Substation secondary principle drawing intelligent identification method based on deep learning

Abstract

The invention relates to the field of intelligent recognition, in particular to a substation secondary principle drawing intelligent recognition method based on deep learning, and the method comprises the steps: obtaining a first-stage electrical primitive recognition network obtained in advance; performing a round of training on the electrical primitive recognition network in the first stage to obtain an output result and a feature image of each convolutional layer; calculating the distinguishing difficulty of the two electrical primitives; obtaining a difference normalized image of every two electrical primitive areas in the secondary principle drawing image; taking the distinguishing difficulty as a weight, and carrying out weighted calculation on the difference normalized image to obtain an attention degree image; extracting key information of the secondary principle drawing image by using the attention image; setting the weight of a network parameter in a regularization item according to the correlation between the feature image and the key information; and intelligent identification is realized. The regularization weight is reasonably set, so that network lightweight is realized, and meanwhile, the influence on the network identification precision can be relatively small.

Description

Technical Field

[0001] The present invention relates to the field of intelligent identification, and more specifically, to a method for intelligent identification of secondary schematic drawings of a substation based on deep learning. Background Art

[0002] Substation secondary schematics play an important role in the construction, expansion, technical transformation, operation and maintenance of substations. However, the management of substation secondary schematics is a time-consuming and labor-intensive manual work. Therefore, how to intelligently manage substation secondary schematics has become a new development demand. To achieve effective management of substation secondary schematics, it is necessary to accurately identify the valid information in the substation secondary schematics.

[0003] Since the electrical elements in the secondary schematic drawings of the substation are relatively large and varied, a relatively large recognition network is required to accurately identify the electrical elements in the secondary schematic drawings of the substation. However, the recognition efficiency of the large recognition network is low and real-time performance cannot be guaranteed. In order to improve the efficiency of the recognition of the secondary schematic drawings of the substation, the recognition network needs to be lightweight; wherein network regularization can reduce the number of non-zero network parameters by suppressing the network parameters, so the network regularization can be used to achieve the effect of network pruning and realize network lightweighting. In the process of network regularization, the regularization weight parameter will affect the degree of suppression of the network parameters. If the regularization weight parameter is not set properly, it is easy to suppress some important network parameters, and some unimportant parameters are not suppressed, thereby reducing the accuracy of the recognition network. Therefore, how to reasonably set the regularization weight parameter becomes the research focus of the present invention.

[0004] The patent application document with publication number CN115223189A discloses a substation secondary drawing recognition method and system, and a retrieval method and system. The focus of the method in the patent application document is to improve the accuracy of text recognition by enhancing the text data in the substation secondary drawings. The method in the patent application document does not involve network lightweight processing, so the method in the patent application cannot solve the problem in the present invention. Summary of the invention

[0005] In order to solve the problem of how to reasonably set the regularization weight parameters, the present invention proposes a substation secondary schematic drawing intelligent recognition method based on deep learning, which includes the following steps: Obtain a data set, the data set consisting of labeled secondary schematic drawing images; Obtain the first-stage electrical element recognition network obtained in advance; use the secondary schematic drawing image to train the first-stage electrical element recognition network for a round of training, obtain the output results and the feature images of each convolution layer; calculate the difficulty of distinguishing two electrical elements , the identification frame containing any one of the two electrical diagram elements in the output result is recorded as a spare identification frame, They represent the probability that the area within the kth spare identification box in the sth secondary schematic drawing image is the jth electrical element and the ith electrical element, respectively. represents the category recognition error of the kth alternative recognition box in the sth secondary schematic drawing image, They represent the label values ​​of the j-th category and the i-th category of the area corresponding to the k-th backup recognition box in the s-th secondary schematic drawing image, respectively. represents the number of spare recognition boxes in the s-th secondary schematic drawing image, Indicates the number of secondary schematic drawings images, Indicates the preset anti-zero coefficient. represents the linear normalization function; Obtain a difference normalized image of each two electrical element regions in the secondary schematic drawing image; perform weighted calculation on the difference normalized image using the difficulty of distinction as a weight to obtain an attention image; The key information of the secondary schematic drawing image is extracted using the attention image; the weights of the network parameters in the regularization term are set according to the correlation between the feature image and the key information to achieve intelligent recognition.

[0006] The present invention accurately evaluates the necessity of each network parameter by the contribution of the information extracted by each network parameter to the primitive recognition, and then sets a suitable regularization weight for each network parameter according to the evaluation result; further, considering that the correlation between the information extracted by each network parameter and the important information required for primitive recognition can reflect the contribution of the information extracted by each network parameter, the key information index of the secondary principle drawing image is introduced to accurately reflect the important information required for primitive recognition; further, when focusing on the key information of the secondary principle drawing image, considering that the distinguishing information of the primitives with greater recognition difficulty should be focused on, the key information of the secondary principle drawing image is introduced. Indicators such as the difficulty of distinguishing are used to accurately reflect the difficulty of identifying graphic elements; further, when obtaining the key information of the secondary schematic drawing image, taking into account that the difference information of the graphic elements can better distinguish different types of graphic elements, the difference analysis of different types of electrical graphic elements is introduced to accurately screen out the key information for graphic element recognition in the secondary schematic drawing image; further, when analyzing the difficulty of distinguishing different types of graphic elements, taking into account that a large distinction difficulty will lead to low recognition accuracy, and at the same time, a large contribution of recognition error to the difficulty of distinguishing graphic elements is large, these characteristics are introduced to accurately analyze the difficulty of distinguishing graphic elements.

[0007] Preferably, the step of obtaining the pre-obtained first-stage electrical graphic element recognition network includes: A pre-built electrical graphic element recognition network is obtained; and a round of training is performed on the electrical graphic element recognition network using the data set to obtain a first-stage electrical graphic element recognition network.

[0008] The present invention allows the electrical graph element recognition network to learn partial information by performing a round of training on the electrical graph element recognition network, thereby providing a basis for the subsequent analysis of the ability to extract information from each network parameter.

[0009] Preferably, the step of obtaining a difference normalized image of each two electrical graphic element regions in the secondary schematic drawing image includes: Extract each electrical element area in the secondary schematic drawing image according to the label; Adjust all electrical graphic element areas to the same size; and perform subtraction of every two electrical graphic element areas after the size adjustment to obtain a difference image; The pixels in the difference image are normalized to obtain a difference normalized image of each two electrical graphic element regions.

[0010] The present invention accurately extracts the difference information of different types of graphic elements through difference analysis, and provides a basis for the subsequent extraction of key information for graphic element recognition in the secondary principle drawing image.

[0011] Preferably, the step of performing weighted calculation on the difference normalized image to obtain the attention image by taking the difficulty of differentiation as a weight includes: Taking the difficulty of distinction as a weight, the weighted sum of the difference normalized images between any electrical graphic element area and all other electrical graphic element areas is performed to obtain an initial attention image; Obtain outer envelope pixels of an electrical primitive in an electrical primitive region; The values ​​of the outer envelope pixels in the initial attention image are adjusted to 1 to obtain the attention image.

[0012] When acquiring an attention image, the present invention introduces information on difficulty of distinction and difference, so that not only information for distinguishing different types of primitives can be extracted from the attention image, but also information that is difficult to distinguish can be extracted, thereby providing a basis for subsequent extraction of key information; further, considering that outer envelope pixels will affect the positioning of the primitive area, the values ​​of outer envelope pixels are adjusted, so that the information of outer envelope pixels can be extracted when extracting key information.

[0013] Preferably, the extracting key information of the secondary schematic drawing image by using the attention image includes: The key information of the secondary schematic drawing image is obtained by multiplying each electrical element area in the secondary schematic drawing image by the attention image.

[0014] Preferably, the step of setting the weight of the network parameter in the regularization term according to the correlation between the feature image and the key information includes: All feature images corresponding to each network parameter are obtained, and the correlation between the feature image and the key information is calculated; the weight of each network parameter is obtained by taking the inverse of the mean of the correlation obtained from all feature images of each network parameter, and the first-stage electrical element recognition network after one round of training is recorded as the second-stage electrical element recognition network, and the regularization term of the loss function of the second-stage electrical element recognition network is constructed based on the weight.

[0015] The present invention reduces the number of non-zero network parameters and improves network recognition efficiency by adding regularization terms; further, by setting the weights of each network parameter in the regularization term, the network recognition accuracy can be less affected when the network is lightweight.

[0016] Preferably, the regularization term adopts L2 regularization.

[0017] Preferably, the intelligent identification includes: Based on the loss function after adding the regularization term, the electrical graph element recognition network of the second stage is continuously trained using the data set until the preset training completion cutoff condition is met, thereby obtaining the electrical graph element recognition network that has been finally trained. The newly acquired secondary schematic drawing image is input into the finally trained electrical graphic element recognition network to obtain the recognition result.

[0018] The present invention has the following beneficial effects: The present invention accurately evaluates the necessity of each network parameter by the contribution of the information extracted from each network parameter to primitive recognition, and then sets a suitable regularization weight for each network parameter according to the evaluation result; Furthermore, considering that the correlation between the information extracted from each network parameter and the important information required for primitive recognition can reflect the contribution of the information extracted from each network parameter, the key information indicators of the secondary schematic drawing image are introduced to accurately reflect the important information required for primitive recognition; Furthermore, when focusing on the key information of the secondary schematic drawing image, considering that the distinguishing information of the graphics elements with greater recognition difficulty should be focused on, the distinction difficulty and other indicators are introduced to accurately reflect the difficulty of graphics element recognition; Furthermore, when acquiring key information of the secondary schematic drawing image, considering that the difference information of the graphic element can better distinguish different types of graphic elements, the difference analysis of different types of electrical graphic elements is introduced to accurately screen out the key information for graphic element recognition in the secondary schematic drawing image; Furthermore, when analyzing the difficulty of distinguishing different types of primitives, taking into account the fact that high difficulty in distinguishing leads to low recognition accuracy and that high difficulty in distinguishing primitives contributes a large amount to recognition errors, the difficulty of distinguishing primitives is accurately analyzed by introducing recognition errors of primitives and error difference analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 It is a flowchart of the steps of a method for intelligently identifying secondary schematic drawings of a substation based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0021] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0022] See also Figure 1 , which shows a flowchart of a method for intelligently identifying secondary schematic drawings of a substation based on deep learning according to an embodiment of the present invention, the method comprising the following steps: S1: Get the dataset.

[0023] Specifically, secondary schematic drawing images of several substations are obtained, and the secondary schematic drawing images are manually labeled. The labels of the secondary schematic drawing images are the width and height of each electrical element in the secondary schematic drawing image, the center position of the electrical element, and the classification one-hot encoding of the electrical element. All labeled secondary schematic drawing images constitute a data set.

[0024] S2: Obtain the first-stage electrical element recognition network obtained in advance; use the secondary schematic drawing image to train the first-stage electrical element recognition network, obtain the output results and the feature image of each convolution layer; calculate the difficulty of distinguishing two electrical elements.

[0025] S20: Acquire the first-stage electrical element recognition network obtained in advance.

[0026] It should be noted that in order to realize secondary schematic drawing recognition, a recognition network must be built first.

[0027] Preferably, as an example, obtaining a pre-obtained first-stage electrical graphic element recognition network includes: A pre-built electrical graphic element recognition network is obtained; and a round of training is performed on the electrical graphic element recognition network using the data set to obtain a first-stage electrical graphic element recognition network.

[0028] Construct an electrical graphic element recognition network. This embodiment uses the YoloV2 network as the electrical graphic element recognition network. Other embodiments may use other networks, and this embodiment does not make specific restrictions.

[0029] Each labeled secondary schematic drawing image in the data set is input into the electrical element recognition network in turn to train the electrical element recognition network. The electrical element recognition network after a round of training for all labeled secondary schematic drawing images is recorded as the first-stage electrical element recognition network.

[0030] It should be noted that it is an existing technology to sequentially input each labeled secondary schematic drawing image in the data set into the electrical graphic element recognition network to train the electrical graphic element recognition network, which will not be described in detail here.

[0031] It should be further explained that in order to set the regularization weights of the network parameters in the recognition network, it is necessary to first determine the importance of each network parameter. Since the network parameters in the initial recognition network are randomly initialized, it is impossible to determine the importance of the network parameters through the initial recognition network. Therefore, the recognition network needs to be trained for a period of time to give it a certain recognition ability, which provides a basis for its subsequent determination of the importance of network parameters.

[0032] S21: Obtain the first-stage electrical element recognition network obtained in advance; use the secondary schematic drawing image to perform a round of training on the first-stage electrical element recognition network to obtain the output results and the feature image of each convolution layer; calculate the difficulty of distinguishing two electrical elements.

[0033] It should be noted that in order to set appropriate regularization weights for network parameters, it is necessary to determine the contribution of each network parameter to drawing recognition. If the information extracted by the network parameter contributes more to recognition, the network parameter contributes more to drawing recognition, and therefore the network parameter should not be suppressed, and the regularization weight of the network parameter should be set smaller. If the information extracted by the network parameter contributes less to recognition, or even interferes with recognition, the network adoption number contributes less to drawing recognition, and therefore the network parameter should be suppressed, and the regularization weight of the network parameter should be set larger.

[0034] It should be further explained that different types of electrical elements have different differences. Some types of electrical elements have large differences, while some types of electrical elements have small differences. Among them, the electrical elements with large differences are more difficult to classify, while the electrical elements with small differences are less difficult to classify. If the network parameter has a stronger ability to extract distinguishing information for electrical elements with greater classification difficulty, the network parameter is more important. Therefore, it is necessary to judge the extraction of distinguishing information of electrical elements with different classification difficulties by each network parameter. First, obtain the difficulty of distinguishing different types of electrical elements.

[0035] Preferably, as an example, the secondary schematic drawing image is input into the electrical element recognition network of the first stage for a round of training, and the output result and the feature image of each convolution layer are obtained, including: The labeled secondary schematic drawing images are sequentially input into the first-stage electrical element recognition network, and the first-stage electrical element recognition network is trained to obtain the output results obtained when each secondary schematic drawing image is input and the feature image of each convolution layer.

[0036] It can be understood that the output results can reflect the recognition of each element in the secondary schematic drawing image by the electrical element recognition network in the first stage, and the feature image of each convolutional layer reflects the extraction of each information in the secondary schematic drawing image by the electrical element recognition network in the first stage. These data are used to analyze the extraction of information distinguishing electrical elements of different classification difficulties by each network parameter, so they should be extracted first.

[0037] Preferably, as an example, calculating the difficulty of distinguishing between every two electrical graphic elements includes:

[0038] The identification frame containing any one of the two electrical elements in the output result is recorded as a spare identification frame. Indicates the probability that the area within the kth spare identification box in the sth secondary schematic drawing image is the jth electrical element represents the probability that the area within the kth spare identification box in the sth secondary schematic drawing image is the i-th electrical element, Indicates the label value of the jth category of the area corresponding to the kth backup recognition box in the sth secondary schematic drawing image They represent the label value of the i-th category of the area corresponding to the k-th backup recognition box in the s-th secondary schematic drawing image, represents the category recognition error of the kth alternative recognition box in the sth secondary schematic drawing image, represents the number of spare recognition boxes in the s-th secondary schematic drawing image, Indicates the number of secondary schematic drawings images, represents the linear normalization function, It represents a preset anti-zero coefficient, which is used to prevent the denominator from being 0. This embodiment is described by taking the preset anti-zero value of 0.001 as an example. Other embodiments may take other values, and this embodiment does not make any specific limitation. Indicates the difficulty of distinguishing the i-th electrical element from the j-th electrical element.

[0039] It should be added that an example is given to illustrate the process of obtaining the label value of each category in the area corresponding to the standby identification frame: Assume that there are three categories of electrical graphics elements in total, and the one-hot encodings of these three categories of electrical graphics elements are [1,0,0], [0,1,0] and [0,0,1] respectively. The area where each electrical graphics element in the secondary schematic drawing is located is obtained according to the label. If the area where the second category of electrical graphics elements is located exists in the standby identification frame, the label value of the first category in the area corresponding to the standby identification frame is 0, the label value of the second category in the area corresponding to the standby identification frame is 1, and the label value of the third category in the area corresponding to the standby identification frame is 0.

[0040] It should be further added that the process of obtaining the category recognition error of the spare identification frame is as follows: according to the label, the category of the graphic element in the area of ​​the spare identification frame is obtained, which is recorded as the reference graphic element category; the probability that the area in the spare identification frame is the reference graphic element category is obtained, and the category recognition error of the spare identification frame is obtained by subtracting the label value from the probability. The method for obtaining the probability that the area in the spare identification frame is various electrical graphic elements is a prior art and will not be described in detail here.

[0041] Understandably, It reflects the category recognition error of the kth spare recognition box in the sth secondary schematic drawing image. The larger the value, the more difficult it is for the recognition network to recognize the type of the element in the spare recognition box. It reflects the recognition error of the recognition network for the jth electrical element. It reflects the recognition error of the recognition network for the i-th electrical element. The closer it is to 1, the more similar the error rates of the two electrical elements are, which further indicates that the recognition errors of the ith or jth electrical element are caused by misidentification of each other. For example, the area in the spare identification box is the ith electrical element, where the probability of the area in the spare identification box being identified as the ith electrical element is 0.8, and the probability of being identified as the jth electrical element is 0.2. The recognition errors of these two electrical elements are both 0.2, which means that the jth electrical element interferes with the recognition result of the ith electrical element, resulting in a larger error. The output result corresponding to the kth alternative recognition box in the sth secondary schematic drawing image reflects the difficulty of distinguishing the jth electrical element from the ith electrical element. The larger the value, the greater the difficulty of the recognition network in identifying the element category in the kth alternative recognition box in the sth secondary schematic drawing image, and the high recognition difficulty is caused by the difficulty in distinguishing the jth electrical element from the ith electrical element, which further proves the difficulty of distinguishing the jth electrical element from the ith electrical element.

[0042] S3: Obtain a difference normalized image of each two electrical element regions in the secondary schematic drawing image; and perform weighted calculation on the difference normalized image using the difficulty of distinction as a weight to obtain an attention image.

[0043] It should be noted that in order to judge the extraction of information for distinguishing electrical elements of different classification difficulties by each network parameter, it is necessary to first analyze which information should be focused on for different types of electrical element distinctions, and then determine the accuracy of the electrical element distinction information extracted by each network parameter based on the correlation between the information extracted by the network parameters and the information that should be paid attention to in the electrical element distinction.

[0044] S30: Obtaining a difference normalized image of each two electrical graphic element regions in the secondary schematic drawing image.

[0045] Preferably, as an example, obtaining a difference normalized image of each two electrical graphic element regions in the secondary schematic drawing image includes: Extract each electrical element area in the secondary schematic drawing image according to the label; Adjust all electrical graphic element areas to the same size; and perform subtraction of every two electrical graphic element areas after the size adjustment to obtain a difference image; The pixels in the difference image are normalized to obtain a difference normalized image of each two electrical graphic element regions.

[0046] It can be understood that the difference normalized image can reflect the distinguishing information of the two electrical graphic element regions.

[0047] S31: Taking the difficulty of distinction as a weight, weighted calculation is performed on the difference normalized image to obtain an attention image.

[0048] Preferably, as an example, taking the difficulty of distinction as a weight, weighted calculation is performed on the difference normalized image to obtain the attention image, including: Taking the difficulty of distinction as a weight, the weighted sum of the difference normalized images between any electrical graphic element area and all other electrical graphic element areas is performed to obtain an initial attention image; Obtain outer envelope pixels of an electrical primitive in an electrical primitive region; The values ​​of the outer envelope pixels in the initial attention image are adjusted to 1 to obtain the attention image.

[0049] It should be noted that the outer envelope pixels of the electrical primitive are the outermost circle of pixels of the electrical primitive.

[0050] It is understandable that the greater the difficulty of distinction, the greater the difficulty of recognition network. The greater the difficulty of recognition, the more information the recognition network should absorb in this aspect to accurately recognize, so the weight of information with high difficulty in distinction is increased. By weighting the distinguishing information, the attention information of electrical graphic element distinction can be obtained. In addition, graphic element recognition not only needs to identify the type, but also accurately locate the graphic element area. The outer contour of the graphic element is the main information basis for locating the graphic element area, so the outer contour of the graphic element also needs to be focused on, so the attention of the outer contour pixels of the graphic element should be increased.

[0051] S4: Use the attention image to extract the key information of the secondary schematic drawing image; set the weight of the network parameters in the regularization term according to the correlation between the feature image and the key information; so as to realize the intelligent recognition of the secondary schematic drawing of the substation.

[0052] S40: extracting key information of the secondary schematic drawing image using the attention image.

[0053] Preferably, as an example, using the attention image to extract key information of the secondary schematic drawing image includes: The key information of the secondary schematic drawing image is obtained by multiplying each electrical graphic element area in the secondary schematic drawing image with the corresponding pixel of the attention image.

[0054] It can be understood that the key information of the secondary schematic drawing image is important information for distinguishing the types of electrical graphic elements.

[0055] S41: Setting the weight of the network parameters in the regularization term according to the correlation between the feature image and the key information.

[0056] It should be noted that in order to determine the accuracy of information extraction of each network parameter, it is necessary to determine the consistency between the extracted information and the information that should be focused on, and then determine the importance of each network parameter based on the consistency, and then set regularization weights for the network parameters based on the importance.

[0057] Preferably, as an example, the weight of the network parameter in the regularization term is set according to the correlation between the feature image and the key information, including: Obtain all feature images corresponding to each network parameter, and calculate the correlation between the feature image and the key information; take the inverse of the mean of the correlation obtained from all feature images of each network parameter to obtain the weight of each network parameter; The electrical graphic element recognition network of the first stage after the new round of training is completed is recorded as the electrical graphic element recognition network of the second stage; The weights are used as the weights of the corresponding network parameters in the regularization term to construct the regularization term of the loss function of the second-stage electrical element recognition network.

[0058] It should be supplemented that the method for calculating the correlation between the feature image and the key information includes: adjusting the feature image into an image of the same size as the key information through up- and down-sampling processing, splicing the rows of the resized feature image together to obtain a feature sequence, splicing the rows of the key information together to obtain a key information sequence, and taking the absolute value of the Pearson correlation between the feature sequence and the key information sequence as the correlation between the feature image and the key information.

[0059] It should be further supplemented that this embodiment constructs the regularization term of the loss function based on L2 regularization, and other embodiments may construct the loss function based on other regularization methods, which is not specifically limited in this embodiment.

[0060] It should be noted that when the weights are known, the method for constructing the regularization term is a prior art and will not be described in detail here.

[0061] S42: To realize intelligent recognition of substation secondary schematic drawings.

[0062] Preferably, as an example, to realize intelligent recognition of substation secondary schematic drawings, the method includes: Based on the loss function after adding the regularization term, the electrical graph element recognition network of the second stage is continuously trained using the data set until the preset training completion cutoff condition is met, and the finally trained electrical graph element recognition network is obtained; this embodiment uses the convergence of the loss value as the preset training completion cutoff condition, and other embodiments may adopt other conditions, which are not specifically limited in this embodiment.

[0063] The newly acquired secondary schematic drawing image is input into the finally trained electrical graphic element recognition network to obtain the recognition result.

[0064] At this point, this embodiment is completed.

[0065] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A deep learning-based intelligent recognition method for substation secondary schematic drawings, characterized in that: include: Obtain a data set, the data set consisting of labeled secondary schematic drawing images; Acquire a first-stage electrical element recognition network obtained in advance; Use the secondary schematic image to train the electrical element recognition network in the first stage, obtain the output results and the feature images of each convolution layer; calculate the difficulty of distinguishing the two electrical elements , the identification frame containing any one of the two electrical diagram elements in the output result is recorded as a spare identification frame, They represent the probability that the area within the kth spare identification box in the sth secondary schematic drawing image is the jth electrical element and the ith electrical element, respectively. represents the category recognition error of the kth alternative recognition box in the sth secondary schematic drawing image, They represent the label values ​​of the j-th category and the i-th category of the area corresponding to the k-th backup recognition box in the s-th secondary schematic drawing image, respectively. represents the number of spare recognition boxes in the s-th secondary schematic drawing image, Indicates the number of secondary schematic drawings images, Indicates the preset anti-zero coefficient. represents the linear normalization function; Obtain a difference normalized image of each two electrical element regions in the secondary schematic drawing image; perform weighted calculation on the difference normalized image using the difficulty of distinction as a weight to obtain an attention image; The focus image is used to extract the key information of the secondary schematic drawing image; The weights of network parameters in the regularization term are set according to the correlation between the feature image and the key information to achieve intelligent recognition.

2. The method for intelligent identification of secondary schematic drawings of substations based on deep learning according to claim 1 is characterized in that: The step of obtaining the first-stage electrical graphic element recognition network obtained in advance includes: A pre-built electrical graphic element recognition network is obtained; and a round of training is performed on the electrical graphic element recognition network using the data set to obtain a first-stage electrical graphic element recognition network.

3. The method for intelligent identification of secondary schematic drawings of substations based on deep learning according to claim 1 is characterized in that: The step of obtaining a difference normalized image of each two electrical graphic element regions in the secondary schematic drawing image includes: Extract each electrical element area in the secondary schematic drawing image according to the label; Adjust all electrical graphic element areas to the same size; and perform subtraction of every two electrical graphic element areas after the size adjustment to obtain a difference image; The pixels in the difference image are normalized to obtain a difference normalized image of each two electrical graphic element regions.

4. The method for intelligent identification of secondary schematic drawings of substations based on deep learning according to claim 1 is characterized in that: The method of performing weighted calculation on the difference normalized image to obtain the attention image by taking the difficulty of distinction as the weight includes: Taking the difficulty of distinction as a weight, the weighted sum of the difference normalized images between any electrical graphic element area and all other electrical graphic element areas is performed to obtain an initial attention image; Obtain outer envelope pixels of an electrical primitive in an electrical primitive region; The values ​​of the outer envelope pixels in the initial attention image are adjusted to 1 to obtain the attention image.

5. The method for intelligent identification of secondary schematic drawings of substations based on deep learning according to claim 1 is characterized in that: The method of extracting key information of the secondary schematic drawing image by using the attention image includes: The key information of the secondary schematic drawing image is obtained by multiplying each electrical element area in the secondary schematic drawing image by the attention image.

6. The method for intelligent identification of secondary schematic drawings of substations based on deep learning according to claim 1 is characterized in that: The step of setting the weight of the network parameter in the regularization term according to the correlation between the feature image and the key information includes: All feature images corresponding to each network parameter are obtained, and the correlation between the feature image and the key information is calculated; the weight of each network parameter is obtained by taking the inverse of the mean of the correlation obtained from all feature images of each network parameter, and the first-stage electrical element recognition network after one round of training is recorded as the second-stage electrical element recognition network, and the regularization term of the loss function of the second-stage electrical element recognition network is constructed based on the weight.

7. The method for intelligent identification of secondary schematic drawings of substations based on deep learning according to claim 6 is characterized in that: The regularization term adopts L2 regularization.

8. The method for intelligently identifying secondary schematic drawings of substations based on deep learning according to claim 6 is characterized in that: The intelligent identification is realized by: Based on the loss function after adding the regularization term, the electrical graph element recognition network of the second stage is continuously trained using the data set until the preset training completion cutoff condition is met, thereby obtaining the electrical graph element recognition network that has been finally trained. The newly acquired secondary schematic drawing image is input into the finally trained electrical graphic element recognition network to obtain the recognition result.

Images