CN115526255A - Power distribution station equipment identification method and device based on unsupervised deep learning - Google Patents

Abstract

The invention provides a power distribution station equipment identification method and device based on unsupervised deep learning, and relates to the technical field of power distribution detection. The method comprises the following steps: collecting power utilization time sequence data in a preset time period on a secondary side of a main transformer of a power distribution station area; extracting change characteristic data in the power consumption time sequence data based on a preset rule; inputting the change characteristic data into a preset convolutional neural network, and extracting characteristic graphs with different scales; and transmitting the characteristic diagram to a preset detection module to be matched with each electric device in the power distribution area, and identifying a target electric device associated with the power utilization time sequence data. The invention realizes the equipment identification of the distribution station area by using the existing voltage and current sensors as much as possible under the condition of not increasing additional equipment identification hardware.

Description

Power distribution station equipment identification method and device based on unsupervised deep learning

Technical Field

The invention relates to the technical field of power distribution detection, in particular to a power distribution station equipment identification method and device based on unsupervised deep learning.

Background

The intelligent power grid system consists of a master station, a concentrator, a collector and an electric energy meter. Electric power companies often collect power consumption information of electric meters by performing measurements at the main transformer.

In recent years, in order to respond to the call of the national great promotion of artificial intelligence and improve the marketing management level of a power grid, the low-voltage transformer area centralized meter reading transformation engineering is gradually developed in various places. However, due to lack of strict control of construction quality, a line connection failure sometimes occurs, resulting in a wrong design of a distribution substation. In addition, some old urban areas have complex lines, incomplete transformer area maintenance and incomplete table exchange information updating, so that equipment in the distribution station area is difficult to identify.

Disclosure of Invention

In view of the above problems, the present invention provides a power distribution area device identification method and apparatus based on unsupervised deep learning.

The invention provides a power distribution station equipment identification method based on unsupervised deep learning, which comprises the following steps:

collecting power utilization time sequence data in a preset time period at a secondary side of a main transformer of a power distribution station area;

extracting change characteristic data in the power consumption time sequence data based on a preset rule;

inputting the change characteristic data into a preset convolutional neural network, and extracting characteristic graphs with different scales;

and transmitting the characteristic diagram to a preset detection module to be matched with each electric device in the power distribution area, and identifying a target electric device associated with the power utilization time sequence data.

The invention provides a power distribution station equipment identification device based on unsupervised deep learning, which comprises the following components:

the data acquisition module is used for acquiring power utilization time sequence data in a preset time period on the secondary side of a main transformer of the power distribution station area;

the change feature extraction module is used for extracting change feature data in the power utilization time sequence data based on a preset rule;

the characteristic diagram extraction module is used for inputting the change characteristic data into a preset convolutional neural network and extracting characteristic diagrams with different scales;

and the equipment identification module is used for transmitting the characteristic diagram to a preset detection module to be matched with each electric equipment in the power distribution station area, and identifying target electric equipment associated with the power utilization time sequence data.

A third aspect of the present invention provides an electronic device comprising: one or more processors; a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the above unsupervised deep learning-based distribution substation equipment identification method.

Compared with the prior art, the power distribution station equipment identification method and device based on unsupervised deep learning provided by the invention at least have the following beneficial effects:

(1) Under the condition of not adding extra equipment identification hardware, the existing voltage and current sensors are utilized as much as possible to realize equipment identification of the power distribution station area;

(2) The relation between the user load equipment and the main transformer in the power distribution area can be obtained through statistical analysis of the power consumption data of the users, and the method has important theoretical research significance and practical application value;

(3) Important characteristics of the current equipment operation condition of the power distribution station area are obtained in an unsupervised learning mode, and the interference of the complex line condition on the result is reduced;

(4) Interlayer jumping is carried out through a large number of residual error structure blocks, storage space occupation is reduced, multi-layer feature fusion is achieved, and multi-scale prediction is introduced to improve accuracy.

Drawings

The above and other objects, features and advantages of the present invention will become more apparent from the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

fig. 1 schematically illustrates an application scenario of a power distribution station equipment identification method based on unsupervised deep learning according to an embodiment of the present invention;

fig. 2 schematically illustrates a flow diagram of a power distribution substation device identification method based on unsupervised deep learning according to an embodiment of the present invention;

figure 3 schematically illustrates a lightweight centret network architecture according to an embodiment of the invention;

fig. 4 is a block diagram schematically illustrating a structure of a power distribution station equipment identification apparatus based on unsupervised deep learning according to an embodiment of the present invention;

fig. 5 schematically shows a block diagram of an electronic device adapted to implement an unsupervised deep learning-based power distribution substation device identification method according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to specific embodiments and the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The terms "comprises," "comprising," and the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. It is noted that the terms used herein should be interpreted as having a meaning that is consistent with the context of this specification and should not be interpreted in an idealized or overly formal sense.

Fig. 1 schematically shows an application scenario diagram of a power distribution station equipment identification method based on unsupervised deep learning according to an embodiment of the present invention. It should be noted that fig. 1 is only an example of an application scenario in which the embodiment of the present invention may be applied to help those skilled in the art understand the technical content of the present invention, and does not mean that the embodiment of the present invention may not be applied to other devices, systems, environments or scenarios.

As shown in fig. 1, an application scenario according to this embodiment may be a power distribution substation 100, where the power distribution substation 100 specifically includes a main transformer 101, a branch control cabinet 102, a user load device 103, and a voltage and/or current sensor 104.

The secondary side of the main transformer 101 is connected to a plurality of different user load devices 103 through a plurality of branch control cabinets 102. The branch control cabinet 102 is used for controlling the power load demand of each user load device 103, and may also adjust some electrical parameters of each user load device 103. A voltage and/or current sensor 104 is installed on the secondary side of the main transformer 101 for collecting time series data.

It should be understood that the number of main transformers, branch control cabinets, user load devices, and voltage and/or current sensors in fig. 1 are merely illustrative. There may be any number of main transformers, branch control cabinets, user load devices, and voltage and/or current sensors, as desired for the implementation.

The method of the embodiment of the present invention will be described in detail below with reference to fig. 2 to 3 based on the application scenario described in fig. 1.

Fig. 2 schematically shows a flowchart of a power distribution station equipment identification method based on unsupervised deep learning according to an embodiment of the present invention.

As shown in fig. 2, the power distribution substation equipment identification method based on unsupervised deep learning according to the embodiment may include operations S210 to S240.

In operation S210, power consumption timing data within a preset time period is collected at a secondary side of a main transformer of a distribution substation area.

In operation S220, change feature data in the power usage time series data is extracted based on a preset rule.

In operation S230, the variation feature data is input into a preset convolutional neural network, and feature maps of different scales are extracted.

In operation S240, the feature map is transmitted to a preset detection module to be matched with each electric device in the power distribution area, and a target electric device associated with the power utilization time sequence data is identified.

The main transformer may be, for example, the main transformer 101 shown in fig. 1, and the consumer may be, for example, the consumer load device 103 shown in fig. 1.

Through the embodiment, the power distribution station equipment identification method based on unsupervised deep learning provided by the invention realizes equipment identification of the power distribution station by using the existing voltage and current sensors as much as possible under the condition of not adding extra equipment identification hardware. The invention can obtain the relation between the user load equipment and the main transformer in the power distribution area through the statistical analysis of the power consumption data of the users, and has important theoretical research significance and practical application value.

In the embodiment of the invention, a plurality of electric equipment in the power distribution area are connected to a main transformer through at least one branch control cabinet; the method comprises the steps that voltage and/or current sensors are mounted on the secondary side of a main transformer of a power distribution station area, and power utilization time sequence data in a preset time period are collected.

In the embodiment of the invention, the electricity utilization time sequence data comprise voltage time sequence data and current time sequence data, and the change characteristic data comprise voltage change characteristic data and current change characteristic data. On this basis, the extracting the change feature data in the electricity consumption time series data based on the preset rule in operation S220 may specifically include:

carrying out time sequence segmentation according to the time point of sudden change of the voltage time sequence data in the time sequence;

taking the maximum difference value of the voltage values in two adjacent time sequence segments as voltage change characteristic data;

and then intercepting a section of time sequence data by taking the two points of the maximum voltage difference value as a starting point and an end point, solving the maximum value and the minimum value of the current value in the section of time sequence data, and taking the current difference value as current change characteristic data.

Specifically, the collected power utilization time sequence data are segmented according to voltage variation, and the maximum difference value of voltage values in two adjacent time sequence segments is taken as voltage variation characteristic data; and then intercepting a section of time sequence data by taking the two points of the maximum voltage difference value as a starting point and an end point, solving the maximum value and the minimum value of the current value in the section of time sequence data, and taking the current difference value as current change characteristic data. The voltage change characteristic data and the current change characteristic data thus together form change characteristic data. .

Figure 3 schematically shows a lightweight centrnet network architecture according to an embodiment of the invention.

As shown in fig. 3, before inputting the change feature data into the preset convolutional neural network in the embodiment of the present invention, the method further includes:

the method comprises the steps of adopting a lightweight CenterNet network structure to construct a power distribution station equipment identification model, wherein the power distribution station equipment identification model comprises a preset convolutional neural network and a preset detection module.

It should be noted that the centrnet network structure is characterized in that: the target is considered to be a point, the center of the target bounding box. Therefore, the target detection problem can be converted into other target attributes, and the direction and the attitude are subjected to parameter regression by taking the estimation center as a reference. The position of the target is estimated from the upper left and lower right corners to determine the pose of the target. The central net is similar to the first order anchor-based method, in that its central point can be regarded as an anchor point without size, with the important difference that: (1) The CenterNet network structure is only placed at a position without size, and a threshold value for distinguishing a foreground and a background does not need to be manually set; (2) For each target there is only one positive anchor point, the key point is characterized by the local peak above, so that the non-maximum suppression (NMS) algorithm is not needed later.

Further, based on the lightweight centrnet network structure, in the power distribution station equipment identification model, the preset convolutional neural network comprises a CBL component, a Res Unit component and a Res _ N component.

The CBL component is the smallest component among them, and is composed of a convolutional layer (Conv), a Batch normalization layer (Batch Norm), and an activation function layer (Leak ReLU).

The Res Unit component is a residual error structure composed of 2 CBL components and add operation, and specifically means that an initial input value of the Unit is firstly copied to a temporary variable, then the input value is processed by the 2 CBL components, and finally an output result of the input value is added (add) with a value of the temporary variable to obtain a final output value of the Unit, so that residual error jump is realized.

The Res _ N fabric block consists of 1 CBL component and N Res Unit components, including convolutional layers. Where N is a natural number and the 1 st CBL component in the Res _ N structure block sets the convolution step size to 2, thereby changing the feature size and reducing it by half.

Based on the above structural design of the convolutional neural network, please continue to refer to fig. 3, in an embodiment of the present invention, the feature maps with different scales include a first scale feature map, a second scale feature map, and a third scale feature map. In the operation S230, inputting the changed feature data into the preset convolutional neural network, and extracting feature maps with different scales may specifically include the following steps:

processing the change characteristic data by a CBL assembly for 1 time, then sequentially processing by a Res _1 structure block, a Res _2 structure block and a Res _8 structure block, and at the moment, copying an output value and naming the output value as a first output value;

processing the first output value by a Res _8 structure block, and at the moment, copying one output value and naming the copy output value as a second output value;

processing the second output value by a Res _4 structure block, and at the moment, copying one output value and naming the copy output value as a third output value;

processing the third output value by 5 times of CBL assembly, copying one output value and naming the output value as a fourth output value, and processing the fourth output value by 1 time of CBL assembly and 1 convolution layer to obtain a first scale characteristic diagram;

processing the fourth output value by the CBL assembly and the upsampling for 1 time, then splicing (concat) the fourth output value with the second output value to obtain a first spliced value, processing the first spliced value by the CBL assembly for 5 times, copying one output value and naming the output value as a fifth output value, and processing the fifth output value by the CBL assembly for 1 time and 1 convolution layer to obtain a second scale characteristic diagram;

and processing the fifth output value by the CBL module and the upsampling for 1 time, splicing the fifth output value with the first output value to obtain a second spliced value, and processing the second spliced value by the CBL module, the CBL module and the convolution layer for 5 times in sequence to obtain a third scale characteristic diagram.

The acquisition process of the first output value, the second output value and the third output value jointly form a traditional Darknet-53 network, and the 8-fold, 16-fold and 32-fold down-sampling features of the input feature map can be extracted respectively.

Through the embodiment, the convolutional neural network is used for extracting the voltage and current time sequence change characteristic data, and the extracted characteristic diagram is transmitted to the detection module. And moreover, interlayer jumping is carried out through a large number of residual error structure blocks, the storage space occupation is reduced, multi-layer feature fusion is realized, and multi-scale prediction is introduced to improve the accuracy.

Next, in an embodiment of the present invention, the transmitting the feature map to the preset detection module in operation S240 to match with each electric device in the power distribution area may specifically include:

inputting the first scale feature map, the second scale feature map and the third scale feature map into a detection module, generating 3 prior frames with different sizes for the feature map of each scale, and decoding each prior frame to obtain a detection frame;

dividing the feature map of each scale into a plurality of Grid cells (Grid cells) according to preset division parameters;

and detecting the detection frame according to a preset evaluation index threshold value aiming at each grid unit to obtain a detection result, wherein the detection result comprises the position, the detection confidence and the type of the detection frame.

The detection frame location may include, for example, center point coordinates, width, and height of the target bounding box.

Specifically, three feature maps with different scales are obtained after the time sequence change feature data pass through the convolutional neural network, 3 prior frames with different sizes are generated for each feature map, and the prior frames are decoded to obtain the detection frame. In the reasoning process, the lightweight CenterNet network structure divides input data into S multiplied by S grid units according to the size of a final characteristic diagram to be detected, the detection module utilizes a convolutional neural network to obtain the input data in an unsupervised learning mode, a detection confidence function is arranged at the output result of the detection module, and a threshold value is set for an evaluation index. Considering the light weight processing of the network, all output results lower than the evaluation index threshold value are filtered, then one result with the highest evaluation index is selected to be output and used as the final output, and the center point coordinate, the width and the height and the confidence score of the target frame are calculated.

According to the embodiment of the invention, the important characteristics of the current equipment operation condition of the power distribution station area are obtained in an unsupervised learning mode, and the interference of the complex line condition on the result is reduced.

Based on the method disclosed above, the invention further provides a power distribution station equipment identification device based on unsupervised deep learning, which will be described in detail below with reference to fig. 4.

Fig. 4 is a block diagram schematically illustrating a power distribution area equipment identification apparatus based on unsupervised deep learning according to an embodiment of the present invention.

As shown in fig. 4, the power distribution substation equipment identification apparatus 400 based on unsupervised deep learning according to the embodiment includes a data acquisition module 410, a variation feature extraction module 420, a feature map extraction module 430, and an equipment identification module 440.

And the data acquisition module 410 is used for acquiring power utilization time sequence data in a preset time period on the secondary side of a main transformer of the power distribution station area.

And the change feature extraction module 420 is configured to extract change feature data in the power consumption time series data based on a preset rule.

The feature map extracting module 430 is configured to input the changed feature data into a preset convolutional neural network, and extract feature maps of different scales.

And the device identification module 440 is configured to transmit the feature map to a preset detection module to match with each electric device in the power distribution area, and identify a target electric device associated with the power utilization time sequence data.

It should be noted that the embodiment of the apparatus portion is similar to the embodiment of the method portion, and the achieved technical effects are also similar, for details, please refer to the method embodiment, which is not described herein again.

According to the embodiment of the present invention, any plurality of the data collection module 410, the variation feature extraction module 420, the feature map extraction module 430, and the device identification module 440 may be combined into one module to be implemented, or any one of the modules may be split into a plurality of modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of the other modules and implemented in one module. According to an embodiment of the present invention, at least one of the data collection module 410, the changed feature extraction module 420, the feature map extraction module 430, and the device identification module 440 may be implemented at least partially as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented in hardware or firmware in any other reasonable manner of integrating or packaging a circuit, or may be implemented in any one of three implementations of software, hardware, and firmware, or in any suitable combination of any of them. Alternatively, at least one of the data collection module 410, the variation feature extraction module 420, the feature map extraction module 430 and the device identification module 440 may be implemented at least in part as a computer program module that, when executed, may perform a corresponding function.

Fig. 5 schematically shows a block diagram of an electronic device adapted to implement an unsupervised deep learning based power distribution substation device identification method according to an embodiment of the present invention.

As shown in fig. 5, an electronic device 500 according to an embodiment of the present invention includes a processor 501 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 502 or a program loaded from a storage section 508 into a Random Access Memory (RAM) 503. The processor 501 may include, for example, a general purpose microprocessor (e.g., CPU), an instruction set processor and/or associated chipset, and/or a special purpose microprocessor (e.g., application Specific Integrated Circuit (ASIC)), among others. The processor 501 may also include onboard memory for caching purposes. Processor 501 may include a single processing unit or multiple processing units for performing the different actions of the method flows according to embodiments of the present invention.

In the RAM503, various programs and data necessary for the operation of the electronic apparatus 500 are stored. The processor 501, the ROM 502, and the RAM503 are connected to each other by a bus 504. The processor 501 performs various operations of the method flow according to the embodiments of the present invention by executing programs in the ROM 502 and/or the RAM 503. Note that the programs may also be stored in one or more memories other than the ROM 502 and the RAM 503. The processor 501 may also perform various operations of method flows according to embodiments of the present invention by executing programs stored in the one or more memories.

According to an embodiment of the present invention, electronic device 500 may also include an input/output (I/O) interface 505, input/output (I/O) interface 505 also being connected to bus 504. The electronic device 500 may also include one or more of the following components connected to the I/O interface 505: an input portion 506 including a keyboard, a mouse, and the like; an output portion 507 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage portion 508 including a hard disk and the like; and a communication section 509 including a network interface card such as a LAN card, a modem, or the like. The communication section 509 performs communication processing via a network such as the internet. The driver 510 is also connected to the I/O interface 505 as necessary. A removable medium 511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 510 as necessary, so that a computer program read out therefrom is mounted into the storage section 508 as necessary.

Some block diagrams and/or flowcharts are shown in the figures. It will be understood that some blocks of the block diagrams and/or flowchart illustrations, or combinations thereof, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor, create means for implementing the functions/acts specified in the block diagrams and/or flowchart block or blocks.

Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless explicitly specified otherwise. Further, the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements.

The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A power distribution station equipment identification method based on unsupervised deep learning is characterized by comprising the following steps:

collecting power utilization time sequence data in a preset time period at a secondary side of a main transformer of a power distribution station area;

extracting change characteristic data in the power consumption time sequence data based on a preset rule;

inputting the change characteristic data into a preset convolutional neural network, and extracting characteristic graphs with different scales;

and transmitting the characteristic diagram to a preset detection module to be matched with each electric device in the power distribution area, and identifying a target electric device associated with the power utilization time sequence data.

2. The unsupervised deep learning-based distribution substation equipment identification method according to claim 1, wherein the electricity usage time series data comprises voltage time series data and current time series data, and the change characteristic data comprises voltage change characteristic data and current change characteristic data;

wherein the extracting of the change feature data in the power consumption time series data based on the preset rule comprises:

carrying out time sequence segmentation according to the time point of the voltage time sequence data which is subjected to sudden change in time sequence;

taking the maximum difference value of the voltage values in two adjacent time sequence segments as voltage change characteristic data;

and then intercepting a section of time sequence data by taking the two points of the maximum voltage difference value as a starting point and an end point, solving the maximum value and the minimum value of the current value in the section of time sequence data, and taking the current difference value as current change characteristic data.

3. The unsupervised deep learning-based power distribution substation equipment identification method according to claim 2, wherein a plurality of electric devices in the power distribution substation are connected to the main transformer through at least one branch control cabinet;

and acquiring power utilization time sequence data in a preset time period through a voltage and/or current sensor arranged on the secondary side of a main transformer of the distribution area.

4. The unsupervised deep learning-based power distribution station equipment identification method according to claim 1, wherein before the step of inputting the variation characteristic data into a preset convolutional neural network, the method further comprises the following steps:

and constructing a power distribution station equipment identification model by adopting a lightweight CenterNet network structure, wherein the power distribution station equipment identification model comprises the preset convolutional neural network and a preset detection module.

5. The unsupervised deep learning-based distribution substation equipment identification method according to claim 4, wherein said preset convolutional neural network comprises a CBL component, a Res Unit component and a Res _ N component, wherein:

the CBL assembly consists of a convolution layer, a batch normalization layer and an activation function layer;

the Res Unit component is a residual error structure consisting of 2 CBL components and add operations;

the Res _ N structure block is composed of 1 CBL assembly and N Res Unit assemblies, N is a natural number, and the convolution step length of the 1 st CBL assembly in the Res _ N structure block is set to be 2.

6. The unsupervised deep learning-based power distribution station equipment identification method according to claim 5, wherein the feature maps of different scales comprise a first scale feature map, a second scale feature map and a third scale feature map, the change feature data is input into a preset convolutional neural network, and the extracting of the feature maps of different scales specifically comprises:

processing the change characteristic data by 1 CBL assembly, then sequentially processing by a Res _1 structure block, a Res _2 structure block and a Res _8 structure block, and at the moment, copying an output value and naming the output value as a first output value;

processing the first output value by a Res _8 structure block, and at the moment, copying one output value and naming the copy output value as a second output value;

processing the second output value by a Res _4 structure block, and at the moment, copying one output value and naming the copy output value as a third output value;

processing the third output value by 5 times of CBL assembly, copying one output value and naming the output value as a fourth output value, and processing the fourth output value by 1 time of CBL assembly and 1 convolution layer to obtain a first scale characteristic diagram;

processing the fourth output value by a CBL assembly and an upsampling for 1 time, splicing the fourth output value with the second output value to obtain a first spliced value, processing the first spliced value by the CBL assembly for 5 times, copying one output value and naming the output value as a fifth output value, and processing the fifth output value by the CBL assembly for 1 time and 1 convolution layer to obtain a second scale characteristic diagram;

and processing the fifth output value by the CBL assembly and the upsampling for 1 time, splicing the fifth output value with the first output value to obtain a second spliced value, and sequentially processing the second spliced value by the CBL assembly for 5 times, the CBL assembly for 1 time and 1 convolution layer to obtain a third scale characteristic diagram.

7. The power distribution station area equipment identification method based on unsupervised deep learning according to claim 6, wherein the step of transmitting the feature map to a preset detection module to be matched with each electric device in the power distribution station area specifically comprises the steps of:

inputting the first scale feature map, the second scale feature map and the third scale feature map into the detection module, generating 3 prior frames with different sizes for the feature map of each scale, and decoding each prior frame to obtain a detection frame;

dividing the characteristic graph of each scale into a plurality of grid units according to preset division parameters;

and detecting the detection frame according to a preset evaluation index threshold value aiming at each grid unit to obtain a detection result, wherein the detection result comprises a detection frame position, a detection confidence coefficient and a type.

8. The unsupervised deep learning-based distribution substation equipment identification method according to claim 7, wherein the detection frame position comprises center point coordinates, width and height of a target frame.

9. The utility model provides a distribution station equipment recognition device based on unsupervised deep learning which characterized in that includes:

the data acquisition module is used for acquiring power utilization time sequence data in a preset time period on the secondary side of a main transformer of the power distribution station area;

the change feature extraction module is used for extracting change feature data in the power utilization time sequence data based on a preset rule;

the characteristic diagram extraction module is used for inputting the change characteristic data into a preset convolutional neural network and extracting characteristic diagrams with different scales;

and the equipment identification module is used for transmitting the characteristic diagram to a preset detection module to be matched with each electric equipment in the power distribution station area, and identifying target electric equipment associated with the power utilization time sequence data.

10. An electronic device, comprising:

one or more processors;

a storage device for storing one or more programs,

wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of any of claims 1-8.

Images