Intelligent electric meter reading method and system

Abstract

The invention relates to an intelligent electric meter reading method and system, and the method comprises the steps: enabling a high-definition camera to shoot an intelligent electric meter according to a preset meter reading time interval, obtaining an electric meter image, and uploading the electric meter image to an edge computing device through the Internet of Things; the high-definition camera is installed right in front of a dial plate of the intelligent electric meter. The edge computing device carries out image preprocessing on the electric meter image and transmits the electric meter image after image preprocessing to a server; the server recognizes the received electric meter images by adopting a plurality of image recognition models based on a deep learning algorithm, compares and verifies the electric meter readings obtained by recognition of the plurality of image recognition models, and if the comparison and verification are passed, sends the electric meter readings to the server; if yes, the current ammeter reading of the intelligent ammeter is obtained according to the ammeter reading obtained through identification processing of various image identification models; and the server performs associated storage according to the ammeter number of the intelligent ammeter and the current ammeter reading. According to the invention, the meter reading efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of smart meters, and in particular to a smart meter reading method. Background Art

[0002] Smart meters are energy metering devices that combine modern communication technology with power system management. The introduction of smart meters makes energy management more intelligent and efficient, and has a positive effect on the operation of the power system and the electricity consumption behavior of users. In the traditional meter reading method, manual meter reading occupies a dominant position. At present, manual meter reading requires a lot of time and manpower, is inefficient, and increases operating costs. At the same time, due to the subjectivity and uncertainty of manual operation, problems such as unclear handwriting and reading errors often occur, resulting in inaccurate data, affecting the power department's accurate judgment and management of electricity consumption. At the same time, there are differences in the design and display of meters from different manufacturers and models, which brings difficulties to unified meter reading and management. Moreover, it is difficult for traditional meter reading methods to monitor electricity consumption data in real time, and the discovery and processing of abnormal electricity consumption are often delayed, which cannot avoid electricity waste and safety hazards in time. In addition, under complex environmental conditions, such as different light, temperature and humidity, the accuracy and stability of traditional meter reading methods will also be greatly affected.

[0003] With the development of the power industry and the improvement of users' requirements for power service quality, the traditional meter reading method can no longer meet the needs. Therefore, a smart meter reading method is proposed. Summary of the invention

[0004] The purpose of this application is to provide a smart meter reading method and system to improve the meter reading efficiency.

[0005] To achieve the above object, according to a first aspect of the present application, a smart meter reading method is provided, the method comprising:

[0006] Step S10, the high-definition camera photographs the smart meter according to a preset meter reading time interval to obtain an image of the meter, and uploads the image of the meter to an edge computing device through the Internet of Things; the high-definition camera is installed in front of the dial of the smart meter;

[0007] Step S20, the edge computing device performs image preprocessing on the electric meter image, and transmits the preprocessed electric meter image to the server;

[0008] Step S30, the server uses multiple image recognition models based on a deep learning algorithm to respectively recognize the received meter images, and performs comparison verification on the meter readings obtained by the multiple image recognition models. If the comparison verification passes, the server obtains the current meter reading of the smart meter based on the meter readings obtained by the multiple image recognition models.

[0009] Step S40: The server associates and stores the meter number of the smart meter with the current meter reading.

[0010] Optionally, the method further comprises:

[0011] In step S50, the server queries the historical meter readings of the smart meter in the past preset time according to the meter number, and determines whether there is an abnormal fluctuation in the current meter reading based on the historical meter readings of the smart meter in the past preset time and the current meter reading. If so, a warning message is issued to the electricity user of the smart meter.

[0012] Optionally, the step S10 further includes:

[0013] The environmental sensor detects the meter environment information of the smart meter according to the preset meter reading time interval, and uploads the meter environment information to the edge computing device through the Internet of Things; the environmental sensor is installed around the smart meter;

[0014] The step S20 further includes:

[0015] The edge computing device pre-processes the electric meter environment information and transmits the pre-processed electric meter environment information to the server;

[0016] The step S30 further includes:

[0017] The server adopts multiple image recognition models based on deep learning algorithms to respectively identify the meter readings and meter numbers of the received meter images and the meter environmental information, and performs comparative verification on the meter readings obtained by the multiple image recognition models. If the comparative verification passes, the current meter readings of the smart meter are obtained based on the meter readings obtained by the multiple image recognition models.

[0018] Optionally, the environmental sensor includes a temperature sensor, a humidity sensor and a light sensor, and the electric meter environmental information includes temperature information, humidity information and light information.

[0019] Optionally, the step S50 further includes:

[0020] The server associates the meter number of the smart meter with the current meter reading and encrypts and stores it in the blockchain database:

[0021] User access rights are set to authorize or prohibit users from accessing and processing the meter readings in the blockchain database, and users of different levels have different operating permissions.

[0022] Optionally, the multiple image recognition models based on the deep learning algorithm include an image recognition model based on a support vector machine algorithm, an image recognition model based on a decision tree algorithm, and an image recognition model based on a random forest algorithm.

[0023] Optionally, the step S40 further includes:

[0024] When the errors between any two of the meter readings obtained by the multiple image recognition models are less than a preset error threshold, it is determined that the comparison check has passed, otherwise the comparison check has failed.

[0025] Optionally, the step S40 further includes:

[0026] If the comparison verification fails, the server outputs a warning message to notify the meter management staff to conduct manual review.

[0027] Optionally, obtaining the current meter reading of the smart meter according to the meter readings obtained by the recognition and processing of the multiple image recognition models further includes:

[0028] The current meter reading of the smart meter is obtained by weighted summing of the meter readings obtained by the recognition processing of the multiple image recognition models, and the weight of each image recognition model is set according to the recognition accuracy of each image recognition model during testing.

[0029] According to a second aspect of the present application, a smart meter reading system is provided for executing the smart meter reading method according to the first aspect of the present application, including a server, a high-definition camera, an environmental sensor and an edge computing device.

[0030] The smart meter reading method and system described in this application have the following advantages:

[0031] Efficiency: The high-definition camera automatically captures the meter image, eliminating the need for manual on-site meter reading, greatly saving time and labor costs.

[0032] Accuracy: Using deep learning algorithms to identify meter images reduces errors in manual readings and improves data accuracy.

[0033] Real-time: Smart meters can monitor electricity consumption data in real time, detect abnormal electricity consumption in time, and avoid electricity waste and safety hazards.

[0034] Stability: Under complex environmental conditions, the smart meter reading method is less affected by light, temperature and humidity, ensuring the stability of the readings.

[0035] Unified management: Through data processing by servers and edge computing devices, unified meter reading and management of meters of different manufacturers and models can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 The present invention is a flowchart of a smart meter reading method in one embodiment of the present application.

[0038] Figure 2 This is a schematic diagram of the framework of a smart meter reading system in another embodiment of the present application. DETAILED DESCRIPTION

[0039] The detailed description of the drawings is intended as an illustration of the current embodiment of the present application, and is not intended to represent the only form in which the present application can be implemented. It should be understood that the same or equivalent functions can be accomplished by different embodiments intended to be included in the spirit and scope of the present application.

[0040] See also Figure 1 An embodiment of the present application provides a smart meter reading method, comprising the following steps:

[0041] Step S10, a high-definition camera photographs the smart meter according to a preset meter reading time interval to obtain an image of the meter, and uploads the image of the meter to an edge computing device through the Internet of Things; the high-definition camera is installed directly in front of the dial of the smart meter.

[0042] Specifically, a high-definition camera is a device that can capture high-resolution images. The preset meter reading interval refers to a time period set in advance by the system, such as every month. The high-definition camera will automatically perform the shooting task when this time period arrives. Meter reading is generally done on a monthly basis. When the preset time interval arrives, the high-definition camera will automatically aim at the dial of the smart meter to capture the digital display on the meter, which represents the user's electricity consumption. The captured meter image will be sent to a remote server through the Internet of Things (IoT) technology, which enables devices to connect to the network and exchange data. The high-definition camera is placed on the front of the smart meter to ensure that the meter readings can be clearly captured.

[0043] In step S20, the edge computing device performs image preprocessing on the electric meter image, and transmits the preprocessed electric meter image to the server.

[0044] Specifically, in this embodiment, a server is a remote computer system that stores data, processes requests, and performs various computing tasks. An edge computing device is a computer system located at the edge of a network, closer to the data source (in this embodiment, a smart meter), that can perform some data processing tasks, reduce the burden on the central server, and increase processing speed.

[0045] Image preprocessing refers to a series of processing performed on the meter image to improve the image quality and recognition accuracy. Image preprocessing may include but is not limited to the following steps:

[0046] Denoising: removing random noise from an image;

[0047] Contrast enhancement: Improves the contrast of the image to make the meter reading clearer;

[0048] Crop and scale: resize the image to include only the meter reading;

[0049] Lighting Correction: Adjust images to reduce the effects of different lighting conditions.

[0050] Furthermore, the pre-processed electricity meter image will be sent back to the server for further image recognition and processing.

[0051] In summary, in the method of this embodiment, after the edge computing device receives the meter image uploaded by the high-definition camera, it pre-processes the meter image to improve the image quality, and then sends the processed image to the server for subsequent meter reading recognition. This process helps to reduce the workload of the server and can quickly process data close to the data source.

[0052] In step S30, the server uses multiple image recognition models based on deep learning algorithms to respectively identify the received meter images, and compares and verifies the meter readings obtained by the multiple image recognition models. If the comparison and verification passes, the current meter readings of the smart meter are obtained based on the meter readings obtained by the multiple image recognition models.

[0053] Specifically, deep learning is a machine learning technology that can learn how to perform specific tasks, such as image recognition, through large amounts of data and iterative training. In this embodiment, the server uses a variety of trained deep learning models specifically for recognizing numbers on meter images. The server will use these different deep learning models to recognize the pre-processed meter images transmitted back from the edge computing device, and each deep learning model will focus on different aspects of the image recognition or use different techniques to improve the accuracy and reliability of recognition. Each deep learning model will provide a meter reading, and the server will compare these readings to check whether they are consistent. This verification is to ensure the accuracy of the reading, because if multiple independent models give the same reading, then the credibility of the reading will be higher. If the server finds that the readings given by multiple models are consistent, or the difference is within an acceptable range, then the comparison verification is considered to have passed. Once the verification passes, the server will use the readings provided by these models as the current readings of the smart meter, which will be used to record the user's electricity consumption, billing or other power system management activities.

[0054] Step S40: The server associates and stores the meter number of the smart meter with the current meter reading.

[0055] Specifically, each smart meter has a unique identifier, namely the meter number, which is used to distinguish different meters. The current meter reading refers to the meter reading identified and verified by the deep learning model in step S30, which represents the power consumption of the smart meter at a specific point in time. Associative storage refers to storing the meter number and the corresponding meter reading together in the database. This storage method establishes an association between the meter number and the reading, so that each reading record can be traced back to a specific meter, which facilitates subsequent data management and analysis, for example:

[0056] Billing and bill generation: Based on the meter number and readings, an electricity bill can be accurately generated for each user.

[0057] Power usage data analysis: By analyzing the stored data, you can understand the user's power usage pattern and make demand forecasts or energy-saving recommendations.

[0058] Equipment management: The associated storage of meter numbers and readings helps power companies perform equipment maintenance and troubleshooting.

[0059] In some embodiments, the method further comprises:

[0060] In step S50, the server queries the historical meter readings of the smart meter in the past preset time according to the meter number, and determines whether there is an abnormal fluctuation in the current meter reading based on the historical meter readings of the smart meter in the past preset time and the current meter reading. If so, a warning message is issued to the electricity user of the smart meter.

[0061] Specifically, the server will retrieve the historical reading records of the meter in the past period of time (preset time) from the database according to the meter number. The preset time can be several months, depending on the monitoring needs. The server will compare the current meter reading with the historical reading to analyze whether there are significant differences or fluctuations. Abnormal fluctuations indicate power theft, equipment failure or other power usage problems. If the server detects abnormal fluctuations in the meter readings, it will send early warning information to the power users associated with the meter. This information can be sent via SMS, email, application notifications, etc. to remind users of power usage issues. Step S50 increases the monitoring capabilities of the smart meter system, helps to improve the transparency and safety of power use, and also promotes users to use power resources more rationally.

[0062] In some embodiments, the step S10 further includes:

[0063] The environmental sensor detects the meter environment information of the smart meter according to the preset meter reading time interval, and uploads the meter environment information to the edge computing device through the Internet of Things; the environmental sensor is installed around the smart meter;

[0064] The step S320 further includes:

[0065] The edge computing device will preprocess the electric meter environment information and transmit the preprocessed electric meter environment information to the server; specifically, the edge computing device preprocesses the received electric meter environment information, including data cleaning, standardization or conversion steps, to improve the availability and accuracy of the data.

[0066] The step S30 further includes:

[0067] The server adopts multiple image recognition models based on deep learning algorithms to respectively identify the meter readings and meter numbers of the received meter images and the meter environmental information, and performs comparative verification on the meter readings obtained by the multiple image recognition models. If the comparative verification passes, the current meter readings of the smart meter are obtained based on the meter readings obtained by the multiple image recognition models.

[0068] Specifically, when the server recognizes the meter image, it also considers the meter environment information. This means that the deep learning model not only recognizes the reading and number based on the meter image, but also uses the environmental information to improve the accuracy of the recognition. The server uses multiple image recognition models to process the meter image and environmental information respectively, and compares and verifies the obtained meter readings. If the readings given by different models are consistent or the difference is within an acceptable range, and they are still consistent after considering the environmental information, the recognition result is considered reliable. The smart meter reading method not only considers the recognition of the meter image, but also integrates the environmental monitoring data to improve the accuracy and robustness of the entire system. By combining environmental information, the changes in meter readings can be better understood and adjustments can be made in environmental conditions that affect the accuracy of the readings.

[0069] In some embodiments, the environmental sensor includes a temperature sensor, a humidity sensor, and a light sensor, and the electric meter environmental information includes temperature information, humidity information, and light information.

[0070] Specifically, the temperature information is data about the temperature around the meter provided by the temperature sensor, the humidity information is data about the humidity around the meter provided by the humidity sensor, and the light information is data about the light level around the meter provided by the light sensor. Changes in temperature can affect the performance of electronic devices, including the accuracy of smart meters. Too high or too low humidity can cause damage or performance degradation to the meter or other electronic components. Changes in light conditions can affect the quality of the meter image captured by the camera, thereby affecting the accuracy of the meter readings.

[0071] In some embodiments, the step S40 further includes:

[0072] The server associates the meter number of the smart meter with the current meter reading and encrypts and stores it in the blockchain database:

[0073] User access rights are set to authorize or prohibit users from accessing and processing the meter readings in the blockchain database, and users of different levels have different operating permissions.

[0074] Specifically, after associating the meter number with the current meter reading, the server does not store the data in a traditional database, but in a blockchain database. Blockchain is a distributed ledger technology that ensures the security and immutability of data through encryption. When data is stored on the blockchain, it is encrypted to protect privacy and prevent unauthorized access.

[0075] In the blockchain database, the system sets access rights for different users, which means that only users with appropriate permissions can access and process meter reading data. The system assigns different permissions based on the user's role or level. For example, ordinary users can only view their own electricity consumption data, while power company staff have the authority to modify or analyze the data.

[0076] Here is a specific example to illustrate this process:

[0077] Consider a smart meter reading system that uses blockchain to store data:

[0078] 1. Meter ID and reading association: Smart meter A's ID is "12345" and its current reading is "1000kWh". The server associates these two pieces of information.

[0079] 2. Encrypted storage: The server encrypts the associated information and stores it in the blockchain database. This database is a distributed network consisting of multiple nodes, each of which stores a copy of the data.

[0080] 3. User access rights settings:

[0081] Ordinary user: User Zhang can only view the electricity consumption data of his own meter (number "12345"). His authority on the blockchain is limited to querying information related to his meter.

[0082] Power company technicians: Technician Li Si has higher privileges and can access all meter readings for system maintenance and data analysis.

[0083] Administrator: Administrator Wang Wu has the highest authority and can set user access rights, monitor system operation status, and audit or modify data when necessary.

[0084] In this way, blockchain technology not only ensures the transparency and security of data, but also ensures that different users can access and process data according to their needs and permission levels through refined permission management. This architecture is of great significance for protecting user privacy, preventing data tampering and ensuring system security.

[0085] In some embodiments, the multiple image recognition models based on deep learning algorithms include an image recognition model based on a support vector machine algorithm, an image recognition model based on a decision tree algorithm, and an image recognition model based on a random forest algorithm.

[0086] Specifically, here is a simplified example of how to use an image recognition model based on support vector machines (SVM), decision trees, and random forest algorithms to identify readings from a smart meter:

[0087] Suppose we have an image of a smart meter and need to recognize the digital reading on it.

[0088] Image recognition model based on support vector machine algorithm:

[0089] 1. Feature extraction: First, the model extracts feature vectors from the meter image. These features include the outline, texture, color, or other visual features of the numbers.

[0090] 2. Number category judgment: Then, a support vector machine (SVM) algorithm is used to judge the category of each number. For example, if the feature vector indicates that a number has a round top and a straight bottom, SVM will classify it as the number "0".

[0091] Image recognition model based on decision tree algorithm:

[0092] 1. Construct decision rules: The decision tree algorithm constructs a series of decision rules based on the shape, size, and position characteristics of the number. For example, the rule is "if the top of the number is closed, then it is 0, 6, or 9."

[0093] 2. Determine the reading step by step: The algorithm will step through the decision tree according to these rules until it reaches a leaf node, which represents a specific number. For example, if the top of the number is closed, but the bottom extends to the right, then according to the decision rule, it is the number "6".

[0094] Image recognition model based on random forest algorithm:

[0095] 1. Multiple judgments: The random forest algorithm uses multiple decision trees to judge the meter reading image. Each decision tree is trained based on different data subsets and features.

[0096] 2. Voting to determine the final result: Each decision tree will give a prediction result, and then the random forest algorithm will use a voting mechanism to determine the final reading. For example, if 3 of the 5 decision trees judge that a number is "6" and the other 2 judge it to be "9", then the random forest will ultimately determine that the number is "6".

[0097] For example:

[0098] Assume that the meter image shows a reading of "1234kWh".

[0099] The SVM model will identify the feature vector of each number and judge it as "1", "2", "3", and "4" respectively.

[0100] The decision tree model determines the identity of each number through a series of rules based on the shape characteristics of the number.

[0101] The random forest model uses multiple decision trees to identify each digit individually, and then uses a voting mechanism to determine the final reading as "1234".

[0102] These models can work in parallel, and the server compares their outputs. If multiple models give the same reading, the reading is more reliable. If there is a disagreement, the server will take additional verification measures or flag the meter reading for further manual review.

[0103] In some embodiments, the step S40 further comprises:

[0104] When the errors between any two of the meter readings obtained by the multiple image recognition models are less than a preset error threshold, it is determined that the comparison check has passed, otherwise the comparison check has failed.

[0105] Specifically, this embodiment refers to the use of different algorithms (such as SVM, decision tree, random forest, etc.) to identify digital readings on the meter image. The error refers to the difference between the meter readings identified by different models. For example, if the readings identified by two models are 1001kWh and 1002kWh respectively, then the error between them is 1kWh. The preset error threshold is a pre-set value used to determine whether the reading error is within an acceptable range. If the error is less than this threshold, the reading is considered to be reliable. If the error between the meter readings identified by any two models is less than the preset error threshold, then the readings are considered to be consistent and the comparison check passes. If the error between the meter readings identified by any two models exceeds the preset error threshold, then the readings are considered to be inconsistent and the comparison check fails.

[0106] Here is a specific example to illustrate this process:

[0107] Assume that there are three image recognition models A, B, and C, which respectively recognize the reading of a smart meter as:

[0108] Model A: 1001kWh;

[0109] Model B: 1002kWh;

[0110] Model C: 1001kWh;

[0111] The preset error threshold is 2kWh.

[0112] Compare the reads identified by these models:

[0113] The error between model A and model B is 1 kWh (1002 1001 = 1), which is less than 2 kWh.

[0114] The error between Model A and Model C is 0 kWh (1001 1001 = 0), which is less than 2 kWh.

[0115] The error between Model B and Model C is 1 kWh (1002 1001 = 1), which is less than 2 kWh.

[0116] Because the error between any two model-recognized readings is less than the preset error threshold of 2 kWh, the comparison check passes. The system can determine that the current reading of the smart meter is 1001 kWh or 1002 kWh (depending on whether the reading is rounded or otherwise processed).

[0117] Conversely, if the error between Model A and Model C is greater than 2 kWh, for example, Model C identifies a reading of 1005 kWh, then the comparison check will not pass and the system will flag this meter reading for further inspection or re-identification.

[0118] In some embodiments, step S30 further includes:

[0119] If the comparison verification fails, the server outputs a warning message to notify the meter management staff to conduct manual review.

[0120] Specifically, failure of the comparison verification means that in step S30, the error between the meter readings identified by different image recognition models exceeds the preset error threshold, or there is inconsistency in the readings between the models. When the verification fails, the server will generate and output an early warning message. This information is usually used to notify relevant management personnel, indicating that there are potential problems in the system that need attention. The early warning message will be sent to professionals responsible for meter management to inform them that the system has detected a problem and further processing is required. The purpose of the early warning message is to prompt the meter management staff to manually review the problematic meter readings. The review process may include re-checking the meter image, verifying the actual meter readings, checking the working status of sensors and cameras, etc.

[0121] Here is a specific example to illustrate this process:

[0122] Assume that there are three image recognition models A, B, and C, which respectively recognize the reading of a smart meter as:

[0123] Model A: 1001kWh;

[0124] Model B: 1003kWh;

[0125] Model C: 1002kWh;

[0126] The preset error threshold is 2kWh.

[0127] Compare the reads identified by these models:

[0128] The error between model A and model B is 2 kWh (1003 1001 = 2), which is equal to the preset error threshold.

[0129] The error between Model A and Model C is 1 kWh (1002 1001 = 1), which is smaller than the preset error threshold.

[0130] The error between model B and model C is 1 kWh (1003 1002 = 1), which is smaller than the preset error threshold.

[0131] In this case, since the error between model A and model B is equal to the preset error threshold, this may be regarded as a comparison check failure. Therefore, the server will output an early warning message to notify the meter management staff.

[0132] The warning message might look like this: "Attention: The reading verification of meter number 12345 has not passed. Please conduct manual review."

[0133] After receiving this warning, the meter management staff will take actions, such as checking whether the meter is damaged, whether the camera is aligned, whether the environmental conditions affect the reading, etc., to ensure the accuracy of the meter reading.

[0134] In some embodiments, obtaining the current meter reading of the smart meter according to the meter readings obtained by the recognition and processing of the multiple image recognition models further includes:

[0135] The current meter reading of the smart meter is obtained by weighted summing of the meter readings obtained by the recognition processing of the multiple image recognition models, and the weight of each image recognition model is set according to the recognition accuracy of each image recognition model during testing.

[0136] Specifically, different image recognition models can obtain multiple meter readings after processing the meter images. Weighted summation means adding the meter readings identified by each image recognition model according to its weight to obtain a comprehensive reading. Each model has a weight value associated with it, which reflects the accuracy and reliability of the model in identifying meter readings. The weight is set based on the performance of the model in the testing phase, that is, the more accurate the model is in identifying meter readings, the higher its weight is.

[0137] Here is a specific example to illustrate this process:

[0138] Assume that there are three image recognition models A, B, and C, which respectively recognize the reading of a smart meter as:

[0139] Model A identified reading: 1001kWh;

[0140] Model B identified reading: 1002kWh;

[0141] Model C identified reading: 1003kWh;

[0142] During the testing phase, the recognition accuracy of the three models were:

[0143] Accuracy of Model A: 95%

[0144] Model B accuracy: 90%

[0145] Model C accuracy: 92%

[0146] Based on the accuracy, we can set weights for each model:

[0147] Weight of model A: 0.95 / (0.95+0.90+0.92)≈0.326;

[0148] The weight of model B: 0.90 / (0.95+0.90+0.92)≈0.302;

[0149] The weight of model C: 0.92 / (0.95+0.90+0.92)≈0.372;

[0150] Now, we use these weights to calculate the weighted sum of the readings:

[0151] Weighted readings = (readings of model A × weight of model A) + (readings of model B × weight of model B) + (readings of model C × weight of model C);

[0152] Weighted reading = (1001 × 0.326) + (1002 × 0.302) + (1003 × 0.372);

[0153] Weighted reading ≈ 326.326 + 303.604 + 372.109;

[0154] Weighted reading ≈ 1002.039 kWh;

[0155] Therefore, the current meter reading of the smart meter is approximately 1002.039 kWh, which is the weighted sum of the three model recognition results. This method can more accurately reflect the meter reading because it takes into account the recognition accuracy of each model.

[0156] The smart meter reading system needs to be tested before application. The system testing method aims to comprehensively evaluate the performance and reliability of the smart meter reading recognition system. The specific steps are summarized as follows:

[0157] 1. Select diverse test samples:

[0158] Test meters of different types, ages, and reading ranges to simulate various real-world usage scenarios.

[0159] 2. Accuracy Assessment:

[0160] Compare the readings identified by the system with the actual meter readings, calculate the accuracy, and ensure that the system can accurately identify the meter readings under different conditions.

[0161] 3. Stability test:

[0162] Run the system for a period of time to observe whether the system has performance degradation or system failures after long-term operation. Check whether there is data loss or transmission interruption to ensure that the system remains stable during long-term operation.

[0163] 4. Environmental adaptability test:

[0164] Test under different lighting, temperature, humidity and other environmental conditions to verify the system's adaptability in different environments. Simulate abnormal fluctuations in meter readings, network failures and equipment failures to test the system's performance under these abnormal conditions.

[0165] 5. Test result analysis:

[0166] Analyze the accuracy, stability and environmental adaptability of the system based on the test results, and evaluate whether the system performance has reached the expected goal. Summarize the problems found in the test to provide a basis for system optimization and improvement.

[0167] Through the test method, the performance of the smart meter reading recognition system can be comprehensively evaluated to ensure that the system can work accurately and stably under various conditions, thereby improving the reliability and performance of the system.

[0168] like Figure 2 As shown, another embodiment of the present application provides a smart meter reading system for executing the smart meter reading method according to the aforementioned embodiment, including a server, a high-definition camera, an environmental sensor and an edge computing device.

[0169] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A smart meter reading method, characterized in that: The method comprises: Step S10, the high-definition camera photographs the smart meter according to a preset meter reading time interval to obtain an image of the meter, and uploads the image of the meter to an edge computing device through the Internet of Things; the high-definition camera is installed in front of the dial of the smart meter; Step S20, the edge computing device performs image preprocessing on the electric meter image, and transmits the preprocessed electric meter image to the server; Step S30, the server uses multiple image recognition models based on a deep learning algorithm to respectively recognize the received meter images, and performs comparison verification on the meter readings obtained by the multiple image recognition models. If the comparison verification passes, the server obtains the current meter reading of the smart meter based on the meter readings obtained by the multiple image recognition models. Step S40: The server associates and stores the meter number of the smart meter with the current meter reading.

2. The method according to claim 1, characterized in that The method further comprises: In step S50, the server queries the historical meter readings of the smart meter in the past preset time according to the meter number, and determines whether there is an abnormal fluctuation in the current meter reading based on the historical meter readings of the smart meter in the past preset time and the current meter reading. If so, a warning message is issued to the electricity user of the smart meter.

3. The method according to claim 1, characterized in that The step S10 further includes: The environmental sensor detects the meter environment information of the smart meter according to the preset meter reading time interval, and uploads the meter environment information to the edge computing device through the Internet of Things; the environmental sensor is installed around the smart meter; The step S20 further includes: The edge computing device pre-processes the electric meter environment information and transmits the pre-processed electric meter environment information to the server; The step S30 further includes: The server adopts multiple image recognition models based on deep learning algorithms to respectively identify the meter readings and meter numbers of the received meter images and the meter environmental information, and performs comparative verification on the meter readings obtained by the multiple image recognition models. If the comparative verification passes, the current meter readings of the smart meter are obtained based on the meter readings obtained by the multiple image recognition models.

4. The method according to claim 3, characterized in that: The environmental sensors include a temperature sensor, a humidity sensor and a light sensor, and the electric meter environmental information includes temperature information, humidity information and light information.

5. The method according to claim 3, characterized in that: The step S50 further includes: The server associates the meter number of the smart meter with the current meter reading and encrypts and stores it in the blockchain database: User access rights are set to authorize or prohibit users from accessing and processing the meter readings in the blockchain database, and users of different levels have different operating permissions.

6. The method according to claim 1, characterized in that The multiple image recognition models based on deep learning algorithms include an image recognition model based on a support vector machine algorithm, an image recognition model based on a decision tree algorithm, and an image recognition model based on a random forest algorithm.

7. The method according to claim 1, characterized in that The step S40 further comprises: When the errors between any two of the meter readings obtained by the multiple image recognition models are less than a preset error threshold, it is determined that the comparison check has passed, otherwise the comparison check has failed.

8. The method according to claim 7, characterized in that The step S40 further includes: If the comparison verification fails, the server outputs a warning message to notify the meter management staff to conduct manual review.

9. The method according to claim 7, characterized in that: The step of obtaining the current meter reading of the smart meter according to the meter readings obtained by the recognition and processing of the multiple image recognition models further includes: The current meter reading of the smart meter is obtained by weighted summing of the meter readings obtained by the recognition processing of the multiple image recognition models, and the weight of each image recognition model is set according to the recognition accuracy of each image recognition model during testing.

10. A smart meter reading device, characterized in that: Used to execute the smart meter reading method according to any one of claims 1 to 9, comprising a server, a high-definition camera, an environmental sensor and an edge computing device.

Images