Non-intrusive electrical load identification method and device, terminal equipment and storage medium

Abstract

The invention discloses a non-intrusive power utilization load identification method and device, terminal equipment and a storage medium. The method comprises the following steps: acquiring a first power utilization information data packet; wherein the first electricity utilization information data packet comprises current information and voltage information at each moment in a preset time period; generating an electricity utilization information feature code according to the first electricity utilization information data packet; inputting the electricity utilization information feature code into an electric appliance classification model, so that the electric appliance classification model outputs an electric appliance type; wherein the electric appliance classification model is constructed by the following steps of: acquiring an electricity utilization information feature code corresponding to each electric appliance type and a numerical value representation of each electric appliance type; and an initial neural network model is constructed, the electricity utilization information feature codes corresponding to the electric appliance types serve as input, the numerical value representation of the electric appliance types serves as output, the initial neural network model is trained until the initial neural network model converges, and the electric appliance classification model is generated.

Description

Technical Field

[0001] The present invention relates to the technical field of non-intrusive load identification, and in particular to a non-intrusive power load identification method, device, terminal equipment and storage medium. Background Art

[0002] When monitoring the load in the power system, it is usually necessary to use an intrusive method, that is, by installing additional sensors or meters to measure parameters such as current, voltage and power, and determine the type of electrical appliances used by the user through the measured power consumption data. This intrusive method requires the installation of additional instruments, which is not only costly, but also the installation of sensors or instrument lights in the power system for interference and modification may cause power outages or system instability. Due to the above-mentioned shortcomings of the intrusive method, with the improvement of computing power and the development of data analysis technology, load monitoring technology based on non-invasive methods has been born. The existing non-invasive method is to realize the recognition function through the collected high-frequency harmonic features and the decision tree judgment algorithm; however, this method of analyzing the collected high-frequency harmonic features through the decision tree consumes a lot of computing power.

[0003] Therefore, there is an urgent need for a method that does not require decision analysis of high-frequency harmonic features to reduce computing power consumption. Summary of the invention

[0004] The embodiments of the present invention provide a non-invasive power load identification method, apparatus, terminal device and storage medium, which can avoid the analysis of high-frequency harmonic characteristics in power information data packets and reduce the computing power consumed in the analysis of high-frequency harmonic characteristics.

[0005] An embodiment of the present invention provides a non-intrusive power load identification method, comprising:

[0006] Acquire a first power usage information data packet; wherein the first power usage information data packet includes current information and voltage information at each moment in a preset time period;

[0007] generating a power usage information characteristic code according to the first power usage information data packet;

[0008] Inputting the power consumption information feature code into an electrical appliance classification model so that the electrical appliance classification model outputs the electrical appliance type;

[0009] The electrical appliance classification model is constructed in the following manner:

[0010] Obtain the power consumption information characteristic code corresponding to each type of electrical appliance and the numerical representation of each type of electrical appliance;

[0011] An initial neural network model is constructed, with the power consumption information feature code corresponding to each appliance type as input and the numerical representation of each appliance type as output, and the initial neural network model is trained until the initial neural network model converges to generate the appliance classification model.

[0012] Further, generating the power usage information characteristic code according to the first power usage information data packet includes:

[0013] Segmenting the first power usage information data packet in a time dimension to obtain a plurality of second power usage information data packets representing the segmented power usage information data packets;

[0014] Performing feature coding on each second power usage information data packet to obtain power usage information feature coding in a plurality of time dimensions;

[0015] The power consumption information feature codes of several time dimensions are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

[0016] Furthermore, before processing the characteristic encoding of the power consumption information of several time dimensions according to the multi-head perceptron algorithm, it also includes:

[0017] For each time dimension of the electricity consumption information feature code, a corresponding blank feature code is established;

[0018] The method of processing the power consumption information feature codes of several time dimensions according to the multi-head perceptron algorithm to generate the power consumption information feature codes includes:

[0019] The power consumption information feature codes of several time dimensions and the corresponding blank feature codes are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

[0020] Furthermore, the segmenting of the first power usage information data packet in the time dimension includes:

[0021] Obtaining the data packet length of the first power usage information data packet;

[0022] If the length of the data packet is greater than a preset length, the first power usage information data packet is segmented in the time dimension by a fine segmentation method;

[0023] If the length of the data packet is not greater than the preset length, when the first power usage information data packet contains high-frequency short-time transient power usage characteristics, the first power usage information data packet is segmented in the time dimension using redundant segmentation.

[0024] Furthermore, the feature encoding of each second power usage information data packet to obtain power usage information feature encodings in several time dimensions includes:

[0025] The features of each second power consumption information data packet are extracted through a two-dimensional convolution kernel, and the power consumption information features of several time dimensions are encoded.

[0026] Based on the above method embodiment, the present invention provides a corresponding device embodiment;

[0027] An embodiment of the present invention provides a non-intrusive power load identification device, including: a data acquisition module, a feature encoding module and a classification module;

[0028] The data acquisition module is used to acquire a first power usage information data packet; wherein the first power usage information data packet includes current information and voltage information at each moment in a preset time period;

[0029] The feature coding module is used to generate a feature code of power usage information according to the first power usage information data packet;

[0030] The classification module is used to input the power consumption information feature code into the electrical appliance classification model so that the electrical appliance classification model outputs the electrical appliance type; wherein the electrical appliance classification model is constructed in the following manner: obtaining the power consumption information feature code corresponding to each electrical appliance type and the numerical representation of each electrical appliance type; constructing an initial neural network model, using the power consumption information feature code corresponding to each electrical appliance type as input and the numerical representation of each electrical appliance type as output, training the initial neural network model until the initial neural network model converges, thereby generating the electrical appliance classification model.

[0031] Further, generating the power usage information characteristic code according to the first power usage information data packet includes:

[0032] Segmenting the first power usage information data packet in a time dimension to obtain a plurality of second power usage information data packets representing the segmented power usage information data packets;

[0033] Performing feature coding on each second power usage information data packet to obtain power usage information feature coding in a plurality of time dimensions;

[0034] The power consumption information feature codes of several time dimensions are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

[0035] Furthermore, the segmenting of the first power usage information data packet in the time dimension includes:

[0036] Obtaining the data packet length of the first power usage information data packet;

[0037] If the length of the data packet is greater than a preset length, the first power usage information data packet is segmented in the time dimension by a fine segmentation method;

[0038] If the length of the data packet is not greater than the preset length, when the first power usage information data packet contains high-frequency short-time transient power usage characteristics, the first power usage information data packet is segmented in the time dimension using redundant segmentation.

[0039] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a non-intrusive power load identification method described in the above-mentioned embodiment of the invention is implemented.

[0040] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the non-intrusive power load identification method described in the above-mentioned embodiment of the invention.

[0041] The following beneficial effects are achieved by implementing the present invention:

[0042] The present invention provides a non-invasive power load identification method, device, terminal equipment and storage medium. The power load identification method obtains a first power information data packet of current information and voltage information of a user at each moment in a preset time period, generates a feature code based on the obtained power information data packet, and then identifies the type of electrical appliance based on the feature code based on the generated feature code and a pre-trained electrical appliance classification model; by inputting the feature code into the electrical appliance classification model for prediction, the analysis of high-frequency harmonic features in the power information data packet is avoided, and the computing power consumed in the analysis of high-frequency harmonic features is reduced; further, when constructing the electrical appliance classification model, the power information feature code corresponding to each electrical appliance type and the numerical representation of each electrical appliance type are used as training data to train the electrical appliance classification model, so that the model learns the relationship between the power information feature code and the numerical representation of the electrical appliance type. Based on the feature that the numerical representation of the electrical appliance type is a fixed numerical range, the training data can be reduced during training, and the computing power consumption during model training is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a first flow chart of a non-invasive power load identification method provided by an embodiment of the present invention.

[0044] Figure 2 This is a second flow chart of a non-invasive power load identification method provided by an embodiment of the present invention.

[0045] Figure 3 It is a schematic diagram of a hardware architecture for implementing the method of the present invention provided by an embodiment of the present invention.

[0046] Figure 4 It is a schematic diagram of feature mapping using a two-dimensional convolution kernel provided in one embodiment of the present invention.

[0047] Figure 5 It is a schematic diagram of a multi-head perceptron processing flow provided by an embodiment of the present invention.

[0048] Figure 6 It is a structural schematic diagram of a non-intrusive power load identification device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] like Figure 1 and Figure 2 As shown, a non-intrusive power load identification method provided by an embodiment of the present invention includes:

[0051] Step S1: Acquire a first power usage information data packet; wherein the first power usage information data packet includes current information and voltage information at each moment in a preset time period;

[0052] Step S2: generating a power usage information characteristic code according to the first power usage information data packet;

[0053] Step S3: inputting the power consumption information feature code into the electrical appliance classification model so that the electrical appliance classification model outputs the electrical appliance type;

[0054] The electrical appliance classification model is constructed in the following manner:

[0055] Obtain the power consumption information characteristic code corresponding to each type of electrical appliance and the numerical representation of each type of electrical appliance;

[0056] An initial neural network model is constructed, with the power consumption information feature code corresponding to each appliance type as input and the numerical representation of each appliance type as output, and the initial neural network model is trained until the initial neural network model converges to generate the appliance classification model.

[0057] For step S1, current information and voltage information at each moment in a preset time period are obtained to form a first power usage information data packet. Specifically, the present invention is based on the following Figure 3The overall hardware architecture shown is implemented, the metering unit obtains the instantaneous current information at the target user's home entrance through current sensing paths such as Rogowski coils, obtains the instantaneous voltage information at the user's home entrance through voltage parallel access and other voltage sensing paths, and forms a first power usage information data packet based on the acquired voltage instantaneous information and current instantaneous information within a preset time period, and sends the first power usage information data packet to the central processing chip.

[0058] It should be noted that, in order to meet the needs of users in different scenarios, when generating the first power usage information data packet, the length of the generated data packet can be set by the user according to the needs of their own scenarios, that is, the above-mentioned preset time period can be set according to the needs of the user, and there is no restriction here. After obtaining the current information and voltage information at each moment according to the preset time length set by the user, the current information and voltage information at the same moment are stored in the data packet in the form of two-dimensional data, and the data format usually uses the float type.

[0059] For step S2, after the central processing chip receives the first power usage information data packet, it is necessary to perform data preprocessing on the first power usage information data packet, specifically: firstly, the form of the first power usage information data packet is verified by data type comparison and data length comparison to determine whether the form of the first power usage information data packet meets the requirements. After the form of the first power usage information data packet meets the requirements, the first power usage information data packet is subjected to data cleaning operations such as outlier removal and smoothing to obtain a smoother first power usage information data packet without isolated outliers.

[0060] In a preferred embodiment, generating an electricity usage information feature code based on the first electricity usage information data packet includes: segmenting the first electricity usage information data packet in the time dimension to obtain a plurality of second electricity usage information data packets representing the segmented electricity usage information data packets; feature encoding each second electricity usage information data packet to obtain electricity usage information feature codes of a plurality of time dimensions; processing the electricity usage information feature codes of a plurality of time dimensions according to a multi-head perceptron algorithm to generate an electricity usage information feature code.

[0061] Specifically, firstly, data segmentation is performed on the first power usage information data packet in the time dimension to obtain a plurality of second power usage information data packets representing the segmented power usage information data packets.

[0062] In a preferred embodiment, the segmentation of the first power usage information data packet in the time dimension includes: obtaining the data packet length of the first power usage information data packet; if the data packet length is greater than a preset length, segmenting the first power usage information data packet in the time dimension using a fine segmentation method; if the data packet length is not greater than the preset length, when the first power usage information data packet contains high-frequency short-time transient power usage characteristics, segmenting the first power usage information data packet in the time dimension using a redundant segmentation method.

[0063] Specifically, when the first power information data packet is segmented, according to the segmentation requirements, it can be divided into two data segmentation methods: precise segmentation and redundant segmentation. The main basis for determining which data segmentation method to use is the data packet length of the first power information data packet. If the data packet length is too long (i.e., greater than the above-mentioned preset length), then the redundant segmentation method will cause the first power information data packet with a large data volume to further increase the data volume after the redundant segmentation. Therefore, when the data packet length is too long, the first power information data packet is segmented by a fine segmentation method. For example: if the first power information data packet is composed of 1000 sampling voltage and current data at the time dimension, the fine segmentation method is used, and the segmentation length during fine segmentation is set to 100, then the first power information data packet is segmented to obtain 10 sub-power information data packets with a length of 100. If the redundant segmentation method is used, the redundancy is set to 5, then the first power information data packet is segmented to obtain 10 sub-power information data packets with a length of 110. It can be seen that when the data packet length is large, although the redundant segmentation method can obtain a better feature extraction effect, it will bring a greater computing burden to the central processing chip. Therefore, when the feature extraction effect is not considered, the fine segmentation method can be directly selected to segment the data. When the feature extraction effect is considered, the data segmentation method to be selected is determined in combination with the length of the data packet.

[0064] When the length of the data packet is small (i.e. not greater than the above-mentioned preset length), the data segmentation method to be used can be determined based on the power consumption characteristics of the first power consumption information data packet. When the first power consumption information data packet contains more high-frequency short-time transient power consumption characteristics, or in complex scenarios with a wide variety of electrical appliances, such as factories with stamping machines, shared rental houses, and large shopping malls, the power consumption information data packets obtained will contain more high-frequency short-time transient power consumption characteristics. In this case, selecting the redundant segmentation method to process the data can obtain better feature extraction effects.

[0065] It should be added that, since there may be a large number of second power usage information data packets obtained after data segmentation, before feature encoding the second power usage information data packets, the data packets to be feature encoded can be selected according to the length of the data packets; for example: when the length of the data packet is 1000, all of them can be selected; when the length of the data packet is 10000, 100 samples can be intercepted for selection every 900 samples, and 10 portions of 100 samples can be intercepted in total to obtain a data packet with a total length of 1000 samples. In this way, when the computing power of the embedded system is limited, it is possible to better identify the data packets of loads with long-term power usage characteristics.

[0066] After the data segmentation is completed, feature encoding is performed on the second power usage information data packet.

[0067] In a preferred embodiment, the feature encoding of each second power usage information data packet to obtain power usage information feature encodings of several time dimensions includes: feature extraction of each second power usage information data packet through a two-dimensional convolution kernel, and power usage information feature encodings of several time dimensions.

[0068] Specifically, Figure 4 As shown, the present invention uses a two-dimensional convolution kernel to extract features from the segmented data packets (i.e., the second power information data packets mentioned above). The essence of the feature extraction process is the convolution operation between the second power information data packets and the convolution kernel, and the feature codes corresponding to each second power information data packet are obtained through convolution calculation. It should be noted that the two-dimensional convolution kernel is obtained through big data training using its learning algorithm, and in the present invention, it is mounted in the central processing chip, so that the two-dimensional convolution kernel can directly extract the features of the second power information data packet and obtain the feature code by feature mapping, and the computing power consumed by the feature mapping process is relatively small, thereby saving computing power consumption.

[0069] In a preferred embodiment, before processing the electricity usage information feature codes of several time dimensions according to the multi-headed perceptron algorithm, it also includes: establishing corresponding blank feature codes for the electricity usage information feature codes of each time dimension; the processing of the electricity usage information feature codes of several time dimensions according to the multi-headed perceptron algorithm to generate the electricity usage information feature codes includes: processing the electricity usage information feature codes of several time dimensions and the corresponding blank feature codes according to the multi-headed perceptron algorithm to generate the electricity usage information feature codes.

[0070] Specifically, the central processing chip creates a blank feature code of the same size with all data being 0 based on the power consumption information feature code of each time dimension. The established blank feature codes and the power consumption information feature codes of each time dimension are processed by the multi-head perceptron algorithm deployed in the central processing chip. The processing flow is as follows: Figure 5As shown, first, the electricity consumption information feature codes of all time dimensions and the corresponding blank feature codes are normalized, and the normalized data are further normalized after being processed by the multi-head perceptron similarity matching mechanism to obtain the matched feature code (i.e., the above-mentioned electricity consumption information feature code), which is in the form of a one-dimensional data matrix.

[0071] In step S3, the power consumption information feature code obtained above is input into the electrical appliance classification model, so that the electrical appliance classification model outputs the corresponding electrical appliance type.

[0072] Specifically, the electrical appliance classification model is constructed by acquiring the characteristic codes of electricity consumption information corresponding to each electrical appliance type and the numerical representation of each electrical appliance type, and then constructing an initial neural network model. The initial neural network model is trained with the acquired training data so that the initial neural network model learns the relationship between the characteristic codes of electricity consumption information corresponding to each electrical type and the numerical representation of each electrical appliance type, until the model reaches the preset convergence conditions, thereby generating the electrical appliance classification model.

[0073] After the power consumption information feature code is input into the appliance classification model, the output of the appliance classification model is essentially a numerical representation used to characterize the appliance type. The corresponding appliance type can be determined by consulting the corresponding table according to the numerical representation defined during model training. For example: 256 appliance types are represented by a numerical value of 0-255, and the corresponding appliance type can be determined according to the numerical value output by the classification model. After obtaining the appliance type, it is output through a communication module directly connected to the central processing chip. In some scenarios, the results can be temporarily stored in Flash as needed, and the communication module will output them in batches after the set conditions are met.

[0074] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0075] like Figure 6 As shown, an embodiment of the present invention provides a non-intrusive power load identification device, including: a data acquisition module, a feature encoding module and a classification module;

[0076] The data acquisition module is used to acquire a first power usage information data packet; wherein the first power usage information data packet includes current information and voltage information at each moment in a preset time period;

[0077] The feature coding module is used to generate a feature code of power usage information according to the first power usage information data packet;

[0078] The classification module is used to input the power consumption information feature code into the electrical appliance classification model so that the electrical appliance classification model outputs the electrical appliance type; wherein the electrical appliance classification model is constructed in the following manner: obtaining the power consumption information feature code corresponding to each electrical appliance type and the numerical representation of each electrical appliance type; constructing an initial neural network model, using the power consumption information feature code corresponding to each electrical appliance type as input and the numerical representation of each electrical appliance type as output, training the initial neural network model until the initial neural network model converges, thereby generating the electrical appliance classification model.

[0079] In a preferred embodiment, generating the power usage information characteristic code according to the first power usage information data packet includes:

[0080] Segmenting the first power usage information data packet in a time dimension to obtain a plurality of second power usage information data packets representing the segmented power usage information data packets;

[0081] Performing feature coding on each second power usage information data packet to obtain power usage information feature coding in a plurality of time dimensions;

[0082] The power consumption information feature codes of several time dimensions are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

[0083] In a preferred embodiment, the segmenting of the first power usage information data packet in the time dimension includes:

[0084] Obtaining the data packet length of the first power usage information data packet;

[0085] If the length of the data packet is greater than a preset length, the first power usage information data packet is segmented in the time dimension by a fine segmentation method;

[0086] If the length of the data packet is not greater than the preset length, when the first power usage information data packet contains high-frequency short-time transient power usage characteristics, the first power usage information data packet is segmented in the time dimension using redundant segmentation.

[0087] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0088] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0089] Based on the above method item embodiments, the present invention provides corresponding terminal device item embodiments.

[0090] An embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a non-intrusive power load identification method as described in any one of the present invention is implemented.

[0091] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0092] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0093] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med i aCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0094] Based on the above method item embodiments, the present invention provides a corresponding storage medium item embodiment.

[0095] An embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute any one of the non-intrusive power load identification methods described in the present invention.

[0096] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0097] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A non-intrusive power load identification method, characterized in that: include: Acquire a first power usage information data packet; wherein the first power usage information data packet includes current information and voltage information at each moment in a preset time period; generating a power usage information characteristic code according to the first power usage information data packet; Inputting the power consumption information feature code into an electrical appliance classification model so that the electrical appliance classification model outputs the electrical appliance type; The electrical appliance classification model is constructed in the following manner: Obtain the power consumption information characteristic code corresponding to each type of electrical appliance and the numerical representation of each type of electrical appliance; An initial neural network model is constructed, with the power consumption information feature code corresponding to each appliance type as input and the numerical representation of each appliance type as output, and the initial neural network model is trained until the initial neural network model converges to generate the appliance classification model.

2. A non-intrusive power load identification method as claimed in claim 1, characterized in that: The step of generating a power usage information characteristic code according to the first power usage information data packet includes: Segmenting the first power usage information data packet in a time dimension to obtain a plurality of second power usage information data packets representing the segmented power usage information data packets; Performing feature coding on each second power usage information data packet to obtain power usage information feature coding in a plurality of time dimensions; The power consumption information feature codes of several time dimensions are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

3. A non-intrusive power load identification method as claimed in claim 2, characterized in that: Before processing the characteristic encoding of the power consumption information of several time dimensions according to the multi-head perceptron algorithm, it also includes: For each time dimension of the electricity consumption information feature code, a corresponding blank feature code is established; The method of processing the power consumption information feature codes of several time dimensions according to the multi-head perceptron algorithm to generate the power consumption information feature codes includes: The power consumption information feature codes of several time dimensions and the corresponding blank feature codes are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

4. A non-intrusive power load identification method as claimed in claim 2 or 3, characterized in that: The segmenting of the first power usage information data packet in the time dimension includes: Obtaining the data packet length of the first power usage information data packet; If the length of the data packet is greater than a preset length, the first power usage information data packet is segmented in the time dimension by a fine segmentation method; If the length of the data packet is not greater than the preset length, when the first power usage information data packet contains high-frequency short-time transient power usage characteristics, the first power usage information data packet is segmented in the time dimension using a redundant segmentation method.

5. A non-intrusive power load identification method as claimed in claim 2 or 3, characterized in that: The feature encoding of each second power usage information data packet to obtain power usage information feature encodings in several time dimensions includes: The features of each second power consumption information data packet are extracted through a two-dimensional convolution kernel, and the power consumption information features of several time dimensions are encoded.

6. A non-intrusive electrical load identification device, characterized in that: include: Data acquisition module, feature encoding module and classification module; The data acquisition module is used to acquire a first power usage information data packet; wherein the first power usage information data packet includes current information and voltage information at each moment in a preset time period; The feature coding module is used to generate a feature code of power usage information according to the first power usage information data packet; The classification module is used to input the power consumption information feature code into the electrical appliance classification model so that the electrical appliance classification model outputs the electrical appliance type; wherein the electrical appliance classification model is constructed in the following manner: obtaining the power consumption information feature code corresponding to each electrical appliance type and the numerical representation of each electrical appliance type; constructing an initial neural network model, using the power consumption information feature code corresponding to each electrical appliance type as input and the numerical representation of each electrical appliance type as output, training the initial neural network model until the initial neural network model converges, thereby generating the electrical appliance classification model.

7. A non-intrusive electrical load identification device as claimed in claim 6, characterized in that: The step of generating a power usage information characteristic code according to the first power usage information data packet includes: Segmenting the first power usage information data packet in a time dimension to obtain a plurality of second power usage information data packets representing the segmented power usage information data packets; Performing feature coding on each second power usage information data packet to obtain power usage information feature coding in a plurality of time dimensions; The power consumption information feature codes of several time dimensions are processed according to the multi-head perceptron algorithm to generate the power consumption information feature codes.

8. A non-intrusive electrical load identification device as claimed in claim 7, characterized in that: The segmenting of the first power usage information data packet in the time dimension includes: Obtaining the data packet length of the first power usage information data packet; If the length of the data packet is greater than a preset length, the first power usage information data packet is segmented in the time dimension by a fine segmentation method; If the length of the data packet is not greater than the preset length, when the first power usage information data packet contains high-frequency short-time transient power usage characteristics, the first power usage information data packet is segmented in the time dimension using a redundant segmentation method.

9. A terminal device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a non-intrusive power load identification method as claimed in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute a non-intrusive power load identification method as described in any one of claims 1 to 5.

Images