1. A non-invasive load decomposition method based on deep learning is characterized by comprising the following steps:

s1, acquiring short-term load historical data and total load historical data of each industrial device;

s2, training the initial model by adopting the short-term load historical data and the total load historical data in the step S1 to obtain a typical load model of the industrial equipment;

and S3, inputting the total load parameters obtained in real time into the typical load model of the industrial equipment to obtain the load of each industrial equipment.

2. The deep learning-based non-intrusive load splitting method of claim 1, wherein the total load history data includes total active power, total reactive power, current, and power factor.

3. The deep learning-based non-invasive load splitting method according to claim 1 or 2, wherein the short-term load history data is a differential value of corresponding data, and the differential value d is a differential value_i,tThe calculation formula of (2) is as follows:

d_i,t＝P_i,t-P_i,t-1,i＝1,2,…,N

wherein i represents the ith industrial equipment, N is the number of industrial equipment, t is time, P_i,tIs the data collected.

4. The deep learning based non-invasive load decomposition method according to claim 3, wherein the specific operation method of step S2 is: obtaining a differential value of certain data of certain industrial equipment in a preset time period, generating a power differential curve, carrying out fast Fourier change on the power differential curve to obtain an amplitude sequence, and inputting the amplitude sequence into the initial model for cluster training, thereby obtaining a typical load model of the industrial equipment.

5. The deep learning-based non-invasive load splitting method according to any one of claims 1 to 4, wherein the initial model is a long-short term memory artificial neural network model.

6. A deep learning based non-intrusive load splitting system, comprising the steps of:

the data acquisition module is used for acquiring short-term load historical data and total load historical data of each industrial device;

the model training module is used for training the initial model according to the short-term load historical data and the total load historical data to obtain a typical load model of the industrial equipment;

and the output module is used for inputting the total load parameters obtained in real time into the typical load model of the industrial equipment to obtain the load of each piece of industrial equipment.

7. A computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing industrial device, cause the computing industrial device to perform the deep learning based non-intrusive load splitting method according to any of claims 1 to 5.

8. A computing device, comprising: one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing the deep learning based non-invasive load splitting method of any of claims 1-5.

Background

Sensing, analyzing and managing industrial electric loads are important steps for promoting the construction of smart power grids. Load resolution is an important technique for sensing and identifying the load components and their status of a subject, also known as non-invasive load monitoring (NILM). The load decomposition problem mainly comprises three parts, namely data acquisition, load characteristic extraction and load decomposition method.

The load feature extraction method is mainly based on steady-state features and transient-state features of industrial equipment. Proposed methods include averaging, variable point monitoring, fourier decomposition, wavelet analysis, regression analysis, Hidden Markov Models (HMMs), and the like.

The load decomposition method mainly comprises an optimization method and a learning algorithm. The optimization method is mainly used for solving the state of the current electric industrial equipment through a planning method or a heuristic algorithm, such as a particle swarm algorithm, a genetic algorithm and the like. The learning algorithm decomposes the overall characteristics by learning the load characteristics of each industrial device, and mainly comprises a support vector machine, a K-nearest neighbor algorithm, an HMM-based algorithm and the like.

However, the load decomposition method in the prior art is almost all the target of commercial and residential electric loads. Different from commercial and residential electric loads, industrial equipment has the characteristics of strong periodicity, strong continuity and low stability, so that the industrial equipment and the load are difficult to decompose by using a method in the prior art.

Disclosure of Invention

In view of the above problems, it is an object of the present invention to provide a deep learning-based non-intrusive load splitting method, system, medium, and device, which can efficiently model loads of industrial devices and obtain loads of respective industrial devices through total loads of the industrial devices based on the model.

In order to achieve the purpose, the invention adopts the following technical scheme: a deep learning-based non-intrusive load decomposition method comprises the following steps: s1, acquiring short-term load historical data and total load historical data of each industrial device; s2, training the initial model by adopting the short-term load historical data and the total load historical data in the step S1 to obtain a typical load model of the industrial equipment; and S3, inputting the total load parameters obtained in real time into the typical load model of the industrial equipment to obtain the load of each industrial equipment.

Further, the total load history data includes total active power, total reactive power, current, and power factor.

Further, the short-term load history data is a differential value d of the corresponding data_i，tThe calculation formula of (2) is as follows:

d_i，t＝P_i，t-P_i，t-1，i＝1，2，...，N

wherein i represents the ith industrial equipment, N is the number of industrial equipment, t is time, P_i，tIs the data collected.

Further, the specific operation method of step S2 is: and obtaining a differential value of certain data of certain industrial equipment in a preset time period, generating a power differential curve, carrying out fast Fourier change on the power differential curve to obtain an amplitude sequence, and inputting the amplitude sequence into an initial model for cluster training, thereby obtaining a typical load model of the industrial equipment.

Further, the initial model is a long-term and short-term memory artificial neural network model.

The invention also discloses a non-invasive load decomposition system based on deep learning, which comprises the following steps: the data acquisition module is used for acquiring short-term load historical data and total load historical data of each industrial device; the model training module is used for training the initial model according to the short-term load historical data and the total load historical data to obtain a typical load model of the industrial equipment; and the output module is used for inputting the total load parameters obtained in real time into the typical load model of the industrial equipment to obtain the load of each piece of industrial equipment.

The present invention also discloses a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing industrial device, cause the computing industrial device to perform the privacy protection based image processing method according to any of the above.

The invention also discloses a computing device, comprising: one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing the privacy-based image processing method according to any of the above.

Due to the adoption of the technical scheme, the invention has the following advantages:

1. the invention can effectively model the load of the industrial equipment, and obtains the load of each industrial equipment through the total load of the industrial equipment based on the model and the non-intrusive load decomposition based on deep learning.

2. The invention can avoid the high cost caused by installing the load monitoring device for the industrial equipment, and simultaneously, the system only uses the bus load of the industrial user to protect the data security of the user.

Detailed Description

The present invention will be described in detail by way of specific examples in order to enable those skilled in the art to better understand the technical solutions of the present invention. It should be understood, however, that the detailed description is provided for a better understanding of the invention only and that they should not be taken as limiting the invention. In describing the present invention, it is to be understood that the terminology used is for the purpose of description only and is not intended to be indicative or implied of relative importance.

The invention provides a non-intrusive load decomposition method, a non-intrusive load decomposition system, a non-intrusive load decomposition medium and industrial equipment based on deep learning, and the use objects of the non-intrusive load decomposition system comprise industrial large users, power departments and other users needing to monitor the load state of industrial equipment produced by an industrial production enterprise. Aiming at the application requirement that a user needs to monitor the state of the industrial equipment load in industrial production, a non-invasive load decomposition system comprising an industrial user, an industrial user bus load monitoring device and a remote data acquisition terminal is constructed. The invention designs the modeling method of the industrial equipment load aiming at the characteristics of strong periodicity, strong continuity and weak stability of the industrial equipment load produced by a large industrial user, decomposes the total load of the industrial user by using a deep learning algorithm in the system to obtain the load state of each industrial equipment produced by the user, avoids high cost of installing and monitoring each industrial equipment and protects the data security of the user. The present invention will be described in detail with reference to specific examples.

Example one

The embodiment discloses a deep learning-based non-intrusive load decomposition method, which comprises the following steps:

s1, acquiring short-term load historical data and total load historical data of each industrial device; the total load history data includes, but is not limited to, total active power, total reactive power, current, and power factor. Taking the active power as an example, it is assumed that there are 4 industrial users and industrial devices under the users, that is, N is 4, and the active power of each industrial device at time t is P_i，tN, t is a time parameter, with a 15 minute per unit time interval, the bus load comprising the total active power of each industrial plant

S2, training the initial model by using the short-term load historical data and the total load historical data in the step S1 to obtain a typical load model of the industrial equipment.

The specific operation method of step S2 is: and obtaining a differential value of certain data of certain industrial equipment in a preset time period, generating a power differential curve, carrying out fast Fourier change on the power differential curve to obtain an amplitude sequence, and inputting the amplitude sequence into an initial model for cluster training, thereby obtaining a typical load model of the industrial equipment. Such as the active power obtained above, is differentially processed to capture active changes in the industrial equipment. Differential value d_i，tIs defined as d_i，t＝P_i，t-P_i，t-11, 2. Wherein i represents the ith industrial equipment, N is the number of industrial equipment, t is time, P_i，tIs the data collected. While the total active power differential value isGenerally speaking, the load state of each industrial device under the bus is difficult to be identified only through the total power value at the current moment, and the state change characteristics of the industrial devices can be shown only through a power curve for a period of time, so that the identification effect is achieved. Here, a section of the power difference curve v before the current time is taken_i，t＝[d_i，t-l+1，d_i，t-l，…，d_i，t]L is the length of the difference curve, which is subjected to Fast Fourier Transform (FFT) to obtain the amplitude sequence thereof

A_i，t＝{a₀，a₁，...，a_l-1}

In the above formula, a_jJ-0, 1, l-1 is the amplitude magnitude at different frequency values. Because the differential curve represents the change characteristics of the active power of the industrial equipment along with time, the amplitude sequence of the differential curve can provide unique load marks for different states and during state change. Therefore, the amplitude sequence is input into the initial model for clustering analysis, and a typical load model of the industrial equipment is obtained and used for representing the load state category of the whole industrial equipment. The initial model in this embodiment is preferably a long-short term memory artificial neural network model.

The clustering analysis in this embodiment adopts a K-means + + clustering algorithm, but other clustering methods may be adopted.

The specific method of the K-means + + clustering algorithm comprises the following steps: randomly selecting a sample from the amplitude sequence set as an initial clustering center c₁Calculating the shortest distance D (A) between the samples except the selected sample and the current initial cluster center_i，t)＝min(A_i，t-c_k)²And thus the probability of other samples being selected as the next cluster center is calculated: p (A)_i，t)＝D(A_i，t)²/∑_kD(A_i，k)²Selecting the next clustering center according to a turntable method; and repeating the steps until K clustering centers are selected.

Calculating the distances from the samples except the K clustering centers to the K clustering centers, and classifying the samples except the K clustering centers into the nearest clustering centers; and re-calculating the clustering center according to the current clustering result, and repeating the steps until the clustering center is not changed any more.

And S3, inputting the total load parameters obtained in real time into the typical load model of the industrial equipment to obtain the load of each industrial equipment.

In the embodiment, the state recognition accuracy of the typical load model of the industrial equipment can reach 88% after being tested.

Example two

Based on the same inventive concept, the embodiment discloses a non-intrusive load decomposition system based on deep learning, which comprises the following steps:

the data acquisition module is used for acquiring short-term load historical data and total load historical data of each industrial device;

the model training module is used for training the model according to the short-term load historical data and the total load historical data to obtain a typical load model of the industrial equipment;

and the output module is used for inputting the total load parameters obtained in real time into the typical load model of the industrial equipment to obtain the load of each piece of industrial equipment.

EXAMPLE III

Based on the same inventive concept, the present embodiments disclose a computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing industrial device, cause the computing industrial device to perform the privacy-based image processing method according to any one of the above.

Example four

Based on the same inventive concept, the present embodiment discloses a computing device, comprising: one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing the privacy-based image processing method according to any of the above.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, industrial devices (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing industrial equipment to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing industrial equipment, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing industrial equipment to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing industrial equipment to cause a series of operational steps to be performed on the computer or other programmable industrial equipment to produce a computer implemented process such that the instructions which execute on the computer or other programmable industrial equipment provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims. The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application should be defined by the claims.