A non-invasive load decomposition improved method, device and storage medium

Through the improvement method based on the user behavior network diagram, the timing characteristics and user behavior characteristics are integrated, and the problem of insufficient utilization of timing characteristics and user behavior changes in the prior art is solved, and the accuracy and efficiency of non-invasive load monitoring are improved. For multiple types of electrical appliances, the model size has been reduced.

CN114692740BActive Publication Date: 2024-11-08SOUTH CHINA UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210275087.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-21
Publication Date
2024-11-08
Estimated Expiration
2042-03-21

AI Technical Summary

Technical Problem

The existing non-invasive load decomposition technology does not fully utilize the timing characteristics and does not consider changes in user electricity consumption behavior, resulting in insufficient classification accuracy and efficiency.

Method used

Adopting an improved method based on user behavior network diagram, by constructing a general network diagram of user behavior, extracting user behavior characteristics, and integrating behavior characteristics and power consumption characteristics, building a multi-label non-invasive load decomposition model, and updating the user behavior network diagram to adapt to Users' electricity usage habits change.

Benefits of technology

It improves the accuracy of non-invasive load monitoring, is suitable for multiple types of electrical appliances, reduces the scale of the model, and improves energy utilization efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114692740B_ABST
    Figure CN114692740B_ABST
Patent Text Reader

Abstract

The present invention discloses an improved method, device and storage medium for non-invasive load decomposition, wherein the method comprises: obtaining home appliance-level data, constructing a user behavior general network diagram based on the home appliance-level data; constructing an initial sub-graph based on the user behavior general network diagram, and extracting user behavior features; and constructing a multi-label non-invasive load decomposition model that integrates user behavior features. The improved model of the present invention based on the user behavior network diagram achieves an effective combination of multi-label classification and advanced time series feature extraction methods, thereby improving the accuracy of non-invasive load monitoring. In addition, the traditional model is only applicable to a single type of electrical appliances, while the present invention integrates multiple types of electrical appliances into one model, greatly reducing the scale of the total model. The present invention can be widely used in the field of non-invasive load decomposition.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of non-invasive load decomposition, and in particular to an improved method, device and storage medium for non-invasive load decomposition. Background Art

[0002] At present, energy shortage and pollution issues have received widespread attention from all countries. In this context, improving energy efficiency is particularly critical. In addition, the current proportion of global energy consumption in households and commercial buildings has exceeded 40%, and the level of electricity consumption of household users continues to rise. Existing studies have shown that this method of providing users with subtle electricity consumption information will save users 10%-20% of electricity consumption. Non-invasive load decomposition is a technology that relies solely on the aggregated data of smart meters to extract household appliance-level power consumption information. Due to its non-invasive and economic advantages, it has become the main means of load decomposition.

[0003] In the prior art, many non-invasive load decomposition methods have been proposed, especially time series models have reached a relatively advanced level in the field of non-invasive load decomposition. However, existing studies mainly complete the load decomposition of a single appliance through the power consumption characteristics of the appliance itself, and few studies have considered the correlation of the use behavior of each appliance. Some existing technologies use MLKNN and REKAL models to consider behavioral correlation for multi-label load classification, and some existing technologies also consider device correlation and complete multi-label classification through deep dictionary learning and deep learning. They can achieve better classification results with a small amount of data, but these existing technologies focus on correlation research and do not make full use of the current more advanced time series feature extraction methods, and do not consider changes in users' power consumption behavior. Summary of the invention

[0004] In order to solve at least one of the technical problems existing in the prior art to a certain extent, the object of the present invention is to provide an improved method, device and storage medium for non-intrusive load decomposition based on a user behavior network graph.

[0005] The technical solution adopted by the present invention is:

[0006] A non-intrusive load decomposition improvement method comprises the following steps:

[0007] Obtain appliance-level data and build a general network diagram of user behavior based on the appliance-level data;

[0008] Construct the initial subgraph based on the user behavior general network graph and extract user behavior features;

[0009] Construct a multi-label non-intrusive load decomposition model that integrates user behavior characteristics;

[0010] Among them, the multi-label non-intrusive load decomposition model refers to an improved multi-label version of the single-label load decomposition model based on the current various sequence-to-point structures. The model output layer is adjusted from single-appliance output to multi-appliance output, and the input and middle layer structures remain unchanged before and after the improvement; fusion of user behavior refers to the fusion of behavioral characteristics and electricity consumption characteristics at the shallow, middle and deep layers of the model.

[0011] Furthermore, the user behavior general network graph includes all electrical appliances appearing in the existing appliance-level data, the nodes of the graph represent various electrical appliances, the node features represent the feature information of the appliances, and the behavior correlation between the appliances is represented by the edge weights;

[0012] The weight represents the probability of the appliances appearing at the same time. The weight is calculated as follows:

[0013]

[0014] Where n j For device m j Number of occurrences, n ij Indicates device m i and m j The number of simultaneous occurrences, P ij Indicates device m j When the device m appears i The probability of occurrence, the final matrix P is called the behavior correlation matrix.

[0015] Furthermore, in order to avoid the influence of overfitting of the behavior correlation matrix and signal noise, the matrix P is binarized by setting a threshold:

[0016]

[0017] in, and are not equal, the final matrix It is called the general behavior correlation matrix, and the graphical form is called the general behavior network diagram. The behavior correlation matrix is ​​not symmetrical.

[0018] Furthermore, the construction of the initial subgraph refers to extracting a corresponding subgraph from the behavior general network graph according to the electrical appliances owned by the user, and using the extracted subgraph as the initial behavior network graph required by the network model;

[0019] The extraction method is as follows: modifying the edge weights connected to the nodes represented by the non-existent electrical appliances to 0, that is, deleting the related edges.

[0020] Furthermore, the extracting of user behavior features includes:

[0021] Based on the user behavior general network graph, a graph convolutional network is used to extract user behavior features;

[0022] The calculation formula of the graph convolutional network is:

[0023]

[0024] Among them, H l ∈|M|×D l is the feature matrix of each node in the lth layer, D l represents the node feature dimension of the lth layer, |M| represents the number of electrical appliances, is the normalized behavior correlation matrix, W l ∈D l ×D l+1 is the parameter to be learned by layer l during the training phase, and σ(·) is the nonlinear activation function Relu.

[0025] Furthermore, the deep fusion method is to directly perform matrix product of features, and the calculation methods of shallow and middle layers are:

[0026] y=Conv1d(x T H) T +x

[0027] Where x∈W×D is the extracted signal feature, W represents the length of the signal feature, D is the number of channels, H∈|M|×D represents the behavioral feature information of each device in each layer, |M| represents the number of electrical appliances, and Conv1d(·) is a one-dimensional convolution process.

[0028] Furthermore, the non-intrusive load decomposition improvement method further includes the step of updating the user behavior general network diagram according to the user's power consumption:

[0029] During the user's electricity consumption process, a behavior correlation matrix representing the user's real electricity consumption habits is regularly generated based on the load identification results. The real behavior correlation matrix is ​​a matrix that has not been binarized. The real behavior correlation matrix and the behavior general network diagram are combined for calculation to complete the update of the network diagram.

[0030] The update process is calculated as follows:

[0031]

[0032] Where A represents the behavior correlation matrix used for calculation, and A′ refers to the newly obtained behavior correlation matrix representing the user's recent real electricity usage habits.

[0033] Furthermore, when the user adds a new appliance midway, the nodes and related edges represented by the corresponding appliance are extracted from the user behavior general network diagram, and spliced ​​with the user's current behavior network diagram to add the new node; when the user deletes an appliance midway, the corresponding appliance node and related edges in the user's current behavior network diagram are directly deleted.

[0034] Another technical solution adopted by the present invention is:

[0035] A non-invasive load decomposition improvement device, comprising:

[0036] at least one processor;

[0037] at least one memory for storing at least one program;

[0038] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0039] Another technical solution adopted by the present invention is:

[0040] A computer-readable storage medium stores a program executable by a processor, wherein the program executable by the processor is used to execute the method described above when executed by the processor.

[0041] The beneficial effects of the present invention are as follows: the improved model based on the user behavior network diagram of the present invention achieves an effective combination of multi-label classification and advanced time series feature extraction methods, thereby improving the accuracy of non-intrusive load monitoring. The traditional model is only applicable to a single type of electrical appliances, while the present invention integrates multiple types of electrical appliances into one model, thereby greatly reducing the scale of the overall model. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the embodiments of the present invention or the drawings of related technical solutions in the prior art are introduced below. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 is a flowchart of the steps of an improved method for non-invasive load decomposition in an embodiment of the present invention;

[0044] Figure 2 is a schematic diagram of a user's electricity consumption behavior network diagram in an embodiment of the present invention;

[0045] Figure 3 It is a basic structural framework diagram of the non-intrusive load monitoring model in an embodiment of the present invention;

[0046] Figure 4 It is a schematic diagram of the updating process of the user's electricity consumption behavior network diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0048] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0049] In the description of the present invention, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0050] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0051] like Figure 1 As shown, this embodiment provides an improved method for non-intrusive load monitoring based on a user behavior network diagram, comprising the following steps:

[0052] Embodiment 1

[0053] S101. Complete the construction of a general network diagram of user behavior through existing home appliance-level data.

[0054] The existing home appliance-level data includes a large amount of long-term household electricity consumption data of various household appliances collected from public data sets or invasive experiments. This example uses a large amount of home appliance-level data including refrigerators, air conditioners, kettles, hair dryers and computers to complete the construction of a general network diagram of user behavior.

[0055] The nodes of the general network diagram of electricity consumption behavior represent various electrical devices, the node features represent the characteristic information of the devices, and the correlation between devices is represented by the weight of the edge. The weight represents the probability that when appliance 1 appears, appliance 2 also appears. The specific calculation method is:

[0056]

[0057] Where n j For device m j Number of occurrences, n ij Indicates device m i and m j The number of simultaneous occurrences, P ij Indicates device m j When the device m appears i The probability of occurrence, the final matrix P is called the behavior correlation matrix.

[0058] In order to prevent the overfitting of the behavior correlation matrix and the influence of signal noise, P is binarized by setting the threshold τ, that is:

[0059]

[0060] In this example, the threshold μ is equal to 0.3. The final matrix is It is called the general behavior correlation matrix, and the graphical form is called the general behavior network diagram. and are not equal, and the behavior correlation matrix is ​​not symmetric. The behavior correlation matrix is ​​asymmetric, and the constructed graph is a directed graph, such as Figure 2 Shown is a general network diagram of the behavior of the above five appliances.

[0061] S102: construct an initial subgraph based on the general behavior network graph and complete the extraction of user behavior features.

[0062] The construction of the initial subgraph refers to extracting the corresponding subgraph from the behavior general network diagram according to the appliances owned by the user, and using the subgraph as the initial behavior network diagram required by the model. The extraction method is: modifying the edge weights connected to the nodes represented by the non-existent appliances to 0, that is, deleting the relevant edges. This is because the general behavior network diagram contains a variety of common appliances, but a user does not own every appliance; for example, assuming that the general behavior network diagram has five appliances, such as Figure 2As shown; however, user 1 does not have an air conditioner at home, but only has four of the five types. Therefore, the air conditioner is the non-existent appliance. The edge weights of the air conditioners connected to each other are changed to 0 to avoid the influence of the non-existent appliance. Therefore, the user in this example has four appliances, namely a refrigerator, a kettle, a computer, and a hair dryer. Therefore, the air conditioner and the edges connected to the air conditioner in the general network graph are deleted, and the subgraph obtained is the initial behavior network graph of the user.

[0063] The method used for extracting user behavior features based on the user behavior network graph is the graph convolutional network (GCN). The calculation of the GCN layer can be expressed as:

[0064]

[0065] Among them, H l ∈|M|×D l is the feature matrix of each node in the lth layer, D l represents the node feature dimension of the lth layer, |M| represents the number of electrical appliances, is the normalized device correlation matrix, W l ∈D l ×D l+1 is the parameter to be learned by layer l during the training phase, and σ(·) is the nonlinear activation function Relu.

[0066] S103. Construct an improved multi-label non-intrusive load decomposition model that integrates behavioral features.

[0067] The multi-label non-intrusive load decomposition model refers to an improved multi-label version of the single-label load decomposition model based on various current sequence-to-point structures. The basic structure of the model is as follows: Figure 3 As shown in the figure, the original model used in this example is a convolutional network model (CNN). The shallow, middle and deep layers of the network in the figure are all composed of CNN layers. The input and middle layer structures of the model remain unchanged before and after the improvement. The output layer is adjusted from single-electrical output to multi-electrical output, and from the original single-value output to multi-value output. The output layer activation function is modified to the Sigmoid function. The function expression is:

[0068]

[0069] Fusion of behavioral features refers to the fusion of behavioral features and power consumption features at the shallow, middle and deep layers of the model, such as Figure 3 As shown. The deep fusion method is to directly perform matrix product of features, and the calculation method of shallow and middle layers is:

[0070] y=Conv1d(x T H) T +x

[0071] Among them, H∈|M|×D represents the behavioral feature information of each device in each layer. x∈W×D is the extracted signal feature, W represents the length of the signal feature, and D is the number of channels, which is the same as the number of channels in H. Conv1d(·) is a one-dimensional convolution process.

[0072] S104: Update the behavior network diagram based on the user's electricity usage.

[0073] The updating of the behavior network diagram refers to the periodic generation of a behavior correlation matrix that represents the user's real power usage habits based on the load identification results during the user's power usage process. The real behavior correlation matrix is ​​a matrix that has not been binarized. The real behavior correlation matrix and the behavior general network diagram are combined for calculation to complete the network diagram update. The updating process is as follows: Figure 4 As shown, the behavioral network Figure 1 is the network diagram before updating. The updated behavior network is obtained by combining the general behavior network diagram and the newly obtained behavior correlation matrix. Figure 2 , the figure shows an update process, and then the update process can be expressed as:

[0074]

[0075] Where A represents the behavior correlation matrix used for calculation, and A′ refers to the newly obtained behavior correlation matrix representing the user's recent real electricity usage habits.

[0076] Embodiment 2

[0077] This embodiment provides an improved method for non-intrusive load monitoring based on a user behavior network graph, comprising the following steps:

[0078] S201. Complete the construction of a general network diagram of user behavior through existing home appliance-level data; this example uses data containing more appliances, a total of 22 types of appliances, to complete the composition, and the method of constructing the network diagram is the same as the method used in Example 1.

[0079] S202. Construct an initial subgraph based on the general behavior network graph and extract the user behavior features. In this example, the user includes 9 types of electrical appliances. Similarly, delete the other 13 types of electrical appliances that do not appear and their connected edges to complete the extraction of the initial behavior network subgraph. The method for extracting behavior features is the same as that used in Example 1.

[0080] S203, constructing a multi-label non-intrusive load decomposition improvement model that integrates behavioral features; the original single-label model used in this example is the long short-term memory network model LSTM, and other model improvement processes are the same as those in Example 1.

[0081] S204, based on the user's electricity usage, the behavior network diagram is updated; the updating process is the same as that in the first embodiment. When the user adds an appliance midway and the appliance belongs to the 13 types of appliances deleted, the corresponding newly added appliance node and its related edges are extracted from the general behavior network diagram, and the extracted nodes and edges are spliced ​​with the user's currently updated behavior network diagram to add the corresponding appliance node. After that, the behavior of the spliced ​​behavior network diagram is continuously updated, and the updating method is the same as that in the first embodiment.

[0082] Embodiment 3

[0083] This embodiment provides an improved method for non-intrusive load monitoring based on a user behavior network graph, comprising the following steps:

[0084] S301. Complete the construction of a general network diagram of user behavior through existing home appliance-level data.

[0085] S302: construct an initial subgraph based on the general behavior network graph and complete the extraction of user behavior features.

[0086] S303, constructing a multi-label non-intrusive load decomposition improvement model that integrates behavioral features; the original single-label model used in this example is a bidirectional long short-term memory neural network BiLSTM, and other model improvement processes are the same as those in Example 1.

[0087] S304. Complete the behavior network diagram update based on the user's electricity usage; the update process is the same as that in Example 1. When the user deletes an electrical appliance during electricity usage, the corresponding electrical appliance node and its related edges in the user's current behavior network diagram are directly deleted. When the user replaces an electrical appliance of the same type during electricity usage, the behavior network diagram remains unchanged because the behavior and target object do not change. When different types of electrical appliances are replaced during electricity usage, the electrical appliance nodes and related edges are first deleted, and then the process of adding new appliances is carried out in Example 2. After that, the behavior of the modified behavior network diagram continues to be updated, and the update method is the same as in Example 1.

[0088] In summary, the embodiment of the present invention proposes an improved non-invasive load decomposition method based on a user behavior network graph, which constructs a user behavior general network graph through a large amount of home appliance-level data, performs multi-label modification on the existing single-label load monitoring model of the sequence-to-point structure, and constructs a non-invasive load decomposition improvement model that can integrate electricity consumption characteristics with behavior-related characteristics. As the user uses electricity, the user's behavior network graph will be combined with the general behavior network graph and the real behavior correlation matrix to regularly update the behavior, completing the self-learning of the user-specific behavior network graph. Compared with the existing methods, the improved model based on the user behavior network graph completes the effective combination of multi-label classification and advanced time series feature extraction methods, improves the accuracy of non-invasive load monitoring, and the traditional model is only applicable to a single type of electrical appliances. The present invention integrates multiple types of electrical appliances into one model, greatly reducing the scale of the total model.

[0089] This embodiment also provides a non-invasive load decomposition improvement device, comprising:

[0090] at least one processor;

[0091] at least one memory for storing at least one program;

[0092] When the at least one program is executed by the at least one processor, the at least one processor implements the above Figure 1 The method shown.

[0093] A non-invasive load decomposition improvement device of this embodiment can execute a non-invasive load decomposition improvement method provided by the method embodiment of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0094] The present application also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes Figure 1 The method shown.

[0095] This embodiment also provides a storage medium storing instructions or programs that can execute an improved non-invasive load decomposition method provided by an embodiment of the method of the present invention. When the instructions or programs are run, any combination of implementation steps of the method embodiment can be executed, and the corresponding functions and beneficial effects of the method can be obtained.

[0096] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0097] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0098] If the functions 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 invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0099] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0100] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0101] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0102] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0103] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0104] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A non-invasive load decomposition improvement method, characterized in that: The following steps are involved: Obtain appliance-level data and build a general network diagram of user behavior based on the appliance-level data; Construct the initial subgraph based on the user behavior general network graph and extract user behavior features; Construct a multi-label non-intrusive load decomposition model that integrates user behavior characteristics; Among them, the multi-label non-intrusive load decomposition model refers to an improved multi-label version of the single-label load decomposition model based on the sequence-to-point structure. The model output layer is adjusted from single-appliance output to multi-appliance output, and the input and middle layer structures remain unchanged before and after the improvement; the fusion of user behavior refers to the fusion of behavioral characteristics and power consumption characteristics at the shallow, middle and deep layers of the model; The user behavior general network graph includes all electrical appliances appearing in the appliance-level data. The nodes of the graph represent various electrical appliances. The node features represent the characteristic information of the appliances. The behavior correlation between the appliances is represented by the edge weights. The weight represents the probability of the appliances appearing at the same time. The weight is calculated as follows: Where n j For device m j Number of occurrences, n ij Indicates device m i and m j The number of simultaneous occurrences, P ij Indicates device m j When the device m appears i The probability of occurrence, the final matrix P is called the behavior correlation matrix; In order to avoid the influence of overfitting of the behavior correlation matrix and signal noise, the matrix P is binarized by setting a threshold: in, and are not equal, the final matrix It is called the general behavior correlation matrix, and the graphical form is called the general behavior network diagram. The behavior correlation matrix is ​​not symmetrical; The extracting of user behavior features includes: Based on the user behavior general network graph, a graph convolutional network is used to extract user behavior features; The calculation formula of the graph convolutional network is: Among them, H l ∈|M|×D l is the feature matrix of each node in the lth layer, D l represents the node feature dimension of the lth layer, |M| represents the number of electrical appliances, is the normalized behavior correlation matrix, W l ∈D l ×D l+1 is the parameter to be learned by layer l during the training phase, and σ(·) is the nonlinear activation function Relu.

2. The non-intrusive load decomposition improvement method according to claim 1, characterized in that: The construction of the initial subgraph refers to extracting a corresponding subgraph from the behavior general network graph according to the electrical appliances owned by the user, and using the extracted subgraph as the initial behavior network graph required by the network model; The extraction method is as follows: modifying the edge weights connected to the nodes represented by the non-existent electrical appliances to 0, that is, deleting the related edges.

3. The non-intrusive load decomposition improvement method according to claim 1, characterized in that: The deep fusion method is to directly perform matrix product of features, and the calculation methods of shallow and middle layers are: y=Conv1d(x T H) T +x Where x∈W×D is the extracted signal feature, W represents the length of the signal feature, D is the number of channels, H∈|M|×D represents the behavioral feature information of each device in each layer, |M| represents the number of electrical appliances, and Conv1d(·) is a one-dimensional convolution process.

4. The non-intrusive load decomposition improvement method according to claim 1, characterized in that: The non-intrusive load decomposition improvement method further includes the step of updating the user behavior general network diagram according to the user's power consumption: During the user's electricity consumption process, a behavior correlation matrix representing the user's real electricity consumption habits is regularly generated based on the load identification results. The real behavior correlation matrix is ​​a matrix that has not been binarized. The real behavior correlation matrix and the behavior general network diagram are combined for calculation to complete the update of the network diagram. The update process is calculated as follows: Where A represents the behavior correlation matrix used for calculation, and A′ refers to the newly obtained behavior correlation matrix representing the user's recent real electricity usage habits.

5. The non-invasive load decomposition improvement method according to claim 1, characterized in that: When a user adds a new appliance midway, the nodes and related edges represented by the corresponding appliance are extracted from the user behavior general network diagram, and spliced ​​with the user's current behavior network diagram to add a new node; when a user deletes an appliance midway, the corresponding appliance node and related edges in the user's current behavior network diagram are directly deleted.

6. A non-invasive load decomposition improvement device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 5.

7. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to perform the method according to any one of claims 1 to 5 when executed by the processor.