Intelligent electric energy metering box state identification system

Abstract

The application discloses a kind of intelligent electric energy metering box state identification systems, the present application relates to image data processing technical field, including image acquisition module, image quality analysis module, semantic analysis module, associated feature module and state identification module;Image quality analysis module analyzes image quality, obtains image quality score, and sets quality threshold, determines whether image is qualified, state identification module is combined with environmental parameters, semantic feature credibility and abnormal association score to identify the state of electric energy meter, with the unqualified image of key feature loss filtered by image quality analysis, combined with temperature and humidity quantitative semantic feature credibility attenuation amplitude, avoid misjudgment caused by environmental interference;Associated feature module quantifies physical connection strength and historical fault correlation degree, identifies real-time anomaly and implicit risk, reduces missed judgment, and state identification module multi-dimensional weighted grading positioning, subdivides abnormal type, and greatly improves operation efficiency.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology, specifically to an intelligent power metering box status recognition system. Background Technology

[0002] A metering box is a comprehensive set of metering instruments and auxiliary equipment necessary for measuring electrical energy, including electricity meters, voltage and current transformers and their secondary circuits, electricity metering panels, cabinets, and boxes. Metering boxes can be installed in a suspended or floor-mounted manner. To ensure installation quality, suspended boxes must be secured using internal expansion bolts, with bolt dimensions meeting safety and secure requirements for outdoor installation. In the context of the ongoing development of smart grids, smart electricity metering boxes, as key terminal equipment, play a crucial role in enabling remote operation and maintenance and automated inspection by identifying the status of components such as electricity meters within them.

[0003] Because the lighting conditions in the deployment scenarios of smart energy metering boxes vary greatly, and internal components such as energy meters are prone to reflection, the current standard for judging quality is usually based solely on image clarity. This fails to quantify the impact of lighting deviation, reflection level, and environmental temperature and humidity factors on the integrity of image features. This can easily lead to misjudgments where the clarity meets the standard but key features have been lost. Furthermore, the impact of related components on key components such as energy meters is not considered, which can easily result in missed judgments, leading to untimely identification of abnormal states and increasing safety risks. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent power metering box status identification system, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent power metering box status recognition system, comprising an image acquisition module, an image quality analysis module, a semantic analysis module, an association feature module, and a status recognition module;

[0006] The image acquisition module acquires image data from different locations of the electricity meter, including images of the display screen, wiring terminals, and sealing device;

[0007] The image quality analysis module extracts image information from the image data, including sharpness, reflectivity and light intensity, analyzes the image quality, obtains an image quality score, sets a quality threshold, determines whether the image is qualified, and re-acquires image data if it is unqualified.

[0008] The semantic analysis module acquires environmental parameters and compares them with their standard values ​​to analyze the interference of the environment on feature recognition and obtain the credibility of semantic features.

[0009] The correlation feature module quantifies the abnormal impact of the correlation components on the electricity meter based on the real-time physical state and historical fault data of the correlation components, and obtains an abnormal correlation score.

[0010] The status recognition module evaluates the abnormal status of the electricity meter by combining environmental parameters, semantic feature credibility, and abnormal correlation score to obtain a comprehensive status abnormality score. Based on the comprehensive status abnormality score and the value range of each component data, the module identifies the status of the electricity meter.

[0011] Optionally, the image quality analysis module first sets a brightness threshold, compares the pixels in the image data with the brightness threshold to obtain reflective bright pixels, then calculates the ratio of reflective bright pixels to the total number of pixels to obtain the reflectivity, and introduces corresponding weights to obtain a reflectivity suppression term. Next, it calculates the deviation between the illumination intensity and the ideal illumination intensity, and introduces corresponding weights to obtain an illumination adaptation correction term. Finally, it combines the image sharpness, the reflectivity suppression term, and the illumination adaptation correction term to obtain the image quality score F. imq ;

[0012] When the image quality score F imq A score ≥0.7 indicates that the image data is acceptable. When the image quality score F... imq If the value is less than 0.7, it indicates that the image data is unqualified. In this case, the image acquisition module will re-acquire the image. If the image data fails to meet the requirements three times in a row, the administrator will be notified to conduct a manual inspection.

[0013] Optionally, the semantic analysis module determines the image quality score based on the image quality score F. imq The credibility of the original semantic features under ideal conditions is corrected to obtain the credibility of the basic semantic features;

[0014] Next, acquire ambient temperature and humidity data, calculate the difference between the ambient temperature and humidity data and their corresponding working thresholds, and adjust the correction strength of the basic semantic feature credibility based on the degree to which the ambient temperature and humidity exceed the working thresholds, ultimately obtaining the semantic feature credibility S. cor .

[0015] Optionally, after the associated feature module obtains the real-time physical status and historical fault data of the associated components of the energy meter, it first calculates the physical connection strength, calculates the conduction reliability by the deviation ratio between the measured circuit resistance and the standard resistance, obtains the screw offset by image recognition, calculates the ratio of the screw offset to the maximum allowable offset to obtain the terminal tightness, and adds the conduction reliability and terminal tightness by weight to obtain the physical connection strength value.

[0016] Then, the number of shared failures of related components and the electricity meter over a period of time is counted, and the ratio of this number to the total number of failures of related components over the same period is calculated to obtain the correlation score. The physical connection strength score and the correlation score are then weighted and added together to obtain the abnormal correlation score S. rel .

[0017] Optionally, the state recognition module obtains the environmental anomaly degree S by calculating the maximum percentage of environmental temperature and humidity exceeding the working threshold. env The semantic feature credibility S cor Abnormal correlation score S rel and environmental anomaly S env The weighted sum is used to obtain the comprehensive score A for abnormal state. tot ;

[0018] The abnormal state comprehensive score was A. tot Next, set a state division threshold to divide it into different states. The process is as follows:

[0019] When the overall score for abnormal status is A tot A value less than 0.3 is considered normal.

[0020] When 0.3 ≤ Comprehensive score for abnormal status A tot When the value is less than 0.7, it is in a warning state;

[0021] When the overall score for abnormal status is A tot A value ≥0.7 indicates an abnormal state.

[0022] Optionally, in the warning state, the minimum value of the warning state range is 0.3, when the semantic feature confidence level S cor If the percentage exceeds 50%, it is considered a semantic feature warning.

[0023] When the abnormal association score S rel When the proportion exceeds 40%, it is judged as a warning for related components;

[0024] Environmental anomaly S env If the proportion exceeds 30%, it is judged as an environmental warning, and corresponding measures are taken according to different warning states;

[0025] Under abnormal conditions, the minimum value of the abnormal condition range is 0.7, when the semantic feature credibility S cor When the percentage exceeds 60%, it is judged as semantic feature abnormality;

[0026] When the abnormal association score S rel and environmental anomaly S env When the total percentage exceeds 50%, it is judged as a system-level anomaly, and corresponding measures are taken according to different anomaly states.

[0027] Optionally, the image acquisition module has a built-in angle calibration unit that automatically takes a preset image of the reference mark inside the box every 24 hours, and automatically triggers a correction reminder when the offset exceeds the set value.

[0028] Optionally, when acquiring images from the display screen, the image acquisition module activates a polarizing filter and combines it with RGB fill light, continuously acquiring three high-definition images. At the same time, it eliminates local shadows through a multi-frame fusion algorithm. When acquiring images of the wiring terminals and sealing devices, it activates infrared fill light.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] I. This invention quantifies the impact of illumination deviation and reflectivity on the integrity of image features through image quality analysis, and filters out unqualified images to avoid situations where the image has acceptable clarity but key features are lost. Furthermore, the semantic analysis module quantifies the correction strength of semantic feature credibility when temperature and humidity exceed thresholds. The more severe the environment, the greater the attenuation of semantic feature credibility, reflecting the impact of the environment on the accuracy of state recognition. This ensures that feature judgment under environmental interference is more in line with the actual scenario and avoids misjudgment caused by environmental deviation.

[0031] Second, the correlation feature module of this invention quantifies the real-time physical connection strength by analyzing the reliability of secondary circuit conduction and the tightness of wiring terminals. Then, it quantifies the historical fault correlation by analyzing the proportion of the number of simultaneous faults of the related components to the total number of faults in the electricity meter. Finally, the two are combined to obtain an abnormal correlation score, which can identify the abnormal state of the real-time physical connection and discover hidden risks through the historical fault patterns, identify abnormal states in a timely manner, avoid missed judgments, and reduce safety hazards.

[0032] Third, this invention combines semantic feature credibility, anomaly correlation score, and environmental anomaly through a state recognition module, and divides the results into normal, warning, and abnormal states by setting a state division threshold. At the same time, for warning and abnormal states, it further divides them according to the proportion of the scores of the constituent data, including semantic feature warning, associated component warning, environmental warning, semantic feature anomaly, and system-level anomaly, so that operation and maintenance management personnel can carry out targeted maintenance without having to conduct a comprehensive investigation every time, thus improving operation and maintenance efficiency. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the system flow of the present invention;

[0034] Figure 2 This is a state recognition block diagram of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] For examples, please refer to Figure 1 and Figure 2 This embodiment provides an intelligent power meter box status recognition system, including an image acquisition module, an image quality analysis module, a semantic analysis module, an association feature module, and a status recognition module;

[0037] The image acquisition module acquires image data from different locations of the electricity meter, including images of the display screen, wiring terminals, and sealing devices. The image acquisition module has a built-in angle calibration unit that automatically takes a preset image of the reference mark inside the box every 24 hours. When the deviation exceeds the set value by more than 5°, a correction reminder is automatically triggered to avoid the image acquisition device from shifting after long-term use and affecting the accuracy of the electricity meter box status identification.

[0038] The image acquisition module includes image acquisition devices installed inside the electricity metering box at different angles to acquire image data from the electricity meter at different angles. The installation positions of the angles are as follows:

[0039] Viewpoint 1 is located directly in front of the display screen, at a 0° angle to the plane of the electricity meter display screen, to suppress glass reflection and to capture images of the display screen;

[0040] The second viewpoint is located diagonally above the left terminal block, at a 30° angle to the terminal block plane, allowing for a clear view of the terminal block's secure position.

[0041] The third viewpoint is located directly above the sealing device, at a 0° angle to the sealing plane, and can fully capture the sealing texture and number;

[0042] When acquiring images from the display screen, the image acquisition module activates the polarizing filter and combines it with RGB fill light to continuously acquire three high-definition images. It then uses a multi-frame fusion algorithm to eliminate local shadows. When acquiring images of the wiring terminals and sealing devices, it activates infrared fill light to avoid visible light interference with the reflections of the metal terminals.

[0043] The image quality analysis module extracts image information from the image data, including sharpness, reflectivity, and illumination intensity, analyzes the image quality, and obtains an image quality score. The process is as follows:

[0044]

[0045] In the above formula, F imqImage quality score quantifies the effective information content of the image. The larger the value, the clearer the image features, which meets the requirements for energy meter status recognition.

[0046] C represents image sharpness, which is the mean gradient of the image edges calculated based on the Laplacian edge detection operator. It is used to quantify the basic detail retention capability of the image and is the basis for energy meter digital recognition, seal texture matching, and terminal status judgment. The higher the value, the richer the edge details and the higher the image quality.

[0047] w1 is the sharpness impact coefficient, with a value of 0.5. Image sharpness has the highest priority and is the basis for image feature recognition.

[0048] L represents the actual light intensity, reflecting the actual lighting environment during image acquisition;

[0049] L opt The optimal light intensity for identifying the characteristics of the electricity meter is determined by on-site testing and serves as a reference value for light adaptability.

[0050] w2 is the illumination adaptability influence coefficient, with a value of 0.3. This is a lighting adaptation correction term, applied when the actual lighting intensity L differs from the ideal lighting intensity L. opt When the absolute difference becomes increasingly large, it indicates that the ambient light is too dark or too bright, making it difficult to identify image features and reducing the image quality score F. imq The lighting conditions in the installation scenarios of electricity meters vary greatly. Traditional assessments that rely solely on clarity are prone to misjudgment. For example, if the clarity is sufficient but the lighting is too dark or too bright, the image may be misjudged as qualified even though key features have been lost. By setting this option, the impact of ambient light fluctuations on image quality can be quantified, and feature loss problems caused by excessive darkness or brightness can be identified in a timely manner.

[0051] R h To determine the number of bright pixels, a brightness threshold is set, and pixels in key areas of the image whose brightness exceeds the brightness threshold are identified as reflective bright pixels.

[0052] R all Total number of pixels, referring to the total number of pixels in the key region of the image. Reflectivity;

[0053] w3 is the reflection impact factor, with a value of 0.2. The sharpness impact factor w1, the lighting adaptability impact factor w2, and the reflection impact factor w3 can be adjusted according to the actual situation, but their sum is 1. This is a reflection suppression item, quantifying the degree of reflection interference from different components to eliminate feature blurring caused by material reflection. Excessive reflectivity can obscure the numbers on the electricity meter display, conceal loose screw marks on wiring terminals, and blur the texture of serial numbers due to reflections from sealing plastic. Higher reflectivity results in a lower score for this item, thus reducing the image quality score (F). imq The impact of reflections on image quality is fully considered.

[0054] Obtain the image quality score F imq Then, the quality threshold was set to 0.7, and when the image quality score F... imq A score ≥0.7 indicates that the image data is acceptable. When the image quality score F... imq When the value is less than 0.7, it indicates that the image data is unqualified. At this time, the image acquisition module will re-acquire the image. If the image fails three times in a row, the administrator will conduct a manual inspection to check whether there is a problem with the image acquisition equipment or whether the installation angle is off. This is done by quantifying the impact of illumination deviation and reflectivity on the integrity of image features, so as to avoid the situation where the image is clear but key features are lost.

[0055] The semantic analysis module acquires environmental parameters and compares them with their standard values ​​to analyze the interference of the environment on feature recognition, thereby obtaining the semantic feature credibility. The process is as follows:

[0056]

[0057] In the above formula, S cor The semantic feature credibility is such that the closer the value is to 1, the higher the credibility of image feature recognition and the better the condition of the electricity meter.

[0058] S raw Based on the credibility of semantic features, S raw =S raw,ori ×(0.7+0.3×F imq The image quality score F represents the credibility of the original semantic features after image quality correction. imq The higher the value, the higher the accuracy of semantic feature recognition;

[0059] S raw,ori The original semantic feature credibility refers to the original recognition accuracy of the semantic features of the electricity meter image under ideal environmental conditions, without considering the image acquisition quality. For example, the accuracy of recognizing digital readings and backlight status in the electricity meter display image, and the accuracy of recognizing seal texture, number and edge fit in the sealing device image.

[0060] T represents the actual temperature of the electricity meter.

[0061] T thr This refers to the normal operating temperature threshold for the electricity meter.

[0062] ΔT max The maximum allowable over-temperature value, i.e. the range within which the actual temperature is allowed to exceed the threshold, is set to 10℃;

[0063] k1 is the temperature influence coefficient, with a value of 0.5, representing the degree of impact of the electricity meter temperature exceeding the threshold on the credibility of image semantic features. This is a temperature over-threshold correction term, which quantifies the impact of temperature exceeding the normal threshold on the credibility of semantic features. The numerator of this term is set to max(0, T-T). thr The effect is only calculated when the temperature exceeds the threshold, avoiding meaningless adjustments at normal temperatures;

[0064] H represents the actual humidity level measured by the electricity meter.

[0065] H thr This is the humidity threshold for normal operation of the electricity meter. Exceeding this threshold can easily lead to condensation on the display screen and blurring of the seal texture.

[0066] ΔH max The maximum allowable humidity level, i.e. the range within which the actual humidity is allowed to exceed the threshold, is set at 10%RH.

[0067] k2 is the humidity influence coefficient, with a value of 0.5. The sum of the temperature influence coefficient k1 and the humidity influence coefficient k2 is 1. These values ​​can be adjusted according to the specific environment. For example, in high-temperature areas, the temperature influence coefficient k1 can be increased; in high-humidity areas, the humidity influence coefficient k2 can be increased to improve the applicability of this system. This is a humidity over-threshold correction term, which quantifies the effect of humidity exceeding the normal threshold on the semantic feature credibility S. cor The degree of influence is determined by setting the molecular part of this term to max(0,H-H). thr The calculation is only initiated when the humidity exceeds the standard. By converting the degree of excessive humidity into a relative proportion, the degree of interference to image feature recognition when the humidity exceeds the threshold is calculated.

[0068] The semantic analysis module quantifies the credibility S of semantic features when temperature and humidity exceed thresholds. cor The correction strength reflects the impact of the environment on the accuracy of electricity meter box identification; the harsher the environment, the lower the semantic feature reliability S. cor The greater the attenuation, the more it affects the subsequent state recognition results. This ensures that the feature judgment under environmental interference is more in line with the actual scene and avoids misjudgment caused by environmental deviation.

[0069] The correlation feature module quantifies the abnormal impact of the correlation components on the electricity meter based on the real-time physical status and historical fault data of the related components, and obtains an abnormal correlation score. The specific process is as follows:

[0070]

[0071] In the above formula, S rel The score represents the correlation score; the higher the score, the greater the impact of the correlated component on the energy meter, and the higher the probability that the correlated component and the energy meter will both malfunction simultaneously.

[0072] β1 is the influence coefficient of physical connection strength, with a value of 0.6. Physical connection strength is a key factor affecting real-time performance, and its value is higher than that of historical fault factors.

[0073] The physical connection strength term represents the physical connection strength between the associated component and the energy meter, quantifies the current hard connection reliability between the associated component and the energy meter, and reflects the impact of real-time physical status on the energy meter.

[0074] γ is the conduction reliability influence coefficient, with a value of 0.7. Conduction is the basis for the operation of the energy meter and has a higher priority than terminal tightness.

[0075] The tightness of the terminal block is given by d, where d is the screw offset. max The larger the screw offset d is, the lower the quality of the terminal fastening.

[0076] R con The reliability of the secondary circuit is scored, with a higher score indicating more reliable conduction. The process is as follows:

[0077]

[0078] Where R act This represents the actual resistance of the secondary circuit.

[0079] R min Design the minimum resistance for the secondary circuit;

[0080] R max Design the maximum allowable resistance for the secondary circuit, when the actual resistance R of the secondary circuit is... act Minimum resistance R in secondary circuit design min When the circuit has no additional contact resistance, the continuity reliability is 1, and the minimum design resistance R of the secondary circuit is 1. min < Actual resistance R of the secondary circuit act <Maximum allowable resistance R in secondary circuit design max When the resistance increases, it indicates that the reliability of conduction decreases linearly, reflecting the severity of problems such as poor contact and wire aging. When the actual resistance R of the secondary circuit... act >Maximum allowable resistance R in secondary circuit design max When the circuit is open, it indicates that there is a fault such as a broken wire or a severely loose terminal, and the continuity reliability is directly set to 0.

[0081] β2 is the historical fault impact coefficient, with a value of 0.4, which reflects the impact of faults in the electricity meter's associated components on the electricity meter in historical data;

[0082] N co This refers to the number of co-occurrence faults between related components and the electricity meter, indicating that when the related component malfunctions, the electricity meter also malfunctions. It is obtained by analyzing one year's worth of fault records extracted from historical data.

[0083] N all This represents the total number of failures of the associated components;

[0084] This section defines historical fault correlation items, quantifying the probability of meter malfunction when associated components are abnormal. It reflects the co-occurrence pattern of faults in associated components and meters. By analyzing the co-occurrence patterns of meters and associated components in historical faults of the electricity metering box, the blind spots of real-time physical detection can be compensated for. For example, latent contact problems may not be detected in real-time, but historical data can reflect their probability. If the real-time physical connection is normal but the historical correlation is high (e.g., the tightness of the transformer terminals is normal, but historical data shows that a loose transformer of that model is highly likely to cause meter malfunction), then the fault correlation score S is calculated. rel The score will remain high, enabling timely identification of abnormal states, avoiding missed diagnoses, and reducing safety hazards. If a real-time connection anomaly occurs but historical correlation is low (e.g., a slight loosening of the transformer terminals), but historically such situations have a low probability of causing meter malfunctions, the anomaly correlation score S will be low. rel This will reduce and avoid over-amplifying the influence of related components, decrease misjudgments, and improve the accuracy of the results data.

[0085] The status recognition module assesses the abnormal status of the electricity meter by combining environmental parameters, semantic feature credibility, and anomaly correlation score, obtaining a comprehensive status anomaly score. Based on the comprehensive status anomaly score and the value range of each component data, the module identifies the status of the electricity meter to achieve intelligent status recognition. The specific process is as follows:

[0086]

[0087] In the above formula, A tot The overall score for abnormal status is as follows: the higher the value, the higher the degree of abnormality of the electricity meter.

[0088] θ1 is the semantic feature anomaly influence coefficient, with a value of 0.5. Semantic features are the key factor in judging the anomaly of the electricity meter, so they have the highest weight.

[0089] S cor The credibility of semantic features can directly reflect the explicit anomalies of the electricity meter itself;

[0090] θ2 is the abnormal correlation influence coefficient, with a value of 0.3. Abnormalities in related components have an indirect influence and are therefore weighted less.

[0091] S rel For anomaly correlation scoring, the indirect impact of anomalies in correlated components on the electricity meter is quantified. For example, a loose connection in the current transformer among the correlated components of the electricity meter, coupled with a high historical correlation, will result in an anomaly correlation score of S. rel Even if the electricity meter itself has no obvious abnormalities, this will increase the overall abnormality score (A). tot ;

[0092] θ3 is the environmental anomaly impact coefficient, with a value of 0.2. Environmental anomalies represent long-term potential impacts and have the lowest weight.

[0093] S env The environmental anomaly level is indicated by a higher value, signifying a more severe environment. This quantifies the long-term impact of environmental anomalies on the electricity meter and provides early warnings of potential hidden faults caused by environmental factors. For example, if the electricity meter is exposed to high temperatures for an extended period, even if the display shows normal digit recognition, the overall anomaly score will be A. tot This will also increase, thereby adjusting subsequent state recognition results to achieve early warning effects, environmental anomaly degree S env The process of obtaining the result is as follows:

[0094]

[0095] T represents the actual temperature of the electricity meter.

[0096] T thr This refers to the normal operating temperature threshold for the electricity meter.

[0097] ΔT max This is the maximum allowable over-temperature value, i.e., the range within which the actual temperature is allowed to exceed the threshold.

[0098] H represents the actual humidity level measured by the electricity meter.

[0099] H thr This is the humidity threshold for normal operation of the electricity meter. Exceeding this threshold can easily lead to condensation on the display screen and blurring of the seal texture.

[0100] ΔH max This is the maximum allowable humidity level, i.e., the range within which the actual humidity is allowed to exceed the threshold.

[0101] The abnormal state comprehensive score was A. tot Subsequently, based on experimental testing, a state division threshold was set to divide the state into different states, as follows:

[0102] When the overall score for abnormal status is A tot When the value is less than 0.3, it is considered normal, indicating that the electricity meter itself, its associated components, and the environment are all within a safe range, with no obvious or hidden abnormal risks.

[0103] When 0.3 ≤ Comprehensive score for abnormal status A tot When the value is less than 0.7, it is a warning state. At this time, it is further divided according to the value of each component data:

[0104] Taking the lowest value of 0.3 in the warning state range as an example, under the warning state, such as the semantic feature confidence S cor The proportion exceeds 50%, that is, θ1×(1-S) cor When the value is greater than 0.15, it is judged as a semantic feature warning, indicating that there is a problem with the electricity meter itself. At this time, the image recognition result can be remotely reviewed, or the electricity meter display screen and seal integrity can be manually checked on-site.

[0105] Under alert status, the abnormal correlation score S rel The proportion exceeds 40%, that is, θ2×S rel When the value is greater than 0.12, it is determined to be a warning for related components. At this time, the physical connection status of related components, such as current transformers and secondary circuits, should be checked.

[0106] Under warning conditions, the environmental anomaly degree S env The proportion exceeds 30%, that is, θ3×S env When the value is greater than 0.09, an environmental warning is issued, indicating that the ambient temperature or humidity is too high. At this time, check the ventilation and waterproofing of the electricity meter box; adjust the position of the temperature and humidity sensor and confirm the accuracy of the data; if the environment is abnormal for a long period of time, install ventilation and heat dissipation or dehumidification devices.

[0107] When the overall score for abnormal status is A tot A value ≥ 0.7 indicates an abnormal state. Taking the lowest value of the abnormal state range as 0.7 as an example, if the semantic feature credibility S is ≥ 0.7 at this time... cor The proportion exceeds 60%, that is, θ1×(1-S) cor If the value is greater than 0.42, it is judged as semantic feature abnormality, indicating that there is a serious problem with the electricity meter itself. The power should be cut off immediately and the electricity meter should be inspected. The damaged display screen should be replaced in time.

[0108] If the environmental anomaly degree S env Score of association with anomalies rel The sum of these proportions exceeds 50%, that is, (θ2×S) rel )+(θ3×S env When the value is greater than 0.42, it is determined to be a system-level anomaly. At this time, the related components of the electricity meter and the environment are abnormal. An emergency power outage is required to check the related components and deal with the environmental anomaly, such as abnormal temperature rise or water ingress into the electricity metering box.

[0109] The semantic feature credibility S is determined by the state recognition module. cor Abnormal correlation score S rel and environmental anomaly S envThe results are weighted and combined, and divided into normal, warning, and abnormal states. Furthermore, for warning and abnormal states, they are further subdivided according to the proportion of the scores of the constituent data to obtain semantic feature warning, associated component warning, environmental warning, semantic feature abnormality, and system-level abnormality. This eliminates the need for manual differentiation and achieves the effect of intelligent status recognition of electricity meters. This allows operation and maintenance personnel to carry out targeted maintenance without having to conduct a comprehensive inspection every time, thus improving operation and maintenance efficiency.

[0110] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart electricity metering box status identification system, characterized in that, It includes an image acquisition module, an image quality analysis module, a semantic analysis module, a correlation feature module, and a state recognition module; The image acquisition module acquires image data from different locations of the electricity meter, including images of the display screen, wiring terminals, and sealing device; The image quality analysis module extracts image information from the image data, including sharpness, reflectivity and light intensity, analyzes the image quality, obtains an image quality score, sets a quality threshold, determines whether the image is qualified, and re-acquires image data if it is unqualified. The semantic analysis module acquires environmental parameters and compares them with their standard values ​​to analyze the interference of the environment on feature recognition and obtain the credibility of semantic features. The correlation feature module quantifies the abnormal impact of the correlation components on the electricity meter based on the real-time physical state and historical fault data of the correlation components, and obtains an abnormal correlation score. The status recognition module evaluates the abnormal status of the electricity meter by combining environmental parameters, semantic feature credibility, and abnormal correlation score to obtain a comprehensive status abnormality score. Based on the comprehensive status abnormality score and the value range of each component data, the module identifies the status of the electricity meter.

2. The intelligent power metering box status identification system according to claim 1, characterized in that: The image quality analysis module first sets a brightness threshold, compares the pixels in the image data with the brightness threshold to obtain reflective high-brightness pixels, then calculates the ratio of reflective high-brightness pixels to the total number of pixels to obtain the reflectivity, and introduces corresponding weights to obtain a reflectivity suppression term. Next, it calculates the deviation between the illumination intensity and the ideal illumination intensity, introduces corresponding weights to obtain an illumination adaptation correction term, and finally combines the image sharpness, reflectivity suppression term, and illumination adaptation correction term to obtain the image quality score F. imq ; When the image quality score F imq A score ≥0.7 indicates that the image data is acceptable. When the image quality score F... imq If the value is less than 0.7, it indicates that the image data is unqualified. In this case, the image acquisition module will re-acquire the image. If the image data fails to meet the requirements three times in a row, the administrator will be notified to conduct a manual inspection.

3. The intelligent power metering box status identification system according to claim 2, characterized in that: The semantic analysis module is based on the image quality score F. imq The credibility of the original semantic features under ideal conditions is corrected to obtain the credibility of the basic semantic features; Next, acquire ambient temperature and humidity data, calculate the difference between the ambient temperature and humidity data and their corresponding working thresholds, and adjust the correction strength of the basic semantic feature credibility based on the degree to which the ambient temperature and humidity exceed the working thresholds, ultimately obtaining the semantic feature credibility S. cor .

4. The intelligent power metering box status identification system according to claim 3, characterized in that: After the associated feature module obtains the real-time physical status and historical fault data of the associated components of the energy meter, it first calculates the physical connection strength, calculates the conduction reliability by the deviation ratio between the measured circuit resistance and the standard resistance, obtains the screw offset by image recognition, calculates the ratio of the screw offset to the maximum allowable offset to obtain the terminal tightness, and adds the conduction reliability and terminal tightness by weight to obtain the physical connection strength value. Then, the number of shared failures of related components and the electricity meter over a period of time is counted, and the ratio of this number to the total number of failures of related components over the same period is calculated to obtain the correlation score. The physical connection strength score and the correlation score are then weighted and added together to obtain the abnormal correlation score S. rel .

5. The intelligent power metering box status identification system according to claim 4, characterized in that: The state recognition module calculates the maximum percentage of environmental temperature and humidity exceeding the working threshold to obtain the environmental anomaly degree S. env The semantic feature credibility S cor Abnormal correlation score S rel and environmental anomaly S env The weighted sum is used to obtain the comprehensive score A for abnormal state. tot ; The abnormal state comprehensive score was A. tot Next, set a state division threshold to divide it into different states. The process is as follows: When the overall score for abnormal status is A tot A value less than 0.3 is considered normal. When 0.3 ≤ Comprehensive score for abnormal status A tot When the value is less than 0.7, it is in a warning state; When the overall score for abnormal status is A tot A value ≥0.7 indicates an abnormal state.

6. The intelligent power metering box status identification system according to claim 5, characterized in that: In the warning state, the minimum value of the warning state range is 0.3, when the semantic feature credibility S cor If the percentage exceeds 50%, it is considered a semantic feature warning. When the abnormal association score S rel When the proportion exceeds 40%, it is judged as a warning for related components; When the environmental anomaly degree S env When the proportion exceeds 30%, it is determined to be an environmental warning, and corresponding measures are taken according to different warning states; Under abnormal conditions, the minimum value of the abnormal condition range is 0.7, when the semantic feature credibility S cor When the percentage exceeds 60%, it is judged as semantic feature abnormality; When the abnormal association score S rel and environmental anomaly S env When the total percentage exceeds 50%, it is judged as a system-level anomaly, and corresponding measures are taken according to different anomaly states.

7. The intelligent power metering box status identification system according to claim 1, characterized in that: The image acquisition module has a built-in angle calibration unit that automatically takes a preset image of the reference mark inside the box every 24 hours. When the deviation exceeds the set value, a correction reminder is automatically triggered.

8. The intelligent power metering box status identification system according to claim 1, characterized in that: When acquiring images from the display screen, the image acquisition module activates a polarizing filter and combines it with RGB fill light, continuously acquiring three high-definition images. At the same time, it eliminates local shadows through a multi-frame fusion algorithm. When acquiring images of the wiring terminals and sealing devices, it activates infrared fill light.

Images