Method and system for detecting electrical performance of switch cabinet based on machine vision

Abstract

The invention provides a method and system for detecting the electrical performance of a switch cabinet based on machine vision, and the method comprises the steps: obtaining an image of a to-be-detected switch cabinet, the image at least comprising visible elements in the switch cabinet and the connection state between the elements; carrying out feature extraction on the image and carrying out element consistency detection based on the extracted features; after the consistency detection is passed, carrying out secondary loop detection by adopting a knowledge graph; and after the secondary loop detection is passed, current loop consistency detection and control loop consistency detection are respectively carried out. According to the invention, based on image acquisition and identification processing of machine vision and in combination with logic association, loop detection and detection sequence control, process detection of the switch cabinet can be realized, and the purpose of efficient and accurate detection on comprehensive performance of element installation, physical connection and electrical connection can be ensured.

Description

Technical Field

[0001] The present invention belongs to the field of physical detection technology, and in particular relates to a method and system for detecting electrical performance of a switch cabinet based on machine vision. Background Art

[0002] Switchgear is a type of electrical equipment widely used in power systems. It primarily houses various control switch components and the cables connecting them. During the switchgear manufacturing process, testing its electrical performance and wiring is crucial.

[0003] Generally speaking, it relies on manual visual inspection to see whether the model specifications in the cabinet are consistent with the design drawings, and to check whether the connection marks of the secondary component wire ends are correct, complete, clear and firm. Then, the auxiliary circuit power-on test is carried out to detect the electrical function of the switch cabinet.

[0004] However, traditional manual inspection has many disadvantages. On the one hand, manual inspection of thread ends is tedious and requires a lot of human resources. Manual operation is prone to fatigue and negligence, making it difficult to ensure the accuracy of inspection. On the other hand, the entire inspection process is time-consuming, seriously affecting production efficiency and increasing production costs.

[0005] As the power system's requirements for switchgear quality and production efficiency continue to increase, there is an urgent need for an efficient and accurate detection system to replace the traditional manual detection method. Summary of the Invention

[0006] To solve the above technical problems, this application proposes a method and system for detecting the electrical performance of switchgear based on machine vision. The specific technical solutions are as follows:

[0007] In a first aspect, the present application provides a method for detecting electrical performance of a switchgear based on machine vision, the method comprising:

[0008] Acquire an image of the switchgear to be inspected, the image including at least visible components inside the switchgear and connection status between the components;

[0009] Extract features from the image and perform component consistency detection based on the extracted features;

[0010] After the consistency test is passed, the knowledge graph is used to perform secondary loop detection;

[0011] After the secondary circuit test is passed, the current circuit consistency test and the control circuit consistency test are carried out respectively.

[0012] As a preferred embodiment of the above solution, voice prompts are further provided, and the voice prompts include status prompts corresponding to different detection steps, that is, prompts of different detection results in each status prompt.

[0013] As a preferred embodiment of the above solution, a method for extracting features from an image and performing component consistency detection based on the extracted features is as follows:

[0014] The characteristic Euclidean distance between the captured image and the standard component image is calculated, and a threshold condition is set. If the calculation result meets the set threshold condition, it is considered to meet the component consistency requirements.

[0015] As a preferred alternative to the above solution, the method for secondary loop detection using knowledge graph is:

[0016] Generate a relational database with component codes, models, component names, terminal numbers and associated relationships based on standardized secondary schematics, and generate a detection rule base;

[0017] According to the detection rules in the detection rule library, the components, terminal numbers and wire connection relationships in the corresponding detection rules are searched in the acquired image to determine whether the connection of the secondary circuit is correct.

[0018] As a preferred embodiment of the above solution, a method for generating a relational database based on a standardized secondary schematic diagram is as follows:

[0019] Separate component graphics from circuit graphics in the schematic diagram through image segmentation algorithm;

[0020] Identify component information using an optical character recognition algorithm, wherein the component information includes component code, model, component name, and terminal number;

[0021] Based on graph theory, components and terminal numbers are used as nodes and lines as edges to construct an electrical connection diagram and obtain the connection relationship between components;

[0022] A relational database is generated based on component information and the connection relationships between components.

[0023] As a preferred embodiment of the above solution, after acquiring the image, the image is subjected to image fusion processing, and / or histogram equalization processing, and / or denoising processing, and / or grayscale processing.

[0024] As a preferred embodiment of the above solution, the method for performing current loop consistency detection and control loop consistency detection is:

[0025] Power on the switch cabinet, obtain the three-phase current and phase data of the cable loop in the switch cabinet, and compare the phase difference and current amplitude. If they are the same, the current loop consistency requirement is met;

[0026] Power on the switch cabinet, obtain the voltage value of the key node in the switch cabinet, and compare it with the standard voltage value. If the deviation is within the allowable range, it meets the control loop consistency requirements.

[0027] In a second aspect, the present application provides a system for detecting electrical performance of a switchgear based on machine vision, characterized in that the system is applied to the above-mentioned method, and the system includes:

[0028] An image acquisition module, used to acquire images of the switch cabinet to be inspected;

[0029] Detection module, used to perform component consistency detection, secondary circuit detection, current circuit consistency detection and control circuit consistency detection;

[0030] and data storage modules.

[0031] As a preferred embodiment of the above scheme, a voice prompt module is further included, and the voice prompt module is set with different prompt states, which at least include an initial state, a state of checking components installed in the cabinet, and a state of detecting a comprehensive protection current loop.

[0032] The beneficial effects of the present invention are:

[0033] This application is based on image acquisition and recognition processing based on machine vision, and combines logical association, loop detection, and detection sequence control to realize process-based detection of switch cabinets, ensuring efficient and accurate detection of the comprehensive performance of component installation, physical connection, and electrical connection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The figure shows a flow chart of a method for inspecting electrical performance of a switchgear based on machine vision. DETAILED DESCRIPTION

[0035] In the following description, certain specific details are set forth in order to provide a thorough understanding of the various embodiments. However, it will be understood by those skilled in the art that the present invention can be practiced without these details. In other cases, well-known structures are not shown or described in detail to avoid unnecessarily obscuring the description of the embodiments. Unless the context requires otherwise, throughout the specification and the appended claims, the word "comprising" should be interpreted in an open, inclusive sense, that is, as in "including but not limited to."

[0036] See also Figure 1 The method for detecting the electrical performance of a switch cabinet based on machine vision in this application includes the following steps:

[0037] (1) Acquire images;

[0038] (2) Carry out component consistency detection, secondary circuit detection, current circuit consistency detection and control circuit consistency detection in sequence.

[0039] For the image collected in the above step (1), it is necessary to identify the components and the corresponding terminal numbers in the image. A convolutional neural network (CNN) can be used to extract features from the image to further identify the components and the corresponding terminal numbers.

[0040] CNNs automatically learn key features in component images by building multiple convolutional and pooling layers, such as feature vectors of component shape, terminal locations, and the number of wires connected to each terminal. Taking the VGG-16 network as an example, it performs convolution operations using a series of 3×3 convolutional kernels to continuously extract local image features. Pooling layers are then used to reduce the resolution of the feature maps, minimizing computational effort.

[0041] In step (1), the collected images can also be pre-processed, mainly including image fusion processing and image quality processing. Among them, image fusion processing fuses images from multiple angles into a more comprehensive component image to overcome the incomplete shape problem or poor accuracy problem of image collection under different angles and lighting conditions; image quality processing uses methods such as histogram equalization to pre-process the collected images, enhance the image contrast, make the component features more obvious, and facilitate subsequent feature extraction and recognition.

[0042] In addition, after step (1) and before step 2, the image should be denoised and grayscaled to facilitate subsequent feature extraction and recognition.

[0043] Before proceeding to step (2), a standard component image that matches the component consistency test should be provided for identification of the matching component. Specifically, the captured image is matched with the features of the standard component image after feature extraction, and the similarity between the feature vectors is calculated using the Euclidean distance;

[0044]

[0045] Where x and y are the feature vectors of the captured image and the reference image, respectively, n is the dimension of the feature vector, and d(x, y) is the Euclidean distance between the two. A smaller distance indicates a higher degree of similarity. When the similarity meets the set threshold, the live image and the reference image are considered to match, and the corresponding component is identified.

[0046] In step (2), after completing the component consistency test, the secondary circuit test is performed, and the secondary circuit test is completed using the knowledge graph. First, a relational database storing component codes, models, component names, and terminal numbers generated based on the input standardized secondary schematic diagram should be provided, and the association relationship between them should be established. For example, for a relay component, its component code is K1, its model is JZC-23F, and its component name is intermediate relay. Its various terminal numbers (such as 1, 2, 3, etc.) and the corresponding functional descriptions and the connection relationships in the secondary schematic diagram are all stored in the database.

[0047] The input and parsing process for the standardized quadratic principle begins with an image segmentation algorithm to separate the component and circuit diagrams in the schematic. Then, optical character recognition (OCR) technology is used to identify textual information such as component codes, model numbers, component names, and terminal numbers. Connection relationships are determined by analyzing the path of the circuits and the connecting nodes. For example, a graph-theory-based approach is used, using component and terminal numbers as nodes and circuits as edges to construct an electrical connection diagram, clearly showing the connections between components.

[0048] When testing secondary circuits, knowledge graph technology is used to logically associate secondary principles, images, and physical component terminal numbers. The knowledge graph graphically displays the relationships between entities (such as components, terminal numbers, etc.). By building a rule base, such as "If terminal 1 of component A and terminal 2 of component B are connected in the secondary schematic diagram, and a wire connection is detected between the corresponding physical terminals during image recognition, then the connection is considered logical." Based on these rules, the system can perform logical reasoning to determine whether the secondary circuit connection is correct.

[0049] In step (2), the secondary circuit detection is the basic detection at the physical level. After completing the secondary circuit detection, it is necessary to perform current circuit consistency detection and control circuit consistency detection. This is a connection level detection to ensure that the switch cabinet can work normally.

[0050] For current loop consistency testing:

[0051] Current sensors are used to detect the magnitude and phase of current in a circuit. When there's a break in the circuit, the current is zero. When a wiring error causes current shunting or a phase anomaly, the current is compared with the standard current and phase values to determine the cause. For example, in a three-phase current circuit, the three phase currents are normally equal in magnitude and 120° out of phase. Assuming the three-phase currents are IAIBIC, then under normal circumstances:

[0052] |I A |=|I B |=|I C |

[0053] ∠I A -∠I B =120°

[0054] ∠I B -∠I C =120°

[0055] ∠I C -∠I A =120°

[0056] By measuring the actual current value and substituting it into the above formula for comparison, if it does not meet the requirements, it means that there is a problem in the current loop.

[0057] For control loop consistency detection:

[0058] Control loop testing primarily relies on detecting voltage signals within the loop. Different wiring states correspond to different voltage values within the control loop. For example, when a switch is closed, the voltage across it should be zero; when the switch is open, the voltage across it should be equal to the power supply voltage. Voltages at key nodes in the loop are measured and compared to standard voltage values. For example, using Ohm's law (U = IR), with a known resistance R and a measured current I, a theoretical voltage U is calculated. This is then compared to the actual measured voltage. If the deviation exceeds the allowable range, a wiring error or other fault is detected in the loop.

[0059] For current and control loop testing, hardware requires the proper selection of current and voltage sensors and ensuring their accurate installation. Software requires filtering of the collected current and voltage signals to remove noise. Algorithms such as Kalman filtering are used to predict and estimate signals, improving their stability and accuracy. When determining whether a circuit has a breakpoint or wiring error, multiple factors, such as current, voltage, and phase relationships, must be considered. By establishing a fault diagnosis model, such as one based on a Bayesian network, multiple detection parameters are input, and probabilistic reasoning is used to determine the presence of a circuit fault and the type of fault.

[0060] In addition, during the entire detection process of step (2), voice prompts are also set to assist in the smooth completion of the detection process. The state machine principle is used to define the visual recognition wiring function detection sequence. The state machine has multiple states, such as the initial state, the state of checking the components installed in the cabinet, the state of detecting the comprehensive protection current loop, etc., among which the state of detecting the comprehensive protection current loop includes the secondary circuit detection state, the current loop consistency detection state and the control loop consistency detection state recorded above. Each state corresponds to a detection step. When a detection step is completed and the result is correct, the state machine transfers to the next state; if the detection result is wrong, the corresponding error handling mechanism is triggered.

[0061] For example, in the initial state, the system is ready to start testing. When the test command is issued, the state machine transfers to the state of checking the components installed in the cabinet. At this time, the system starts to check the components and design Figure 1 Only when the test in this state passes (that is, all installed components are consistent with the design Figure 1 The state machine will transfer to the next state of detecting the comprehensive protection current loop.

[0062] In each of the above-mentioned states, voice prompts are provided, implemented using text-to-speech (TTS) technology. The system generates corresponding prompt text based on the test results, such as "Connection error in contact 1 of the current loop phase A measurement circuit on the integrated protection (ZB)." This text is then input into the TTS engine, which converts it into a voice signal based on a pre-trained voice model. The signal is then played through the speaker, prompting the operator to take appropriate action.

[0063] During the voice prompt stage, it is necessary to ensure that the prompt content is accurate, clear, and easy to understand. When generating prompt text, it is necessary to organize it reasonably based on the detailed information of the test results. For example, for prompts of wiring errors, it is necessary to clearly indicate the circuit where the error occurred, the component name, the terminal number, and the specific error type (such as loose wiring, incorrect wire number marking, etc.). When selecting a TTS engine, it is necessary to consider its speech synthesis quality, the types of languages supported, and its compatibility with the system. At the same time, different voice styles and speaking speeds can be set according to the needs of the testers to improve the effectiveness of the prompts.

[0064] Based on the above description, the present application further provides a system for detecting the electrical performance of a switchgear based on machine vision, the system comprising:

[0065] Image acquisition module, used to collect physical images of various components inside the switchgear;

[0066] The detection module is used to perform component consistency detection, secondary circuit detection, current circuit consistency detection and control circuit consistency detection.

[0067] The data storage module is used to store basic data such as images and standard schematics.

[0068] Among them, the detection module also includes a component consistency detection unit. The component consistency detection unit extracts features after reading the image, matches them with the features of the standard component image, and calculates the Euclidean distance to represent the similarity. The process includes three core steps: image feature extraction, Euclidean distance calculation and result comparison. It can be processed sequentially on a single line or in parallel on multiple lines.

[0069] The detection module also includes a secondary circuit detection unit, which detects the integrity and accuracy of the secondary circuit based on the rule base in the data storage module and the entity image obtained by the image acquisition module.

[0070] The rule base in the data storage module is established based on a relational database. The rule base can be stored in the data storage module through external input or generated by the system. When generated by the system, a file recognition and parsing module is required to identify and parse the standardized secondary schematics input into the system to generate a relational database with component codes, model numbers, component names, terminal numbers, and their respective relationships. The rule base is then established based on the relational database and stored in the data storage module.

[0071] The detection module also includes a comprehensive protection current loop detection unit, which performs current loop consistency detection and control loop consistency detection. This unit primarily determines whether there are problems with the current and control loops by comparing the electrical signal data within the loops. Therefore, the comprehensive protection current loop detection unit must work in conjunction with the current acquisition module and the voltage acquisition module, respectively. The current acquisition module typically uses a current sensor, while the voltage acquisition module typically uses a voltage sensor.

[0072] Based on the above system records, the system of this application is also equipped with a voice prompt module to assist in the voice prompt of the detection process. In this application, the voice prompt module is set with different prompt states, including at least the initial state, the state of checking the components installed in the cabinet, and the state of detecting the comprehensive protection current loop, which respectively represent different detection processes and detection purposes, and these states can only be switched in sequence, and cannot be adjusted arbitrarily between different states, that is, when a detection step is completed and the result is correct, the state machine transfers to the next state; if the detection result is wrong, the corresponding error handling mechanism is triggered.

[0073] The method and system of the present invention achieve efficient and accurate detection of switchgear secondary circuit wiring by integrating multiple principles, including image recognition, logical association, circuit detection, detection sequence control, and voice prompts. During implementation, the key technologies and implementation methods at each stage are analyzed in detail, and the corresponding calculation formulas and analysis procedures are provided. This system is highly innovative and practical and is expected to be widely used in the power industry, improving the safety and reliability of power systems.

[0074] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same.

Claims

1. A method for detecting electrical performance of a switchgear based on machine vision, characterized in that: The method is: Acquire an image of the switchgear to be inspected, the image including at least visible components inside the switchgear and connection status between the components; Extract features from the image and perform component consistency detection based on the extracted features; After the consistency test is passed, the knowledge graph is used to perform secondary loop detection; After the secondary circuit test is passed, the current circuit consistency test and the control circuit consistency test are carried out respectively.

2. The method for detecting electrical performance of a switch cabinet based on machine vision according to claim 1, characterized in that: A voice prompt is set, wherein the voice prompt includes status prompts corresponding to different detection steps, that is, prompts of different detection results in each status prompt.

3. The method for detecting electrical performance of a switch cabinet based on machine vision according to claim 1, characterized in that: The method for extracting features from an image and performing component consistency detection based on the extracted features is as follows: The characteristic Euclidean distance between the captured image and the standard component image is calculated, and a threshold condition is set. If the calculation result meets the set threshold condition, it is considered to meet the component consistency requirements.

4. The method for detecting electrical performance of a switch cabinet based on machine vision according to claim 1, characterized in that: The method of using knowledge graph for secondary circuit detection is: Generate a relational database with component codes, models, component names, terminal numbers and associated relationships based on standardized secondary schematics, and generate a detection rule base; According to the detection rules in the detection rule library, the components, terminal numbers and wire connection relationships in the corresponding detection rules are searched in the acquired image to determine whether the connection of the secondary circuit is correct.

5. The method for detecting electrical performance of a switch cabinet based on machine vision according to claim 1, characterized in that: The method for generating a relational database based on a standardized secondary schematic diagram is: Separate component graphics from circuit graphics in the schematic diagram through image segmentation algorithm; Identify component information using an optical character recognition algorithm, wherein the component information includes component code, model, component name, and terminal number; Based on graph theory, components and terminal numbers are used as nodes and lines as edges to construct an electrical connection diagram and obtain the connection relationship between components; A relational database is generated based on component information and the connection relationships between components.

6. The method for detecting electrical performance of a switch cabinet based on machine vision according to claim 1, characterized in that: After the image is acquired, image fusion processing, and / or histogram equalization processing, and / or denoising processing, and / or grayscale processing are performed on the image.

7. The method for detecting electrical performance of a switch cabinet based on machine vision according to claim 1, characterized in that: The method for performing current loop consistency detection and control loop consistency detection is: Power on the switch cabinet, obtain the three-phase current and phase data of the cable loop in the switch cabinet, and compare the phase difference and current amplitude. If they are the same, the current loop consistency requirement is met; Power on the switch cabinet, obtain the voltage value of the key node in the switch cabinet, and compare it with the standard voltage value. If the deviation is within the allowable range, it meets the control loop consistency requirements.

8. A system for detecting electrical performance of switchgear based on machine vision, characterized in that: The system is applied to the method described in any one of claims 1 to 7, and the system includes: An image acquisition module, used to acquire images of the switch cabinet to be inspected; Detection module, used to perform component consistency detection, secondary circuit detection, current circuit consistency detection and control circuit consistency detection; and data storage modules.

9. The system for detecting electrical performance of switchgear based on machine vision according to claim 8, characterized in that: It also includes a voice prompt module, which is set with different prompt states. The different prompt states at least include an initial state, a state of checking components installed in the cabinet, and a state of detecting a comprehensive protection current loop.

Images