Electric power metering terminal

Abstract

The invention discloses an electric power metering terminal, and relates to the technical field of intelligent power grids. The system comprises an intelligent electric meter terminal, an edge AI gateway and a cloud platform. The intelligent electric meter terminal performs high-frequency sampling; the edge AI gateway realizes anomaly detection and load decomposition through an STM32H7 processor, a neural network accelerator, a lightweight ST-CNN model and an improved Seq2Point model, and communicates with the cloud platform; and the cloud platform realizes data aggregation and management. The method comprises the steps of anomaly detection, load decomposition and energy efficiency optimization. According to the invention, the problems of insufficient real-time performance, low identification precision and weak edge capability in the prior art are solved, millisecond-level response, high-precision load decomposition and low-power-consumption edge calculation are realized, and the annual average electric charge of a family can be reduced by 15-18%.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grids, and in particular to an electric power metering terminal. Background Art

[0002] In the field of smart grid technology, anomaly detection and energy efficiency optimization of power metering terminals are crucial. However, existing technical solutions have many shortcomings, mainly reflected in the following aspects: Lack of real-time performance: The anomaly detection delay is usually greater than 10 seconds, and abnormal conditions in the power metering terminal cannot be discovered and handled in a timely manner, which may lead to power waste, equipment damage, and even safety accidents.

[0003] Low recognition accuracy: In complex power usage scenarios, the misjudgment rate of device operating status exceeds 25%, affecting the accuracy of load decomposition and hindering energy efficiency optimization and power management.

[0004] Weak edge capabilities: AI models cannot be executed locally, and power consumption is greater than 10W, which increases energy consumption and does not meet energy-saving and environmental protection requirements. It also limits its application in some power-sensitive scenarios. Summary of the Invention

[0005] In order to make up for the deficiencies of the prior art, the embodiments of the present application propose an electricity metering terminal to solve the problems existing in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: An electric power metering terminal, including a smart meter terminal, an edge AI gateway, and a cloud platform: The smart meter terminal includes a current sensor, a voltage sensor and a sampling module. The current sensor and the voltage sensor are used to collect power metering data. The sampling module is used to perform high-frequency sampling on the collected power metering data. The edge AI gateway includes an STM32H7 processor, a neural network accelerator, a lightweight ST-CNN model, and an improved Seq2Point model. The STM32H7 processor is used for data processing and control, the neural network accelerator is used to accelerate AI model calculations, the lightweight ST-CNN model is used for anomaly detection, and the improved Seq2Point model is used for load decomposition. The smart meter terminal is communicatively connected to the edge AI gateway, and the edge AI gateway is communicatively connected to the cloud platform.

[0007] As a further technical solution of the present invention: the sampling frequency of the sampling module is 4 kHz, and 256 data points lasting 0.2 seconds are intercepted each time.

[0008] As a further technical solution of the present invention: the smart meter terminal further includes a protective shell, the protection grade of the protective shell is IP65, and a Faraday cage is built in to resist electromagnetic interference.

[0009] As a further technical solution of the present invention: the core board of the smart meter terminal adopts an STM32G474RET6 processor and an AD7606C sampling chip.

[0010] As a further technical solution of the present invention: the edge AI gateway also includes a power supply circuit, SDRAM, FLASH and a dual-mode communication interface. The power supply circuit supplies power to the edge AI gateway, the SDRAM is used for temporary data storage, the FLASH is used for data storage, and the dual-mode communication interface is used for communication connection.

[0011] A method for detecting abnormalities in a power metering terminal, applied to the above-mentioned power metering terminal, comprises the following steps: Step 1: High-frequency sampling: the meter samples the current waveform at a rate of 4000 times per second, capturing 256 data points lasting 0.2 seconds each time. Step 2: Feature extraction: The current waveform is input into the spatiotemporal convolution model, which automatically analyzes three key features: the degree of waveform distortion, the current phase offset angle, and the number of burst pulses. Step 3: Abnormal determination: calculate the degree of deviation between the above features and the normal database. If the deviation value is greater than 3.5, immediately trigger the on-site sound and light alarm, and automatically encrypt and store the waveform evidence 10 seconds before the abnormality occurs.

[0012] As a further technical solution of the present invention: the format of the waveform evidence 10 seconds before the anomaly occurs is: timestamp, GPS location, waveform hash value.

[0013] As a further technical solution of the present invention: the deviation value threshold 3.5 in step 3 is obtained based on statistics of 100,000 groups of normal data.

[0014] A method for decomposing loads of an electric power metering terminal is applied to the electric power metering terminal, comprising the following steps: Step 1: Signal input, obtain the total household current curve, which comes from the real-time data of the electricity meter; Step 2: Intelligent identification uses an improved Seq2Point model to analyze the current segment by segment. Specifically, the process includes: first, scanning the current waveform with a convolutional layer to extract basic features; second, intelligently focusing on key change points through an "attention mechanism"; and third, determining the currently operating device. It can identify 10 common household appliances. Step 3: Output the results and generate a list of equipment starts and stops.

[0015] As a further technical solution of the present invention: the edge AI gateway performs ST-CNN inference once every 200ms, with a 256-point current waveform segment as input, and the model output is an abnormal probability value P, 0≤P≤1. If P>0.98 and lasts for 3 cycles, the buzzer alarm is activated and the encrypted data segment is saved.

[0016] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Significantly improved real-time performance: The entire process of anomaly detection from detection to alarm issuance takes only 0.2 seconds, which is more than 40 times faster than traditional solutions, achieving millisecond-level response and meeting the goal of a response time of less than 3 seconds.

[0017] 2. Significantly improved recognition accuracy: The recognition accuracy of high-power devices such as air conditioners and charging piles in load decomposition reached 93%, and the misjudgment rate decreased by 65%, improving the accuracy of device-level load decomposition and meeting the goal of an accuracy greater than 90%.

[0018] 3. Reduced power consumption at the edge: By adopting a lightweight model and optimized hardware architecture, the power consumption of AI computing at the edge is reduced, meeting the goal of less than 2W of power consumption.

[0019] 4. Achieve energy efficiency optimization: By combining time-of-day electricity prices and weather forecast information, home appliances can be intelligently controlled, reducing the average annual household electricity bill by 15-18%, achieving the effect of energy efficiency optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is the structural diagram of the power metering terminal. DETAILED DESCRIPTION

[0021] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] Example 1, as Figure 1 As shown, an electricity metering terminal includes a smart meter terminal, an edge AI gateway and a cloud platform.

[0023] The smart meter terminal consists of a current sensor, a voltage sensor, a sampling module, and a protective housing. The current and voltage sensors are used to collect electricity metering data. The sampling module operates at a 4kHz sampling rate, capturing 256 data points over a 0.2-second period. The core board uses an STM32G474RET6 processor and an AD7606C sampling chip. The protective housing has an IP65 rating and a built-in Faraday cage to resist electromagnetic interference.

[0024] The edge AI gateway includes an STM32H7 processor, a neural network accelerator, a lightweight ST-CNN model, an improved Seq2Point model, a power supply circuit, SDRAM, Flash, and a dual-mode communication interface. The STM32H7 processor is used for data processing and control; the neural network accelerator accelerates AI model calculations; the lightweight ST-CNN model is used for anomaly detection; the improved Seq2Point model is used for load decomposition; the power supply circuit provides power to the edge AI gateway; SDRAM is used for temporary data storage; Flash is used for data storage; and the dual-mode communication interface is used for communication connections.

[0025] The smart meter terminal communicates with the edge AI gateway, and the edge AI gateway communicates with the cloud platform.

[0026] Embodiment 2, a method for detecting abnormality in a power metering terminal, based on the power metering terminal of embodiment 1, comprises the following steps: Step 1: High-frequency sampling: the meter samples the current waveform at a rate of 4000 times per second, capturing 256 data points lasting 0.2 seconds each time.

[0027] Step 2: Feature extraction: The current waveform is input into the spatiotemporal convolutional neural network (ST-CNN). The model automatically analyzes three key features: the degree of waveform distortion (harmonic distortion rate), the current phase offset angle, and the number of burst pulses.

[0028] Step 3: Anomaly determination: Calculate the degree of deviation between the above features and the normal database (Mahalanobis distance). The threshold is 3.5 based on statistics of 100,000 sets of normal data. If the deviation value is greater than 3.5, a 120-decibel buzzer is immediately triggered for on-site sound and light alarms. The waveform evidence 10 seconds before the anomaly occurs is automatically encrypted and stored. The evidence format is {timestamp, GPS location, waveform hash value}.

[0029] The edge AI gateway performs ST-CNN inference every 200ms. The input is a 256-point current waveform segment. The model output is the abnormality probability value P (0≤P≤1). If P>0.98 and lasts for 3 cycles, the buzzer alarm is activated and the encrypted data segment is saved.

[0030] Example 3, a method for decomposing load of an electric power metering terminal, based on the electric power metering terminal of Example 1, comprises the following steps: Step 1: Signal input, obtain the household total current curve, which comes from the real-time data of the electric meter.

[0031] Step 2: Intelligent identification, using an improved Seq2Point model to analyze the current segment by segment. Specifically, the following steps are performed: first, using a convolutional layer to scan the current waveform and extract basic features; second, using an "attention mechanism" to intelligently focus on key change points, such as the startup moment of the air-conditioning compressor; and third, determining the currently operating device. It can identify 10 types of common household appliances, with an accuracy rate of 93% for high-power devices such as air conditioners and charging piles.

[0032] Step 3: Output the results and generate a device start and stop list, for example, "Air conditioner turned on at 14:30, power 1500W."

[0033] Example 4, an energy efficiency optimization method, based on the power metering terminal of Example 1, comprises the following steps: Step 1: Data fusion, real-time reception of two types of external information, namely time-based electricity price signals (such as 0.8 yuan / kWh during peak hours and 0.3 yuan / kWh during off-peak hours) and weather forecast data (light intensity, temperature, etc.).

[0034] Step 2: Optimization calculation, performing energy efficiency optimization calculation based on the received external information and load decomposition results.

[0035] Step 3: Automatic execution, sending instructions to smart sockets / home appliance controllers via wireless networks to achieve energy efficiency optimization. Empirical data shows that the average annual household electricity bill is reduced by 15-18%.

[0036] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be embraced therein.

[0037] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment have also been appropriately combined to form other implementation methods that are easy for those skilled in the art to understand.

Claims

1. An electric power metering terminal, characterized in that: Including smart meter terminals, edge AI gateways and cloud platforms: The smart meter terminal includes a current sensor, a voltage sensor and a sampling module. The current sensor and the voltage sensor are used to collect power metering data. The sampling module is used to perform high-frequency sampling on the collected power metering data. The edge AI gateway includes an STM32H7 processor, a neural network accelerator, a lightweight ST-CNN model, and an improved Seq2Point model. The STM32H7 processor is used for data processing and control, the neural network accelerator is used to accelerate AI model calculations, the lightweight ST-CNN model is used for anomaly detection, and the improved Seq2Point model is used for load decomposition. The smart meter terminal is communicatively connected to the edge AI gateway, and the edge AI gateway is communicatively connected to the cloud platform.

2. The power metering terminal according to claim 1, characterized in that: The sampling frequency of the sampling module is 4 kHz, and 256 data points lasting 0.2 seconds are captured each time.

3. The power metering terminal according to claim 1, characterized in that: The smart meter terminal further includes a protective shell having a protection grade of IP65 and a built-in Faraday cage to resist electromagnetic interference.

4. The power metering terminal according to claim 1, characterized in that: The core board of the smart meter terminal adopts STM32G474RET6 processor and AD7606C sampling chip.

5. The power metering terminal according to claim 2, characterized in that: The edge AI gateway also includes a power supply circuit, SDRAM, FLASH and a dual-mode communication interface. The power supply circuit provides power to the edge AI gateway, the SDRAM is used for temporary data storage, the FLASH is used for data storage, and the dual-mode communication interface is used for communication connection.

6. A method for detecting abnormality in an electric power metering terminal, applied to the electric power metering terminal according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: High-frequency sampling: the meter samples the current waveform at a rate of 4000 times per second, capturing 256 data points lasting 0.2 seconds each time. Step 2: Feature extraction: The current waveform is input into the spatiotemporal convolution model, which automatically analyzes three key features: the degree of waveform distortion, the current phase offset angle, and the number of burst pulses. Step 3: Abnormal determination: calculate the degree of deviation between the above features and the normal database. If the deviation value is greater than 3.5, immediately trigger the on-site sound and light alarm, and automatically encrypt and store the waveform evidence 10 seconds before the abnormality occurs.

7. The method for detecting abnormality of an electric power metering terminal according to claim 6, wherein: The format of the waveform evidence 10 seconds before the anomaly occurs is: timestamp, GPS location, and waveform hash value.

8. The method for detecting abnormality of an electric power metering terminal according to claim 6, wherein: The deviation value threshold of 3.5 in step 3 is obtained based on statistics of 100,000 sets of normal data.

9. A method for decomposing loads of an electric power metering terminal, applied to the electric power metering terminal according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Signal input, obtain the household total current curve, which comes from the real-time data of the electric meter; Step 2: Intelligent identification, using the improved Seq2Point model to analyze the current segment by segment. Specifically, the first step is to use the convolution layer to scan the current waveform and extract basic features; The second step is to intelligently focus on key change points through the "attention mechanism"; the third step is to determine the currently running device and can identify 10 types of common household appliances; Step 3: Output the results and generate a list of equipment starts and stops.

10. The method for decomposing load of an electric power metering terminal according to claim 9, characterized in that: The edge AI gateway performs ST-CNN inference every 200ms. The input is a 256-point current waveform segment. The model output is the abnormal probability value P, 0≤P≤1. If P>0.98 and lasts for 3 cycles, the buzzer alarm is activated and the encrypted data segment is saved.

Images