Method and system for optimizing and improving intelligent operation inspection image recognition algorithm of transformer substation

Abstract

The invention discloses a method and a system for optimizing and improving an intelligent operation inspection image recognition algorithm of a transformer substation. The method comprises the following steps: establishing a training and evaluation environment for the optimized intelligent operation inspection image recognition algorithm of the transformer substation; performing performance evaluation on the optimized transformer substation intelligent operation inspection image recognition algorithm according to a given transformer substation intelligent operation inspection image test set, and if the performance evaluation does not meet requirements, expanding a given transformer substation intelligent operation inspection image training set based on an artificial intelligence generation content technology; retraining the optimized transformer substation intelligent operation inspection image recognition algorithm by using the expanded transformer substation intelligent operation inspection image training set; and deploying a transformer substation intelligent operation inspection image recognition algorithm. The invention aims to solve the problem of sample confidentiality requirement by carrying out model training from a transformer substation intranet to an algorithm management platform, realize incremental training of a transformer substation intelligent operation inspection image recognition algorithm on the premise of sample confidentiality, and improve the effect of incremental training under the condition of few samples.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent operation and inspection of power systems, and specifically relates to a method and system for optimizing and improving an image recognition algorithm for intelligent operation and inspection of a substation. Background Art

[0002] The intelligent operation and inspection technology of substations is based on artificial intelligence image recognition models. By collecting visible light images of equipment in the station for analysis, it can realize the autonomous identification of equipment appearance defects. At present, there are many intelligent operation and inspection R&D manufacturers in China, and there are many intelligent operation and inspection algorithm models on the market. However, in the actual application of substations, they generally perform poorly, with high false detection rates and low correct detection rates. In addition to inconsistent models, different algorithm manufacturers have inconsistent training sets, different image sources, and uneven annotation quality. As a user, power supply companies have limited means of controlling algorithm models. The existing model capability improvement method is mainly for algorithm manufacturers to collect typical false positive and false negative samples on the user side, add them to the original training set, and retrain. This method is not suitable for the power industry. The first reason is the confidentiality requirements of the power industry data. The confidentiality requirements of the user side for substation defect images make it difficult to obtain actual pictures on site; the second reason is that the defects of actual substations in operation are scarce and the scenes are single, which cannot meet the requirements of the number of samples for model training. The number of newly added samples is too small to have a practical impact on the training results. Summary of the invention

[0003] Technical problem to be solved by the present invention: In view of the above-mentioned problems in the prior art, a method and system for optimizing and improving the image recognition algorithm for intelligent operation and maintenance of substations is provided. The present invention aims to carry out model training from the substation intranet to the algorithm management platform to solve the problem of sample confidentiality requirements, realize incremental training of the image recognition algorithm for intelligent operation and maintenance of substations under the premise of sample confidentiality, and how to improve the effect of incremental training when there are very few samples.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for optimizing and improving a substation intelligent operation and inspection image recognition algorithm comprises the following steps:

[0006] S1, build a training and evaluation environment for the optimized substation intelligent operation and inspection image recognition algorithm;

[0007] S2, perform performance evaluation on the optimized substation intelligent operation and inspection image recognition algorithm according to the given substation intelligent operation and inspection image test set, and if the performance evaluation meets the requirements, jump to step S4; otherwise, jump to step S3;

[0008] S3, expanding the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology, using the expanded substation intelligent operation and inspection image training set to retrain the optimized substation intelligent operation and inspection image recognition algorithm, and jumping to step S2;

[0009] S4, deploy substation intelligent operation and maintenance image recognition algorithm for substation intelligent operation and maintenance.

[0010] Optionally, step S2 includes:

[0011] S2.1, respectively inputting image samples in a given substation intelligent operation and inspection image test set of various scenarios into the optimized substation intelligent operation and inspection image recognition algorithm to obtain corresponding target detection results;

[0012] S2.2, calculate the detection rate and false detection rate in each scenario based on the detection box in the target detection result and the label box in the given substation intelligent operation and inspection image sample;

[0013] S2.3, determine whether the detection rate and false detection rate of all scenes meet the requirements. If they do, jump to step S4 if the performance evaluation meets the requirements; otherwise, jump to step S3.

[0014] Optionally, the calculation function expression of the detection rate in step S2.2 is:

[0015] D R =M1 / T,

[0016] In the above formula, D R is the detection rate, M1 is the number of correctly identified detection boxes in the target detection result in this scenario, and T is the number of labeled boxes in the given substation intelligent operation and inspection image test set in this scenario.

[0017] Optionally, the calculation function expression of the false positive rate in step S2.2 is:

[0018] FP R =(M2-m1) / M2,

[0019] In the above formula, FP R is the false detection rate, M1 is the number of correctly identified detection frames in the target detection results in this scenario, and M2 is the total number of detection frames in the target detection results in this scenario.

[0020] Optionally, the calculation function expression of the false positive rate in step S2.2 is:

[0021] FP R =(M2 / (T1+T2)*(T1+a*T2)-M1) / (M2 / (T1+T2)*(T1+aT2)),

[0022] In the above formula, FP R is the false positive rate, M2 is the total number of detection frames in the target detection results in this scenario, T1 is the number of positive samples in this scenario in the given substation intelligent operation and inspection image test set, T2 is the number of negative samples in this scenario in the given substation intelligent operation and inspection image test set, a is the ratio of positive and negative samples of the label in the given substation intelligent operation and inspection image test set, and M1 is the number of correctly identified detection frames in the target detection results in this scenario.

[0023] Optionally, after inputting the image samples in the given substation intelligent operation and inspection image test set of multiple scenarios into the optimized substation intelligent operation and inspection image recognition algorithm to obtain the corresponding target detection results in step S2.1, it also includes a double-blind review mode for the substation intelligent operation and inspection images with false detection and missed detection: first, the annotation boxes of the substation intelligent operation and inspection images with false detection and missed detection are removed; then two operation and maintenance personnel perform a secondary annotation on the same sample to generate an annotation box, and if the annotation boxes annotated by the two people are consistent, the jointly annotated annotation box is used as the basis for re-inputting the optimized substation intelligent operation and inspection image recognition algorithm to obtain the corresponding target detection results; if the annotation boxes annotated by the two people are inconsistent, and the annotation box annotated by one of them is consistent with the detection box in the target detection result, it is considered that the substation intelligent operation and inspection image is not missed or detected, and the detection result is corrected.

[0024] Optionally, in step S3, expanding a given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology includes:

[0025] S3.1, obtaining a scenario where the detection rate or false detection rate in a given substation intelligent operation and inspection image test set does not meet the requirements;

[0026] S3.2. For each scenario where the detection rate or false detection rate does not meet the requirements, determine the substation object category in the scenario. If the detection rate of a substation object category in the scenario does not meet the requirements, then expand the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of negative samples of the substation object category in the scenario, and the increased number of negative samples is proportional to the difference between the detection rate of the substation object category in the scenario and the detection rate of the substation object category that meets the requirements; if the false detection rate of a substation object category does not meet the requirements, then expand the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of positive samples of the substation object category in the scenario, and the increased number of positive samples is proportional to the difference between the detection rate of the substation object category in the scenario and the detection rate of the substation object category that meets the requirements.

[0027] In addition, the present invention also provides a substation intelligent operation and inspection image recognition algorithm optimization and improvement system, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method.

[0028] In addition, the present invention also provides a computer-readable storage medium, which stores a computer program or instruction, and the computer program or instruction is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method through a processor.

[0029] In addition, the present invention also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method through a processor.

[0030] Compared with the prior art, the present invention mainly has the following advantages: the method of the present invention includes building a training and evaluation environment for an optimized substation intelligent operation and inspection image recognition algorithm; performing a performance evaluation on the optimized substation intelligent operation and inspection image recognition algorithm according to a given substation intelligent operation and inspection image test set, and if the performance evaluation does not meet the requirements, expanding the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology, and retraining the optimized substation intelligent operation and inspection image recognition algorithm using the expanded substation intelligent operation and inspection image training set; deploying the substation intelligent operation and inspection image recognition algorithm, the present invention solves the problem of sample confidentiality requirements by carrying out model training from the substation intranet to the algorithm management platform, and can realize incremental training of the substation intelligent operation and inspection image recognition algorithm under the premise of sample confidentiality, thereby being able to improve the effect of incremental training when there are very few samples, and can improve the training efficiency and recognition accuracy of the substation intelligent operation and inspection image recognition algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the basic flow of the method of the embodiment of the present invention.

[0032] Figure 2 It is a curve diagram of the detection rate in the embodiment of the present invention.

[0033] Figure 3 4 is a curve diagram of the false detection rate in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm of the present invention aims to increase the number of samples through AIGC technology, and solve the problem of sample confidentiality requirements by connecting the user substation intranet to the algorithm management platform for model training. In order to enable personnel in this technical field to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0035] like Figure 1 As shown, the method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm in this embodiment includes the following steps:

[0036] S1, build a training and evaluation environment for the optimized substation intelligent operation and inspection image recognition algorithm;

[0037] S2, perform performance evaluation on the optimized substation intelligent operation and inspection image recognition algorithm according to the given substation intelligent operation and inspection image test set, and if the performance evaluation meets the requirements, jump to step S4; otherwise, jump to step S3;

[0038] S3, expanding the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology, using the expanded substation intelligent operation and inspection image training set to retrain the optimized substation intelligent operation and inspection image recognition algorithm, and jumping to step S2;

[0039] S4, deploy substation intelligent operation and maintenance image recognition algorithm for substation intelligent operation and maintenance.

[0040] The purpose of the substation intelligent operation and inspection image recognition algorithm is to identify the substation object category in the substation intelligent operation and inspection image according to the input substation intelligent operation and inspection image, including transformers, current transformers, voltage transformers, lightning arresters, transmission lines, towers, insulators, wire clamps, switches, pins, connectors, meters, insulators, etc. As an optional implementation, the substation intelligent operation and inspection image recognition algorithm in this embodiment is YOLOv8, and other substation intelligent operation and inspection image recognition algorithms can also be used as needed.

[0041] In this embodiment, step S2 includes:

[0042] S2.1, respectively inputting image samples in a given substation intelligent operation and inspection image test set of various scenarios into the optimized substation intelligent operation and inspection image recognition algorithm to obtain corresponding target detection results;

[0043] S2.2, calculate the detection rate and false detection rate in each scenario based on the detection box in the target detection result and the label box in the given substation intelligent operation and inspection image sample;

[0044] S2.3, determine whether the detection rate and false detection rate of all scenarios meet the requirements. If they meet the requirements, it is determined that if the performance evaluation meets the requirements, jump to step S4; otherwise, jump to step S3. Qualified basis According to the application of the substation, in this embodiment, the detection rate and false detection rate meet the requirements. The detection rate is greater than the preset threshold of 80%, and the false detection rate is less than the preset threshold of 20%. Undoubtedly, the standard of meeting the requirements here can be determined according to actual needs.

[0045] In this embodiment, the calculation function expression of the detection rate in step S2.2 is:

[0046] D R =M1 / T,

[0047] In the above formula, D R is the detection rate, M1 is the number of correctly identified detection boxes in the target detection result in this scenario, and T is the number of labeled boxes in the given substation intelligent operation and inspection image test set in this scenario.

[0048] In this embodiment, the calculation function expression of the false positive rate in step S2.2 is:

[0049] FP R =(M2-m1) / M2,

[0050] In the above formula, FP R is the false detection rate, M1 is the number of correctly identified detection frames in the target detection results in this scenario, and M2 is the total number of detection frames in the target detection results in this scenario.

[0051] In this embodiment, the calculation function expression of the false positive rate in step S2.2 is:

[0052] FP R =(M2 / (T1+T2)*(T1+a*T2)-M1) / (M2 / (T1+T2)*(T1+aT2)),

[0053] In the above formula, FP R is the false positive rate, M2 is the total number of detection frames in the target detection results in this scenario, T1 is the number of positive samples in this scenario in the given substation intelligent operation and inspection image test set, T2 is the number of negative samples in this scenario in the given substation intelligent operation and inspection image test set, a is the ratio of positive and negative samples of the label in the given substation intelligent operation and inspection image test set, and M1 is the number of correctly identified detection frames in the target detection results in this scenario.

[0054] In order to avoid data inconsistency in the annotation of multiple samples, a multi-person double-blind correction is carried out to improve the accuracy of model evaluation. In step S2.1 of this embodiment, after inputting the image samples in the given substation intelligent operation and inspection image test set of multiple scenes into the optimized substation intelligent operation and inspection image recognition algorithm to obtain the corresponding target detection results, it also includes a double-blind review mode for the substation intelligent operation and inspection images that are misdetected and missed: first, the annotation box of the substation intelligent operation and inspection image that is misdetected and missed is removed; then two operation and maintenance personnel perform a secondary annotation on the same sample to generate an annotation box. If the annotation boxes annotated by the two people are consistent, the annotation box annotated by the two people is used as the basis for re-inputting the optimized substation intelligent operation and inspection image recognition algorithm to obtain the corresponding target detection result; if the annotation boxes annotated by the two people are inconsistent, and the annotation box annotated by one of them is consistent with the detection box in the target detection result, it is considered that the substation intelligent operation and inspection image is not missed or misdetected, and the detection result is corrected.

[0055] When the given substation intelligent operation and inspection image training set is expanded based on the artificial intelligence generated content (AIGC) technology in step S3 of this embodiment, the training set is expanded based on the AIGC technology, and the unqualified scenes should be further subdivided before expansion, and the unqualified classes after subdivision are expanded in a targeted manner. Specifically, the given substation intelligent operation and inspection image training set is expanded based on the artificial intelligence generated content technology in step S3 of this embodiment, including:

[0056] S3.1, obtaining a scenario where the detection rate or false detection rate in a given substation intelligent operation and inspection image test set does not meet the requirements;

[0057] S3.2. For each scenario where the detection rate or false detection rate does not meet the requirements, determine the substation object category in the scenario. If the detection rate of a substation object category in the scenario does not meet the requirements, then expand the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of negative samples of the substation object category in the scenario, and the increased number of negative samples is proportional to the difference between the detection rate of the substation object category in the scenario and the detection rate of the substation object category that meets the requirements; if the false detection rate of a substation object category does not meet the requirements, then expand the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of positive samples of the substation object category in the scenario, and the increased number of positive samples is proportional to the difference between the detection rate of the substation object category in the scenario and the detection rate of the substation object category that meets the requirements. For example, in a certain scenario P, the substation object category is divided into three categories, namely, category 1, category 2, and category 3. The detection rate and false positive rate of the subdivided scenarios of category 1, category 2, and category 3 are further calculated under scenario P to further determine whether each category meets the index requirements after subdivision. If category 1 meets the requirements, and category 2 and category 3 do not meet the requirements, the samples of category 2 and category 3 are expanded in a targeted manner. If the detection rate does not meet the requirements, the number of negative samples of category 2 and category 3 is increased in a targeted manner, and the number of increases is proportional to the difference between the detection rate of category 2 and category 3 and the detection rate of category 1. If the false positive rate does not meet the requirements, the number of positive samples of category 2 and category 3 is increased in a targeted manner, and the number of increases is proportional to the difference between the false positive rate of category 2 and category 3 and the false positive rate of category 1.

[0058] In this embodiment, expanding a given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of negative samples / positive samples of the substation object category in the scenario includes: obtaining a given substation intelligent operation and inspection image training set to increase the negative samples and positive samples of the substation object category in the scenario; respectively inputting the negative samples and positive samples into a large language model that supports artificial intelligence content generation technology (such as GPT-4o, ChatGPT, StableDiffusion, Gemini 1.5Pro, etc.), and giving a prompt: Please generate diversified negative samples / positive samples based on the given negative samples and positive samples, and the diversified methods include perspective changes, composition changes, light changes, style and texture changes, proportion and size changes, stretching and rotation changes, image expansion changes, local masking and blurring. Add the negative samples / positive samples generated by the large language model to the given substation intelligent operation and inspection image training set, so as to increase the number of negative samples / positive samples of the substation object category in the scenario.

[0059] In order to verify the effectiveness of the optimization and improvement method of the substation intelligent operation and inspection image recognition algorithm in this embodiment, the optimization and improvement method of the substation intelligent operation and inspection image recognition algorithm in this embodiment is used to optimize YOLOv6, YOLOv7, and YOLOv8 respectively in this embodiment, and the curves of the detection rate and the false detection rate changing with time are shown as follows: Figure 2 and Figure 3 As shown, Figure 2 and Figure 3 The horizontal axis is weeks, and the vertical axis is the detection rate or false positive rate. "Series 1" represents YOLOv6, "Series 2" represents YOLOv7, and "Series 3" represents YOLOv8. Figure 2 and Figure 3 It can be seen that after optimizing YOLOv6, YOLOv7, and YOLOv8 using the optimization and improvement method of the substation intelligent operation and inspection image recognition algorithm of this embodiment, the detection rate of YOLOv6, YOLOv7, and YOLOv8 three substation intelligent operation and inspection image recognition algorithms is increased by more than 20% on average, and the false detection rate is reduced by more than 20% on average. It can be seen that the optimization and improvement method of the substation intelligent operation and inspection image recognition algorithm of this embodiment carries out model training from the substation intranet to the algorithm management platform to solve the problem of sample confidentiality requirements, and can realize the incremental training of the substation intelligent operation and inspection image recognition algorithm under the premise of sample confidentiality, thereby improving the effect of incremental training when there are few samples, and can improve the training efficiency and recognition accuracy of the substation intelligent operation and inspection image recognition algorithm.

[0060] In addition, this embodiment also provides a substation intelligent operation and inspection image recognition algorithm optimization and improvement system, including an interconnected microprocessor and a memory, and the microprocessor is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method.

[0061] In addition, this embodiment also provides a computer-readable storage medium, which stores a computer program or instruction, and the computer program or instruction is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method through a processor.

[0062] In addition, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method through a processor.

[0063] Those skilled in the art will appreciate that the technical solutions provided by the embodiments of the present application may be in the form of methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0064] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for optimizing and improving the image recognition algorithm for intelligent operation and inspection of substations, characterized in that: The steps include: S1, build a training and evaluation environment for the optimized substation intelligent operation and inspection image recognition algorithm; S2, performing a performance evaluation on the optimized substation intelligent operation and inspection image recognition algorithm according to a given substation intelligent operation and inspection image test set, and jumping to step S4 if the performance evaluation meets the requirements; Otherwise, jump to step S3; S3, expanding the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology, using the expanded substation intelligent operation and inspection image training set to retrain the optimized substation intelligent operation and inspection image recognition algorithm, and jumping to step S2; S4, deploy substation intelligent operation and maintenance image recognition algorithm for substation intelligent operation and maintenance.

2. The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm according to claim 1 is characterized in that: Step S2 includes: S2.1, respectively inputting image samples in a given substation intelligent operation and inspection image test set of various scenarios into the optimized substation intelligent operation and inspection image recognition algorithm to obtain corresponding target detection results; S2.2, calculate the detection rate and false detection rate in each scenario based on the detection box in the target detection result and the label box in the given substation intelligent operation and inspection image sample; S2.3, determine whether the detection rate and false detection rate of all scenes meet the requirements. If they do, jump to step S4 if the performance evaluation meets the requirements; otherwise, jump to step S3.

3. The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm according to claim 2 is characterized in that: The calculation function expression of the detection rate in step S2.2 is: D R =M1 / T, In the above formula, D R is the detection rate, M1 is the number of correctly identified detection boxes in the target detection result in this scenario, and T is the number of labeled boxes in the given substation intelligent operation and inspection image test set in this scenario.

4. The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm according to claim 2 is characterized in that: The calculation function expression of the false positive rate in step S2.2 is: FP R =(M2-M1) / M2, In the above formula, FP R is the false detection rate, M1 is the number of correctly identified detection frames in the target detection results in this scenario, and M2 is the total number of detection frames in the target detection results in this scenario.

5. The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm according to claim 2 is characterized in that: The calculation function expression of the false positive rate in step S2.2 is: <h2 style=";text-align:left;direction:ltr">FP<h2 style=";text-align:left;direction:ltr"> R <h2 style=";text-align:left;direction:ltr"> (M2 / (T1+T2)*(T1+a*T2)-M1) / (M2 / (T1+T2)*(T1+aT2))) In the above formula, FP R is the false positive rate, M2 is the total number of detection frames in the target detection results in this scenario, T1 is the number of positive samples in this scenario in the given substation intelligent operation and inspection image test set, T2 is the number of negative samples in this scenario in the given substation intelligent operation and inspection image test set, a is the ratio of positive and negative samples of the label in the given substation intelligent operation and inspection image test set, and M1 is the number of correctly identified detection frames in the target detection results in this scenario.

6. The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm according to claim 2 is characterized in that: After inputting the image samples in the given substation intelligent operation and inspection image test set of various scenes into the optimized substation intelligent operation and inspection image recognition algorithm to obtain the corresponding target detection results in step S2.1, it also includes a double-blind review mode for the substation intelligent operation and inspection images with false detection and missed detection to conduct a secondary review: first, the annotation boxes of the substation intelligent operation and inspection images with false detection and missed detection are removed; Then, two operation and maintenance personnel will perform a second annotation on the same sample to generate an annotation box. If the annotation boxes marked by the two people are consistent, the commonly marked annotation box will be used as the basis to re-enter the optimized substation intelligent operation and maintenance image recognition algorithm to obtain the corresponding target detection result; if the annotation boxes marked by the two people are inconsistent, and the annotation box marked by one of them is consistent with the detection box in the target detection result, it is considered that the substation intelligent operation and maintenance image is not missed or misdetected, and the detection result is corrected.

7. The method for optimizing and improving the substation intelligent operation and inspection image recognition algorithm according to claim 2 is characterized in that: In step S3, the given substation intelligent operation and inspection image training set is expanded based on artificial intelligence content generation technology, including: S3.1, obtaining a scenario where the detection rate or false detection rate in a given substation intelligent operation and inspection image test set does not meet the requirements; S3.

2. For each scenario where the detection rate or false detection rate does not meet the requirements, determine the substation object category in the scenario. If the detection rate of a substation object category in the scenario does not meet the requirements, then expand the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of negative samples of the substation object category in the scenario, and the increased number of negative samples is proportional to the difference between the detection rate of the substation object category in the scenario and the detection rate of the substation object category that meets the requirements; if the false detection rate of a substation object category does not meet the requirements, then expand the given substation intelligent operation and inspection image training set based on artificial intelligence content generation technology to increase the number of positive samples of the substation object category in the scenario, and the increased number of positive samples is proportional to the difference between the detection rate of the substation object category in the scenario and the detection rate of the substation object category that meets the requirements.

8. A substation intelligent operation and inspection image recognition algorithm optimization and improvement system, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method described in any one of claims 1 to 7 through a processor.

10. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the substation intelligent operation and inspection image recognition algorithm optimization and improvement method described in any one of claims 1 to 7 through a processor.

Images