized in that: In step C, the redundant data is data monitored in real time by sensors; The method for determining and processing redundant data is as follows: for the same pixel position, starting from the last selected non-redundant data, all data within the time window threshold T are eliminated as redundant data. For the eliminated redundant data, the redundant data of all sensors used to measure the same state quantity are weighted averaged and then output.

6. The method for fusion of power equipment status data based on big data according to claim 1, characterized in that: The step C uses the JPDA algorithm to associate spatial, temporal and redundant data.

7. The method for fusion of power equipment status data based on big data according to claim 1, characterized in that: In step B, if the simulation data of a certain logical node type is partially identical to the corresponding standard data, it is split and compared with the standard data one by one. If there is data in the standard data that is not defined by the power equipment information model, it is added or modified accordingly in the power equipment information model and then re-simulated.

8. The method for fusion of power equipment status data based on big data according to claim 1, characterized in that: In step A, the power equipment status data includes static equipment data, spatial information data, and model topology data. The static equipment data includes equipment cloud maintenance, inspection, technical transformation, infrastructure and power grid operation data. The spatial information data includes online data, offline data and video monitoring information data. The model topology data includes functional data for online monitoring of the main network and distribution network of the power equipment.

Images