CN117239743B - Electric energy meter electricity load acquisition method, device, equipment and medium - Google Patents

Abstract

This application discloses a method, device, equipment and medium for obtaining electric load of an electric energy meter, and relates to the field of electric power technology. The solution builds a load identification model through a time-series Transformer neural network, with a multi-head attention mechanism; so that when using the load identification model to predict the load type of the target power consumption data, each element in the target power consumption data can be fully paid attention to and processed , thus improving the accuracy and robustness of the model, solving the problem of low accuracy of current electric load identification by electric energy meters, and improving the utilization efficiency of electric load and the stability of the power grid.

Description

A method, device, equipment and medium for obtaining electrical load of an electric energy meter

Technical field

The present application relates to the field of electric power technology, and in particular to a method, device, equipment and medium for obtaining electric load of an electric energy meter.

Background technique

In daily life, electric energy meter is one of the key devices to measure electricity consumption. Traditional electric energy meters are usually non-intelligent designs and cannot obtain detailed information about the power grid. They can only obtain the total electricity consumption information of end users. In the construction of smart grid, it is necessary to accurately understand the power load status and analyze and utilize the power load. Therefore, the function of electrical load identification and acquisition is added to the electric energy meter.

However, the biggest drawback of current electric energy meters in identifying electrical loads is the low accuracy of identification, including the low accuracy of identifying the category of electrical equipment and the low accuracy of identifying the numerical value of the equipment's power consumption, which is not conducive to The power company monitors and manages the grid load.

In view of the above problems, how to solve the current low accuracy of electrical load identification by electric energy meters is an urgent problem for technicians in this field.

Contents of the invention

The purpose of this application is to provide a method, device, equipment and medium for obtaining the electric load of an electric energy meter, so as to solve the current problem of low accuracy of electric load identification and acquisition by electric energy meters.

In order to solve the above technical problems, this application provides a method for obtaining the electric load of an electric energy meter, including:

Read the target power consumption data through the electric energy meter; wherein the power consumption data includes target current data and target voltage data;

Perform differential processing on the target current data and the target voltage data respectively;

Input the processed target current data and target voltage data into the load identification model to obtain the target load type and target power consumption;

Among them, the construction process of the load identification model includes:

Obtain the current data and voltage data stored in the electric energy meter;

Preprocessing the current data and the voltage data respectively;

Based on the preprocessed current data and voltage data, the load identification model is trained using a sequential Transformer neural network to obtain the load identification model.

On the one hand, preprocessing the current data and the voltage data respectively includes:

Perform normalization processing on the current data and the voltage data respectively;

Perform segmentation processing on the normalized current data and voltage data in time sequence to obtain multiple sequences of current data and voltage data;

Mark the corresponding load type for the current data and voltage data of each sequence;

The current data and voltage data marked with the load type are divided into a training set and a test set according to a preset ratio.

On the other hand, using the sequential Transformer neural network to train the load identification model based on the preprocessed current data and voltage data includes:

The data in the training set is input into each multi-head attention model; wherein the multi-head attention model includes a first multi-head attention model and a second multi-head attention model, and the first multi-head attention model Occlusion operation is adopted;

In each of the multi-head attention models, obtain a preset number of training feature matrices of load types based on the data in the training set, and obtain an initial weight matrix;

Obtain the weight of the corresponding self-attention layer according to the training feature matrix and the initial weight matrix to obtain the corresponding output result of the self-attention layer;

Aggregate the output results of the self-attention layer corresponding to each of the load types to obtain the aggregated output result of the self-attention layer;

Obtain the corresponding output result of the multi-head attention model according to the weight coefficient and the aggregated output result;

Process the output results of each of the multi-head attention models respectively to obtain the final results corresponding to each of the multi-head attention models;

Aggregate the final results of each multi-head attention model, and use a normalized exponential function to output the aggregated final results to obtain the classification results of the temporal Transformer neural network and obtain the initial load identification model;

Obtain a new weight matrix according to the classification result and the cross-entropy loss function;

Perform backpropagation according to the cross-entropy loss function and the new weight matrix;

Determine whether the load identification model meets the preset requirements;

If yes, then the training of the load identification model is ended.

On the other hand, determining whether the load identification model meets preset requirements includes:

Verify the load identification model based on the data in the test set and determine whether the accuracy of the load identification model is greater than a threshold;

If so, confirm that the load identification model meets the preset requirements;

If not, it is confirmed that the load identification model does not meet the preset requirements.

On the other hand, before reading the target power consumption data through the electric energy meter, it also includes:

deploy the load identification model in the electric energy meter;

The electric energy meter electrical signal detection sensor of the electric energy meter is activated so as to obtain the target power consumption data through the electric energy meter electrical signal detection sensor.

On the other hand, after obtaining the target load type and target power consumption, it also includes:

Generate a target load curve and target load statistical information according to the target load type and the target power consumption;

The target load curve and the target load statistical information are stored in the storage space of the electric energy meter.

On the other hand, after obtaining the target load type and target power consumption, it also includes:

Generate acquisition logs of the target load type and target power consumption;

The log is stored in the storage space of the electric energy meter.

In order to solve the above technical problems, this application also provides an electric load acquisition device for an electric energy meter, including:

A reading module, configured to read target power consumption data through an electric energy meter; wherein the power consumption data includes target current data and target voltage data;

A differential processing module, configured to perform differential processing on the target current data and the target voltage data respectively;

A prediction module, configured to input the processed target current data and target voltage data into the load identification model to obtain the target load type and target power consumption;

Among them, the construction process of the load identification model includes:

Obtain the current data and voltage data stored in the electric energy meter;

Preprocessing the current data and the voltage data respectively;

Based on the preprocessed current data and voltage data, the load identification model is trained using a sequential Transformer neural network to obtain the load identification model.

In order to solve the above technical problems, this application also provides an electric load acquisition device for an electric energy meter, including:

Memory, used to store computer programs;

A processor, configured to implement the above-mentioned steps of the electric load acquisition method of an electric energy meter when executing the computer program.

In order to solve the above technical problems, the present application also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the above-mentioned method for obtaining the electrical load of an electric energy meter is implemented. A step of.

The method for obtaining the electric load of an electric energy meter provided by this application reads the target electric data through the electric energy meter; wherein the electric data includes target current data and target voltage data; differential processing is performed on the target current data and target voltage data respectively; Input the processed target current data and target voltage data into the load identification model to obtain the target load type and target power consumption. The construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; The current data and voltage data are preprocessed respectively; based on the preprocessed current data and voltage data, the time series Transformer neural network is used to train the load identification model to obtain the load identification model. It can be seen that the above scheme builds a load identification model through a time-series Transformer neural network and has a multi-head attention mechanism; so that when using the load identification model to predict the load type of the target power consumption data, each element in the target power consumption data can be Full attention and processing are carried out to improve the accuracy and robustness of the model, solve the problem of low accuracy of current electric load identification by electric energy meters, and improve the utilization efficiency of electric load and the stability of the power grid.

In addition, this application also provides an electrical load acquisition device, equipment and medium for an electric energy meter, with the same effect as above.

Description of drawings

In order to explain the embodiments of the present application more clearly, the drawings required to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, As far as workers are concerned, other drawings can also be obtained based on these drawings without exerting creative work.

Figure 1 is a flow chart of a method for obtaining electric load of an electric energy meter provided by an embodiment of the present application;

Figure 2 is a schematic diagram of the load identification model training process provided by the embodiment of the present application;

Figure 3 is a schematic diagram of an electric load acquisition device for an electric energy meter provided by an embodiment of the present application;

Figure 4 is a schematic diagram of an electric load acquisition device for an electric energy meter provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of this application.

The core of this application is to provide a method, device, equipment and medium for obtaining the electric load of an electric energy meter, aiming to solve the problem of low accuracy in the current electric energy meter's identification and acquisition of the electric load.

In order to enable those skilled in the art to better understand the solution of the present application, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

At present, the biggest drawback of electric energy meters in identifying electrical loads is the low identification accuracy, including the low accuracy in identifying the types of electrical equipment and the low accuracy in identifying the numerical value of the equipment's power consumption, which is not conducive to electric power The company monitors and manages the power grid load. In view of the above problems, this application provides a method for obtaining the electric load of an electric energy meter, aiming to solve the problem of low accuracy in the current electric energy meter's identification and acquisition of the electric load.

Figure 1 is a flow chart of a method for obtaining electric load of an electric energy meter provided by an embodiment of the present application. As shown in Figure 1, methods include:

S10: Read the target power consumption data through the electric energy meter.

Among them, the power consumption data includes target current data and target voltage data.

S11: Perform differential processing on the target current data and target voltage data respectively.

S12: Input the processed target current data and target voltage data into the load identification model to obtain the target load type and target power consumption.

Among them, the construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; preprocessing the current data and voltage data respectively; based on the preprocessed current data and voltage data, using the time series Transformer neural network to perform load analysis Training of the recognition model to obtain the load recognition model.

Specifically, in order to perform load identification and acquisition, the target power consumption data must first be read through the electric energy meter. It can be understood that the target power consumption data specifically includes target current data and target voltage data; the standard electric energy meter needs to output several voltage values and current values per second, and load identification means reading these data every second.

Further, differential processing is performed on the target current data and target voltage data respectively to obtain the latest voltage and current data. Finally, the processed target current data and target voltage data are input into the load identification model to obtain the target load type and target power consumption. It can be understood that the target load type is obtained through the load identification model, and the target power consumption can be directly calculated and obtained based on the target current data and target voltage data.

It should be noted that the load identification model in this embodiment is a model trained using a time series Transformer neural network based on current data and voltage data. Specifically, during the construction process of the load identification model, it is first necessary to obtain the current data and voltage data stored in the electric energy meter. It can be understood that the read-only memory (ROM) of the electric energy meter stores current data and voltage data in multiple time periods, and these data can be used as sample data for model training. Furthermore, due to differences in data quality, it is not appropriate to directly use current data and voltage data for model training. Therefore, some necessary preprocessing, such as filtering, standardization, and normalization, needs to be performed on the original current data and voltage data. etc., so that the quality of the two meets the training requirements of the model. In this embodiment, there is no restriction on the data preprocessing process, and it depends on the specific implementation situation. Finally, based on the preprocessed current data and voltage data, the time series Transformer neural network is used to train the load identification model to obtain the load identification model.

In this embodiment, the load identification model is constructed based on the sequential Transformer neural network. It can be understood that since the load identification model is mainly used to obtain load types, its task is mainly a classification task. Therefore, in this embodiment, the sequential Transformer neural network deletes the decoder of the traditional Transformer neural network. Only the encoder is used for classification. In this embodiment, there is no restriction on the specific construction process of the load identification model, and it depends on the specific implementation situation.

In this embodiment, the target power consumption data is read through the electric energy meter; wherein the power consumption data includes target current data and target voltage data; differential processing is performed on the target current data and target voltage data respectively; the processed target current data and The target voltage data is input into the load identification model to obtain the target load type and target electricity consumption; among which, the construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; conducting the current data and voltage data respectively. Preprocessing: Based on the preprocessed current data and voltage data, the time series Transformer neural network is used to train the load identification model to obtain the load identification model. It can be seen that the above scheme builds a load identification model through a time-series Transformer neural network and has a multi-head attention mechanism; so that when using the load identification model to predict the load type of the target power consumption data, each element in the target power consumption data can be Full attention and processing are carried out to improve the accuracy and robustness of the model, solve the problem of low accuracy of current electric load identification by electric energy meters, and improve the utilization efficiency of electric load and the stability of the power grid.

In order to preprocess the current data and voltage data, based on the above embodiments, in some embodiments, preprocessing the current data and voltage data respectively includes:

S130: Normalize the current data and voltage data respectively.

S131: Process the normalized current data and voltage data in segments according to time sequence to obtain multiple sequences of current data and voltage data.

S132: Mark the corresponding load type for each sequence of current data and voltage data.

S133: Divide the current data and voltage data marked with load types into training sets and test sets according to preset proportions.

Specifically, the current data and voltage data are first normalized and standardized to make the data on the same order of magnitude to facilitate the training and optimization of the neural network. Segment the time series current data and voltage data to form multiple small sequences of current data and voltage data. The data of each sequence is further labeled with the correct load type, and finally the current data and voltage data are divided into training sets and test sets according to preset proportions. In this embodiment, there is no restriction on the preset ratio, and it depends on the specific implementation situation. This completes the preprocessing process of current data and voltage data.

Figure 2 is a schematic diagram of the load identification model training process provided by the embodiment of the present application. Based on the above embodiments, in some embodiments, as shown in Figure 2, based on the preprocessed current data and voltage data, using the sequential Transformer neural network to train the load identification model includes:

S140: Input the data in the training set into each multi-head attention model.

Among them, the multi-head attention model includes the first multi-head attention model and the second multi-head attention model. The first multi-head attention model uses occlusion operation.

Specifically, the data in the training set is input into each multi-head attention (Multi-Head Attention) model. As shown in Figure 2, the multi-head attention model includes the first multi-head attention model and the second multi-head attention model. The difference between the two models is that the first multi-head attention model uses a masked operation. Historical data feeds back information at the current moment, that is, it blocks future information, thereby better reflecting the characteristics of data in time. The network focuses on time step information.

S141: In each multi-head attention model, obtain a preset number of training feature matrices of load types based on the data in the training set, and obtain an initial weight matrix.

S142: Obtain the weight of the corresponding self-attention layer according to the training feature matrix and the initial weight matrix to obtain the output result of the corresponding self-attention layer.

Further, in each multi-head attention model, a preset number is obtained based on the data in the training set Training feature matrix of load type/> , and obtain the initial weight matrix/> . Among them, N is the number of samples. According to the training feature matrix/> and initial weight matrix/> Get the weight of the corresponding self-attention layer/> , to obtain the output result of the corresponding self-attention layer. The output result of the attention layer is specifically/> , . Among them,/> Characterizes the length of the training feature matrix X.

S143: Aggregate the output results of the self-attention layer corresponding to each load type to obtain the aggregated output result of the self-attention layer.

S145: Obtain the output result of the corresponding multi-head attention model based on the weight coefficient and the aggregated output result.

Aggregate the output results of the self-attention layer corresponding to each load type to obtain the aggregated output result of the self-attention layer . Further pass the weight coefficient/> Learn the timing characteristics between different time channels (TimeChannel) in the same time layer, so as to use the weight coefficient/> And the aggregated output result F obtains the output result of the corresponding multi-head attention model:/> .

S146: Process the output results of each multi-head attention model respectively to obtain the final results corresponding to each multi-head attention model.

Subsequently, in order to alleviate the phenomenon of gradient dissipation, the output results of each multi-head attention model are processed separately to obtain the final results corresponding to each multi-head attention model. Specifically, through normalization and accumulation, the purpose of focusing only on the current difference is achieved. The output is: . Furthermore, the feedback network layer is relatively simple. It is a two-layer fully connected layer. The activation function of the first layer is the Relu function. The second layer does not use an activation function. Its output is: , where,/> and/> are the matrix and weight of the third layer respectively,/> and are the matrix and weight of the fourth layer respectively. The purpose of normalization and accumulation of the last layer is consistent with the purpose of normalization and accumulation mentioned above. Therefore, the final result corresponding to the multi-head attention model is:/> .

S147: Aggregate the final results of each multi-head attention model, and use the normalized exponential function to output the aggregated final results to obtain the classification results of the temporal Transformer neural network and obtain the initial load identification model.

Furthermore, the final results of each multi-head attention model are aggregated through the output gate. It should be noted that, unlike the Convolutional Neural Networks (CNN) model that processes time series, this embodiment uses a two-dimensional convolution kernel to focus on both step quantization (step-wise) and channel-wise quantization (channel-wise). ), that is, using the two-tower model, calculate step-wise Attention and channel-wise Attention at the same time. Finally, the normalized exponential function is used Output the final result after aggregation to obtain the classification results of the time series Transformer neural network and obtain the initial load identification model.

S148: Obtain a new weight matrix based on the classification result and the cross-entropy loss function.

S149: Backpropagation based on the cross-entropy loss function and the new weight matrix.

S150: Determine whether the load identification model meets the preset requirements; if so, end the training of the load identification model.

It can be understood that the above process is a forward process of the model. In order to improve the accuracy of classification, the process of reverse iteration updating parameters is added to the temporal Transformer neural network. This embodiment specifically obtains a new weight matrix based on the classification results and the cross-entropy loss function. . Perform several backpropagations based on the cross-entropy loss function and the new weight matrix. Determine whether the load identification model meets the preset requirements; if so, the model is considered to have been constructed and the training of the load identification model is completed. In this embodiment, there is no restriction on the specific process of determining whether the load identification model meets the preset requirements, and it depends on the specific implementation situation. In this way, the construction of the load identification model is realized.

Based on the above embodiments, in some embodiments, determining whether the load identification model meets the preset requirements includes:

S151: Verify the load identification model based on the data in the test set, and determine whether the accuracy of the load identification model is greater than the threshold; if yes, proceed to step S152; if not, proceed to step S153.

S152: Confirm that the load identification model meets the preset requirements.

S153: Confirm that the load identification model does not meet the preset requirements.

Specifically, in order to determine whether the load identification model meets the requirements, the load identification model can be verified based on the data in the test set to determine whether the accuracy of the load identification model is greater than the threshold. In this embodiment, there is no limit to the threshold size. For example, it can be set to 95%, depending on the specific implementation situation. If it is greater than the threshold, it is confirmed that the load identification model meets the preset requirements. If it is not greater than the threshold, it is confirmed that the load identification model does not meet the preset requirements. In this way, it can be judged whether the model meets the requirements.

Based on the above embodiments, in some embodiments, before reading the target power consumption data through the electric energy meter, the method further includes:

S16: Deploy the load identification model in the electric energy meter;

S17: Start the electric energy meter electrical signal detection sensor of the electric energy meter so as to obtain the target power consumption data through the electric energy meter electrical signal detection sensor.

In specific implementation, in order to implement the application of the load identification model, before reading the target power consumption data through the electric energy meter, the load identification model also needs to be deployed in the electric energy meter; at the same time, the electric energy meter electrical signal detection sensor of the electric energy meter is started. , in order to obtain the target electricity consumption data through the electric energy meter electrical signal detection sensor, so as to accurately estimate and identify the user's electricity consumption.

Based on the above embodiments, in some embodiments, after obtaining the target load type and target power consumption, it also includes:

S18: Generate target load curve and target load statistical information based on target load type and target power consumption;

S19: Store the target load curve and target load statistical information in the storage space of the electric energy meter.

In order to better analyze the user's power consumption, in specific implementation, the target load curve and target load statistical information can also be generated based on the finally obtained target load type and target power consumption. It can be understood that the target load curve shows the user's power usage in the form of an image curve, and the target load statistical information shows the user's power usage in the form of a data table, so that the user can learn the power usage more intuitively. The target load curve and target load statistical information are stored in the storage space of the electric energy meter, thereby better preserving the target load curve and target load statistical information.

In addition, in some embodiments, after obtaining the target load type and target power consumption, it also includes:

S20: Generate acquisition logs of target load type and target power consumption;

S21: Store the log in the storage space of the electric energy meter.

It can be understood that after obtaining the target load type and target power consumption, a log can be generated for this acquisition process of the target load type and target power consumption, and the log can be stored in the storage space of the electric energy meter, so as to facilitate Staff can obtain information related to each acquisition process of target load type and target power consumption from the storage space.

In the above embodiments, the method for obtaining the electric load of an electric energy meter is described in detail. This application also provides a corresponding embodiment of a device for obtaining the electric load of an electric energy meter.

Figure 3 is a schematic diagram of an electric load acquisition device for an electric energy meter provided by an embodiment of the present application. As shown in Figure 3, the device includes:

The reading module 10 is used to read the target power consumption data through the electric energy meter; wherein the power consumption data includes target current data and target voltage data.

The differential processing module 11 is used to perform differential processing on the target current data and target voltage data respectively.

The prediction module 12 is used to input the processed target current data and target voltage data into the load identification model to obtain the target load type and target power consumption.

Among them, the construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; preprocessing the current data and voltage data respectively; based on the preprocessed current data and voltage data, using the time series Transformer neural network to perform load analysis Training of the recognition model to obtain the load recognition model.

In this embodiment, the electrical load acquisition device of the electric energy meter includes a reading module, a differential processing module and a prediction module. When the electric energy meter electric load acquisition device is running, it can realize all the steps of the above electric energy meter electric load acquisition method. Read the target power consumption data through the electric energy meter; wherein the power consumption data includes target current data and target voltage data; perform differential processing on the target current data and target voltage data respectively; input the processed target current data and target voltage data to In the load identification model, in order to obtain the target load type and target electricity consumption; among them, the construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; preprocessing the current data and voltage data respectively; based on the pre-processing The processed current data and voltage data are trained on the load identification model using the time series Transformer neural network to obtain the load identification model. It can be seen that the above scheme builds a load identification model through a time-series Transformer neural network and has a multi-head attention mechanism; so that when using the load identification model to predict the load type of the target power consumption data, each element in the target power consumption data can be Full attention and processing are carried out to improve the accuracy and robustness of the model, solve the problem of low accuracy of current electric load identification by electric energy meters, and improve the utilization efficiency of electric load and the stability of the power grid.

Figure 4 is a schematic diagram of an electric load acquisition device for an electric energy meter provided by an embodiment of the present application. As shown in Figure 4, the electric load acquisition equipment of the electric energy meter includes:

Memory 20 is used to store computer programs.

The processor 21 is configured to implement the steps of the method for obtaining the electrical load of an electric energy meter as mentioned in the above embodiment when executing a computer program.

The electric energy meter electric load acquisition device provided in this embodiment may include but is not limited to a smart phone, a tablet computer, a notebook computer or a desktop computer.

The processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 may adopt at least one hardware form among a digital signal processor (Digital Signal Processor, DSP), a field-programmable gate array (Field-Programmable Gate Array, FPGA), and a programmable logic array (Programmable Logic Array, PLA). accomplish. The processor 21 may also include a main processor and a co-processor. The main processor is a processor used to process data in the wake-up state, also called a central processing unit (Central Processing Unit, CPU); the co-processor is A low-power processor used to process data in standby mode. In some embodiments, the processor 21 may be integrated with a graphics processor (Graphics Processing Unit, GPU), and the GPU is responsible for rendering and drawing content to be displayed on the display screen. In some embodiments, the processor 21 may also include an artificial intelligence (Artificial Intelligence, AI) processor, which is used to process computing operations related to machine learning.

Memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory, and non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used to store the following computer program 201. After the computer program is loaded and executed by the processor 21, the relevant steps of the electric energy meter electric load acquisition method disclosed in any of the foregoing embodiments can be implemented. In addition, the resources stored in the memory 20 may also include the operating system 202, data 203, etc., and the storage method may be short-term storage or permanent storage. Among them, the operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, data related to the method for obtaining the electric load of the electric energy meter.

In some embodiments, the electric energy meter electric load acquisition device may also include a display screen 22, an input and output interface 23, a communication interface 24, a power supply 25 and a communication bus 26.

Those skilled in the art can understand that the structure shown in FIG. 4 does not constitute a limitation on the electric load acquisition device of the electric energy meter, and may include more or fewer components than shown in the figure.

In this embodiment, the electric energy meter electrical load acquisition device includes a memory and a processor. Memory is used to store computer programs. The processor is used to implement the steps of the method for obtaining the electrical load of the electric energy meter as mentioned in the above embodiment when executing the computer program. Read the target power consumption data through the electric energy meter; wherein the power consumption data includes target current data and target voltage data; perform differential processing on the target current data and target voltage data respectively; input the processed target current data and target voltage data to In the load identification model, in order to obtain the target load type and target electricity consumption; among them, the construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; preprocessing the current data and voltage data respectively; based on the pre-processing The processed current data and voltage data are trained on the load identification model using the time series Transformer neural network to obtain the load identification model. It can be seen that the above scheme builds a load identification model through a time-series Transformer neural network and has a multi-head attention mechanism; so that when using the load identification model to predict the load type of the target power consumption data, each element in the target power consumption data can be Full attention and processing are carried out to improve the accuracy and robustness of the model, solve the problem of low accuracy of current electric load identification by electric energy meters, and improve the utilization efficiency of electric load and the stability of the power grid.

Finally, this application also provides a corresponding embodiment of a computer-readable storage medium. The computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, the steps recorded in the above method embodiments are implemented.

It can be understood that if the methods in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program code. .

In this embodiment, a computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, the steps described in the above method embodiment are implemented. Read the target power consumption data through the electric energy meter; wherein the power consumption data includes target current data and target voltage data; perform differential processing on the target current data and target voltage data respectively; input the processed target current data and target voltage data to In the load identification model, in order to obtain the target load type and target electricity consumption; among them, the construction process of the load identification model includes: obtaining the current data and voltage data stored in the electric energy meter; preprocessing the current data and voltage data respectively; based on the pre-processing The processed current data and voltage data are trained on the load identification model using the time series Transformer neural network to obtain the load identification model. It can be seen that the above scheme builds a load identification model through a time-series Transformer neural network and has a multi-head attention mechanism; so that when using the load identification model to predict the load type of the target power consumption data, each element in the target power consumption data can be Full attention and processing are carried out to improve the accuracy and robustness of the model, solve the problem of low accuracy of current electric load identification by electric energy meters, and improve the utilization efficiency of electric load and the stability of the power grid.

The method, device, equipment and medium for obtaining the electric load of an electric energy meter provided by this application have been introduced in detail above. Each embodiment in the specification is described in a progressive manner. Each embodiment focuses on its differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. As for the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For relevant details, please refer to the description in the method section. It should be noted that for those of ordinary skill in the art, several improvements and modifications can be made to the present application without departing from the principles of the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

It should also be noted that in this specification, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or operations There is no such actual relationship or sequence between operations. Furthermore, the terms "comprises," "comprises," or any other variations thereof are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements, but also those not expressly listed other elements, or elements inherent to the process, method, article or equipment. Without further limitation, an element defined by the statement "comprises a..." does not exclude the presence of additional identical elements in a process, method, article, or apparatus that includes the stated element.

Claims (8)

1. The method for acquiring the electric load of the electric energy meter is characterized by comprising the following steps of:

reading target electricity consumption data through an electric energy meter; wherein the electricity consumption data comprises target current data and target voltage data;

respectively carrying out differential processing on the target current data and the target voltage data;

inputting the processed target current data and the target voltage data into a load identification model to obtain a target load type and target power consumption; the construction process of the load identification model comprises the following steps: acquiring current data and voltage data stored by the electric energy meter; respectively carrying out normalization processing on the current data and the voltage data; segmenting the normalized current data and the normalized voltage data according to time sequence to obtain a plurality of sequences of the current data and the voltage data; labeling the current data and the voltage data of each sequence with a corresponding load type; dividing the current data and the voltage data marked with the load types into a training set and a testing set according to a preset proportion; inputting the data in the training set into each multi-head attention model; the multi-head attention model comprises a first multi-head attention model and a second multi-head attention model, and the first multi-head attention model adopts shielding operation; in each multi-head attention model, acquiring training feature matrixes of a preset number of load types according to data in the training set, and acquiring an initial weight matrix; acquiring the weight of the corresponding self-attention layer according to the training feature matrix and the initial weight matrix to obtain an output result of the corresponding self-attention layer; aggregating the output results of the self-attention layers corresponding to the load types to obtain an aggregate output result of the self-attention layer; acquiring an output result of the corresponding multi-head attention model according to the weight coefficient and the aggregation output result; processing the output results of the multi-head attention models respectively to obtain the corresponding final results of the multi-head attention models; aggregating the final results of the multi-head attention models, and outputting the aggregated final results by using a normalized exponential function to obtain a classification result of a time sequence transducer neural network, thereby obtaining an initial load identification model; acquiring a new weight matrix according to the classification result and the cross entropy loss function; back-propagating according to the cross entropy loss function and the new weight matrix; judging whether the load identification model meets preset requirements or not; if yes, finishing training of the load identification model to obtain the load identification model.

2. The method for obtaining an electric load of an electric energy meter according to claim 1, wherein the determining whether the load identification model meets a preset requirement comprises:

verifying the load identification model according to the data in the test set, and judging whether the accuracy of the load identification model is greater than a threshold;

if yes, confirming that the load identification model meets the preset requirement;

if not, confirming that the load identification model does not meet the preset requirement.

3. The electric energy meter electricity load acquisition method according to claim 1, characterized by further comprising, before the reading of the target electricity data by the electric energy meter:

deploying the load identification model in the electric energy meter;

and starting an electric energy meter electric signal detection sensor of the electric energy meter so as to obtain the target electricity utilization data through the electric energy meter electric signal detection sensor.

4. The electric energy meter electric load acquisition method according to claim 1, further comprising, after the obtaining the target load type and the target electric power consumption amount:

generating a target load curve and target load statistical information according to the target load type and the target power consumption;

and storing the target load curve and the target load statistical information in a storage space of the electric energy meter.

5. The electric energy meter electric load obtaining method according to any one of claims 1 to 4, characterized by further comprising, after the obtaining of the target load type and the target electric power consumption:

generating an acquisition log of the target load type and the target power consumption;

and storing the log in a storage space of the electric energy meter.

6. An electric energy meter electricity load acquisition device, characterized by comprising:

the reading module is used for reading the target electricity consumption data through the electric energy meter; wherein the electricity consumption data comprises target current data and target voltage data;

the differential processing module is used for respectively carrying out differential processing on the target current data and the target voltage data;

the prediction module is used for inputting the processed target current data and the target voltage data into a load identification model so as to obtain a target load type and target power consumption; the construction process of the load identification model comprises the following steps: acquiring current data and voltage data stored by the electric energy meter; respectively carrying out normalization processing on the current data and the voltage data; segmenting the normalized current data and the normalized voltage data according to time sequence to obtain a plurality of sequences of the current data and the voltage data; labeling the current data and the voltage data of each sequence with a corresponding load type; dividing the current data and the voltage data marked with the load types into a training set and a testing set according to a preset proportion; inputting the data in the training set into each multi-head attention model; the multi-head attention model comprises a first multi-head attention model and a second multi-head attention model, and the first multi-head attention model adopts shielding operation; in each multi-head attention model, acquiring training feature matrixes of a preset number of load types according to data in the training set, and acquiring an initial weight matrix; acquiring the weight of the corresponding self-attention layer according to the training feature matrix and the initial weight matrix to obtain an output result of the corresponding self-attention layer; aggregating the output results of the self-attention layers corresponding to the load types to obtain an aggregate output result of the self-attention layer; acquiring an output result of the corresponding multi-head attention model according to the weight coefficient and the aggregation output result; processing the output results of the multi-head attention models respectively to obtain the corresponding final results of the multi-head attention models; aggregating the final results of the multi-head attention models, and outputting the aggregated final results by using a normalized exponential function to obtain a classification result of a time sequence transducer neural network, thereby obtaining an initial load identification model; acquiring a new weight matrix according to the classification result and the cross entropy loss function; back-propagating according to the cross entropy loss function and the new weight matrix; judging whether the load identification model meets preset requirements or not; if yes, finishing training of the load identification model to obtain the load identification model.

7. An electric energy meter electricity load acquisition device, characterized by comprising:

a memory for storing a computer program;

a processor for implementing the steps of the electric energy meter electric load acquisition method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, which when executed by a processor, implements the steps of the electric load acquisition method for an electric energy meter according to any one of claims 1 to 5.