Intelligent electric meter state evaluation method and system based on neural network model, and medium

Abstract

The application discloses a kind of intelligent electric meter state evaluation method, system and medium based on neural network model, the method includes: data acquisition and pre-processing;Establish convolutional neural network model;Substitute collection data and train convolutional neural network model;Determine surrounding factor influence factor, modify convolutional neural network model;The state of intelligent electric meter is evaluated by the modified convolutional neural network model, and evaluation result is obtained;Intelligent electric meter rotation strategy is established according to evaluation result.The application can improve the level of intelligent electric meter asset management by the way of big data and neural network combination to evaluate the state of intelligent electric meter, realize effective monitoring and operation state evaluation, avoid the waste of funds and energy loss, increase the perception of user at the same time, ensure that metering work quality is efficient, legal and normative.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computer vision, and particularly relates to a smart meter state evaluation method and system based on a neural network model and a medium. BACKGROUND

[0002] According to the current national electric energy metering and verification regulation (JJG596-2012 electronic AC electric energy meter verification regulation), the verification period of a 2-level single-phase smart meter is 8 years. The smart meter has been fully covered by the China Southern Power Grid Company in 2018, and the electronic electric energy meter installed for the first time in 2012 is approaching the 8-year verification period, and millions of smart meters will face the problem of expiration and disassembly for verification every year. Due to the improvement of technology and manufacturing process, the actual operation life of the electric energy meter generally reaches 10-15 years. The International Metrology Organization TC3 / TS4 Working Group prepared the document “Follow-up verification of civil instruments in use by using sampling method” in July 2010, which suggests that if the sampling detection is qualified, the effective period can be extended by 50%. If a large number of electric energy meters with good performance are disassembled and returned according to the verification period, it will lead to a large amount of waste of funds and unpredictable power loss due to meter replacement.

[0003] In order to detect and evaluate the state of the electric meter in the transformer area, it is necessary to apply metrology big data to improve the asset management level of the smart meter, effectively monitor and evaluate the operation state, and then obtain accurate electric meter state evaluation results, so as to facilitate the formulation of electric meter replacement strategy. SUMMARY

[0004] The main purpose of the application is to overcome the shortcomings and deficiencies of the prior art, and provide a smart meter state evaluation method, system and medium based on a neural network model.

[0005] In order to achieve the above purpose, the following technical solutions are adopted in the application:

[0006] In one aspect of the application, a smart meter state evaluation method based on a neural network model is provided, which comprises the following steps:

[0007] Data acquisition and preprocessing;

[0008] Establishing a convolutional neural network model;

[0009] Substituting the collected data and training the convolutional neural network model;

[0010] Determining the surrounding influencing factors and correcting the convolutional neural network model;

[0011] Evaluating the state of the smart meter through the corrected convolutional neural network model to obtain an evaluation result;

[0012] According to the evaluation result, an intelligent electric meter rotation strategy is established.

[0013] As a preferred technical solution, the data collection and preprocessing are specifically:

[0014] Based on big data collection technology, a large amount of existing data is accumulated, and the data is cleaned and selected to obtain a data set that needs to be trained; the data table includes the file information and error information of the intelligent electric meter; the file information includes user number, installation date, transformer area number or transformer area name, electric meter reading, and location information; the error information includes error value, error rate, and out-of-tolerance value information.

[0015] As a preferred technical solution, the convolutional neural network model includes an input layer, an output layer, and a plurality of hidden layers, and the input and output relationship of the neural nodes of each input layer and hidden layer is:

[0016]

[0017] wherein x j is the output of the previous neural node, w ij is the weight from the previous neural node to the i node, b is the bias factor, and f is the activation function.

[0018] As a preferred technical solution, the convolutional neural network model is trained by combining the quasi-Newton method and the least squares method, specifically:

[0019] A unified threshold is used to preprocess the data of each transformer area to achieve a reasonable initial value.

[0020] The recursive least squares method is used to fine-tune the previous calculation error using the new data added each day.

[0021] As a preferred technical solution, the surrounding influencing factors include geographic location, installed transformer area, installation time, power supply information, environmental temperature, and environmental humidity.

[0022] As a preferred technical solution, the weight proportion of each surrounding influencing factor is: geographic location 15%, installed transformer area 10%, installation time 20%, power supply information 15%, environmental temperature 20%, and environmental humidity 20%.

[0023] As a preferred technical solution, the convolutional neural network model is corrected, specifically: according to the weight proportion of the surrounding influencing factors, the weight value of the convolutional neural network model is adjusted.

[0024] Another aspect of the present application provides a neural network model-based intelligent electric meter state evaluation system, applied to the neural network model-based intelligent electric meter state evaluation method, comprising a preprocessing module, a model construction module, a model training module, an influence factor correction module and a prediction result output module.

[0025] The preprocessing module is used for data collection and preprocessing.

[0026] The model construction module is used for establishing a convolutional neural network model.

[0027] The model training module is used for substituting the collected data and training the convolutional neural network model.

[0028] The influence factor correction module is used for determining surrounding influence factors and correcting the convolutional neural network model.

[0029] The prediction result output module is used for evaluating the state of the intelligent electric meter through the corrected convolutional neural network model, obtaining an evaluation result, and establishing an intelligent electric meter rotation strategy according to the evaluation result.

[0030] Another aspect of the present application provides a storage medium storing a program, which, when executed by a processor, implements the neural network model-based intelligent electric meter state evaluation method.

[0031] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0032] (1) The present application evaluates the state of the intelligent electric meter by combining big data with neural networks, which can improve the asset management level of the intelligent electric meter, realize effective monitoring and operation state evaluation, avoid waste of funds and loss of energy, and increase user perception, ensuring efficient and legally standardized metering work quality. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is a flowchart of the neural network model-based intelligent electric meter state evaluation method of the embodiment of the present application;

[0034] Figure 2 is a structural schematic diagram of the convolutional neural network model constructed by the embodiment of the present application;

[0035] Figure 3 is a structural schematic diagram of the neural network model-based intelligent electric meter state evaluation system of the embodiment of the present application;

[0036] Figure 4 is a structural schematic diagram of the storage medium of the embodiment of the present application. DETAILED DESCRIPTION

[0037] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application, so that those skilled in the art can better understand the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0038] Embodiments

[0039] As shown in the figure, the embodiment provides a smart meter state evaluation method based on a neural network model, including the following steps: Figure 1

[0040] S1, data acquisition and preprocessing, specifically:

[0041] Based on big data acquisition technology, a large amount of existing data is accumulated, and the data is preprocessed to obtain a data set that needs to be trained.

[0042] The preprocessing specifically includes data cleaning and screening.

[0043] These data can be determined according to the need to monitor the transformer area, and are obtained through the interface of the electric energy platform and saved to the data table. Including the file information and error information of the smart meter. The file information includes user number, installation date, transformer area number (or transformer area name), meter reading, location information; the error information includes error value, error rate, and excess value information.

[0044] S2, establish a convolutional neural network model, as shown in the figure, including an input layer, a hidden layer and an output layer; in the embodiment, the input of the input layer is 8, the node quantity of the hidden layer is 4, Figure 2 only as an example, the hidden layer therein can be multiple layers. Figure 2

[0045] Each neural node is an input-output relationship:

[0046]

[0047] Wherein, x j is the output of the previous neural node, w ij is the weight from the previous neural node to the i node, b is the bias factor, and f is the activation function (for example, the sigmoid function ).

[0048] S3, substitute the collected data and train the convolutional neural network model;

[0049] ​​This embodiment uses an optimization algorithm that combines the quasi-Newton method and the least squares method to train the convolutional neural network model.

[0050] While the BFGS quasi-Newton method converges quickly, its application requires precise threshold settings. Setting the threshold too low can lead to overfitting, while setting it too high may cause the iteration to stop before the calculated error reaches the threshold and approaches the true optimal value. Furthermore, threshold settings should vary depending on the size of the transformer substation, the load, and the power consumption characteristics. Therefore, it's impossible to set the same threshold for all transformer substations, and it's not easy to flexibly select initial values ​​based on the characteristics of different substations.

[0051] An optimization algorithm combining the quasi-Newton method and the least squares method can improve the above-mentioned shortcomings. This method first uses a uniform threshold to preprocess the data of each transformer area to quickly reach a reasonable initial value. Then, it uses a recursive form of the least squares method to fine-tune the previous calculation error using the newly added data each day, so as to achieve a good estimate of the calculation error for each day, thereby improving the calculation accuracy of the model.

[0052] S4. Identify surrounding influencing factors and revise the convolutional neural network model accordingly;

[0053] Since smart meters are affected by many factors during operation, we include six factors that have a significant impact: geographical location, installation area, installation time, power supply information, ambient temperature, and ambient humidity. These six factors and their corresponding weights are used to refine the network model.

[0054] In this embodiment, the weighting of each factor is as follows: geographical location 15%, installation area 10%, installation time 20%, power supply information 15%, ambient temperature 20%, and ambient humidity 20%.

[0055] By adjusting the network model using various factors and their corresponding weights, the corrected network model becomes more fault-tolerant and accurate.

[0056] S5. The status of the smart meter is evaluated using the modified convolutional neural network model to obtain the evaluation results;

[0057] S6. Establish a smart meter rotation strategy based on the evaluation results.

[0058] like Figure 3 As shown, in another embodiment of this application, a smart meter status evaluation system based on a neural network model is provided. The system includes a preprocessing module, a model building module, a model training module, an influencing factor correction module, and a prediction result output module.

[0059] The preprocessing module is used for data collection and preprocessing;

[0060] The model construction module is used for establishing a convolutional neural network model;

[0061] The model training module is used for substituting the collected data and training the convolutional neural network model;

[0062] The influence factor correction module is used for determining surrounding influence factors and correcting the convolutional neural network model;

[0063] The prediction result output module is used for evaluating the state of the smart meter through the corrected convolutional neural network model, obtaining an evaluation result, and establishing a smart meter rotation strategy according to the evaluation result.

[0064] As shown in Figure 4 In another embodiment of the present application, a storage medium is also provided, which stores a program, and the program is executed by a processor to implement a smart meter state evaluation method based on a neural network model, specifically:

[0065] data collection and preprocessing;

[0066] establishing a convolutional neural network model;

[0067] substituting the collected data and training the convolutional neural network model;

[0068] determining surrounding influence factors and correcting the convolutional neural network model;

[0069] evaluating the state of the smart meter through the corrected convolutional neural network model to obtain an evaluation result;

[0070] establishing a smart meter rotation strategy according to the evaluation result.

[0071] It should be understood that portions of the present application can be implemented in hardware, software, firmware, or combinations thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, implementation can be with any or a combination of the following technologies, which are all well known in the art: a discrete logic circuit having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0072] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited to the above embodiments, and any changes, modifications, substitutions, combinations, simplifications, etc. made without departing from the spirit and principles of the present application are equivalent replacement modes and are included in the protection scope of the present application.

Claims

1. A method for evaluating the status of smart meters based on a neural network model, characterized in that, Includes the following steps: Data acquisition and preprocessing, specifically: Based on big data acquisition technology, existing data is accumulated, and the data is cleaned and filtered to obtain a dataset for training. The dataset includes the smart meter's profile information and error information. The profile information includes user number, installation date, transformer area number or name, meter reading, and location information. The error information includes error value, error rate, and out-of-tolerance information. A convolutional neural network (CNN) model is established. This model is trained by combining the quasi-Newton method with the least squares method. Specifically, a uniform threshold is used to preprocess the data from each transformer substation to achieve a reasonable initial value. A recursive least squares method is then used to fine-tune the previous calculation errors using newly added data each day. The CNN model includes an input layer, an output layer, and several hidden layers. The input-output relationship of each neural node in the input and hidden layers is as follows: ; in, x j This is the output of the previous neural node. w ij For the previous neural node to i Node weights b As the bias factor, f For activation functions; Substitute the collected data and train the convolutional neural network model; Identify surrounding influencing factors and modify the convolutional neural network model accordingly; these factors include geographical location, installation area, installation time, power supply information, ambient temperature, and ambient humidity. The state of the smart meter is evaluated using the modified convolutional neural network model, and the evaluation results are obtained. Establish a smart meter rotation strategy based on the evaluation results.

2. The smart meter status evaluation method based on a neural network model according to claim 1, characterized in that, The weighting of each surrounding influencing factor is as follows: geographical location 15%, installation area 10%, installation time 20%, power supply information 15%, ambient temperature 20%, and ambient humidity 20%.

3. The smart meter status evaluation method based on a neural network model according to claim 1, characterized in that, The modification of the convolutional neural network model specifically involves adjusting the weight values ​​of the convolutional neural network model based on the weight ratio of surrounding influencing factors.

4. A smart meter status evaluation system based on a neural network model, characterized in that, A method for evaluating the status of a smart meter based on a neural network model, applicable to any one of claims 1-3, includes a preprocessing module, a model building module, a model training module, an influencing factor correction module, and a prediction result output module; The preprocessing module is used for data acquisition and preprocessing; The model construction model is used to build a convolutional neural network model; The model training module is used to input the collected data and train the convolutional neural network model. The influencing factor correction module is used to determine the surrounding influencing factors and correct the convolutional neural network model. The prediction result output module is used to evaluate the status of smart meters through a modified convolutional neural network model, obtain evaluation results, and establish a smart meter rotation strategy based on the evaluation results.

5. A storage medium storing a program, characterized in that: When the program is executed by the processor, it implements the smart meter status evaluation method based on a neural network model as described in any one of claims 1-3.

Images