CN115375921A - Two-stage non-intrusive load identification method and terminal - Google Patents
Two-stage non-intrusive load identification method and terminal Download PDFInfo
- Publication number
- CN115375921A CN115375921A CN202210692685.5A CN202210692685A CN115375921A CN 115375921 A CN115375921 A CN 115375921A CN 202210692685 A CN202210692685 A CN 202210692685A CN 115375921 A CN115375921 A CN 115375921A
- Authority
- CN
- China
- Prior art keywords
- load
- switching
- track
- switching load
- sampling frequency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/7625—Hierarchical techniques, i.e. dividing or merging patterns to obtain a tree-like representation; Dendograms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a two-stage non-intrusive load identification method and a terminal, wherein the method comprises the following steps: acquiring current data and voltage data of a switching load to construct a V-I track diagram; HOG feature extraction is carried out on the V-I track graph, and first-level classification is carried out on the HOG feature of the switching load according to a clustering algorithm so as to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; and extracting the multidimensional waveform characteristics of the switching load from the current data and inputting the multidimensional waveform characteristics into a secondary classification model corresponding to the track type of the switching load to obtain the identification result of the switching load. By dividing the loads with similar tracks into one track type, determining the track type of the switching load when identifying the load, and then further classifying the load in the determined track type, the technical problem that the loads with similar tracks cannot be effectively identified in the prior art can be solved, and the accuracy of load identification is effectively improved.
Description
Technical Field
The application belongs to the technical field of load identification, and particularly relates to a two-stage non-intrusive load identification method and a terminal.
Background
The non-intrusive load identification means that monitoring equipment is installed at an electric power inlet of a user, and the type and the operation condition of a single load in a load cluster can be obtained through analyzing signals such as voltage, current and the like at the position by monitoring.
The V-I track of the electric appliance is used as an effective load characteristic and widely applied to non-intrusive load identification, but the V-I track graph is drawn by normalized current and voltage data, so that the magnitude of the current of the electric appliance in working cannot be represented in principle. Therefore, different electrical appliances with similar V-I tracks cannot be effectively distinguished by using a single V-I track.
Disclosure of Invention
In view of this, the invention provides a two-stage non-intrusive load identification method and a terminal, and aims to solve the problem that a V-I locus diagram in the prior art cannot identify similar loads.
The first aspect of the embodiments of the present invention provides a two-stage non-intrusive load identification method, including:
after a load switching event occurs, acquiring current data and voltage data of a switching load to construct a V-I track map;
carrying out feature extraction on the V-I track graph to obtain the HOG feature of the switching load;
carrying out primary classification on the HOG characteristics of the switching load according to a clustering algorithm to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; the secondary classification model is established according to a machine learning algorithm;
extracting the multi-dimensional waveform characteristics of the switching load from the current data;
and inputting the multi-dimensional waveform characteristics into a secondary classification model corresponding to the track type of the switching load to obtain the electric appliance type of the switching load, and taking the electric appliance type of the switching load as the identification result of the switching load.
A second aspect of an embodiment of the present invention provides a two-stage non-intrusive load identification apparatus, including:
the data acquisition module is used for acquiring current data and voltage data of switching loads to construct a V-I track graph after a load switching event occurs;
the first extraction module is used for extracting the characteristics of the V-I track graph to obtain the HOG characteristics of the switching load;
the first classification module is used for carrying out primary classification on the HOG characteristics of the switching load according to a clustering algorithm so as to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; the secondary classification model is established according to a machine learning algorithm;
the second extraction module is used for extracting the multi-dimensional waveform characteristics of the switching load from the current data;
and the second classification module is used for inputting the multi-dimensional waveform characteristics into a secondary classification model corresponding to the switching load track type to obtain the switching load electrical appliance type and taking the switching load electrical appliance type as the switching load identification result.
A third aspect of the embodiments of the present invention provides a non-intrusive load identification terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the two-stage non-intrusive load identification method according to the first aspect when executing the computer program.
A fourth aspect of the embodiments of the present invention provides a computer-readable storage medium, which stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the two-stage non-intrusive load identification method according to the first aspect.
The two-stage non-intrusive load identification method and the terminal provided by the embodiment of the invention comprise the following steps: after a load switching event occurs, acquiring current data and voltage data of a switching load to construct a V-I track graph; HOG feature extraction is carried out on the V-I track graph, and first-level classification is carried out on the HOG feature of the switching load according to a clustering algorithm so as to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; establishing a secondary classification model according to a machine learning algorithm; and extracting the multi-dimensional waveform characteristics of the switching load from the current data and inputting the multi-dimensional waveform characteristics into a secondary classification model corresponding to the track type of the switching load to obtain the identification result of the switching load. The loads with similar tracks are divided into the track types, the track type of the switching load is determined firstly during load identification, and then the loads are further classified in the determined track type, so that the technical problem that the loads with similar tracks cannot be effectively identified in the prior art can be solved, and the accuracy of load identification is effectively improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required to be used in the embodiments or the prior art description will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings may be obtained according to these drawings without inventive labor.
Fig. 1 is an application scenario diagram of a two-stage non-intrusive load identification method according to an embodiment of the present invention;
fig. 2 is a flowchart of an implementation of a two-stage non-intrusive load identification method according to an embodiment of the present invention;
fig. 3 is a V-I trace diagram of various types of loads with 32 x 32 pixels at 6.4kHz according to an embodiment of the present invention;
fig. 4 is a V-I trace diagram of various types of loads with 32 × 32 pixels at 3.2kHz according to an embodiment of the present invention;
FIG. 5 is a graph of the V-I trace for various types of loads having pixels of 20 × 20 at 3.2kHz, according to an embodiment of the present invention;
FIG. 6 is a graph of the sum of squared cluster centers-error for the K-means clustering algorithm provided by the present invention;
FIG. 7 shows a load recognition result of KNN load recognition;
FIG. 8 is a preferred KNN load recognition result based on PSO;
fig. 9 is a schematic structural diagram of a two-stage non-intrusive load identification device according to an embodiment of the present invention;
fig. 10 is a schematic structural diagram of a non-intrusive load identification terminal according to an embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
Fig. 1 is an application scenario diagram of a two-stage non-intrusive load identification method according to an embodiment of the present invention. As shown in fig. 1, the two-stage non-intrusive load identification method provided in the embodiment of the present invention may be applied to, but is not limited to, the application scenario. In this embodiment, the system includes: the system comprises a smart electric meter 11, a non-intrusive load identification terminal 12 and a server 13.
The intelligent electric meter 11 is arranged at an inlet end of the user power utilization system and used for measuring current data and voltage data of the user power utilization system and reporting the current data and the voltage data to the non-invasive load identification terminal 12. The non-intrusive load identification terminal 12 is used for carrying out load identification according to the current data and the voltage data and sending an identification result to the server 13.
The non-intrusive load identification terminal 12 may be integrated in the smart meter 11, or may be a separate device connected to the smart meter 11, which is not limited herein. The consumer power utilization system may be a home power utilization system including a plurality of home power utilization devices, or may be a small area power utilization system including a plurality of companies, shopping malls, and the like, which includes a plurality of devices, and is not limited herein. The sampling frequency of the smart meter can be 6.4KHz. The server 13 may be an independent physical server, or may be a cluster composed of a plurality of servers, and is not limited herein.
Fig. 2 is a flowchart of an implementation of a two-stage non-intrusive load identification method according to an embodiment of the present invention. Referring to fig. 2, in this embodiment, a two-stage non-intrusive load identification method is applied to the non-intrusive load identification terminal 12 shown in fig. 1, and includes:
s201, after a load switching event occurs, acquiring current data and voltage data of a switching load to construct a V-I track diagram.
In this embodiment, the switching load is an electric device switched in or out from a user electric system, and after a load switching event occurs, current data and voltage data of the load during stable operation are obtained, and then the current data and the voltage data before the load switching event occurs are subtracted, so that the current data and the voltage data of the switching load can be obtained.
In this embodiment, after obtaining the current data and the voltage data of the switching load, normalization needs to be performed:
wherein i (t) is the current data of the switching load, v (t) is the voltage data of the switching load, i m (t) Current data for normalized post-load shedding, v m (t) Voltage data of the load cut after normalization, i min And i max Minimum current and maximum current, v, of switching load in one cycle min And v max The minimum voltage and the maximum voltage of the switching load in one period are respectively.
Setting the resolution of the V-I track diagram to be NxN, respectively multiplying the normalized current data and voltage data by N, and carrying out upward value-taking processing to obtain a group of current and voltage data which are greater than or equal to 1 and less than or equal to N, wherein the calculation formula is as follows:
i n (t)＝ceil(i m (t)*N) (3)
v n (t)＝ceil(v m (t)*N) (4)
where ceil (·) is an upward rounding function in MATLAB, N =1,2, \8230, N.
And constructing an NxN zero matrix. For all i n (t) and v n (t) taking the value N-i n (t) +1 and v n (t) N-i respectively representing zero matrix n (t) +1 line and v n And (t) arranging, and changing the value of the corresponding point in the zero matrix into 1 to obtain the V-I locus diagram of the switching load.
And S202, carrying out feature extraction on the V-I track graph to obtain the HOG feature of the switching load.
In this embodiment, a HOG (Histogram of Oriented gradients) feature is a feature descriptor used for object detection in computer vision and image processing, and the HOG feature descriptor is sensitive to gradients and directions and can well describe an image contour. At the same time, the HOG features can remain well invariant to geometric deformation. The basic principle is as follows: dividing the target image into a plurality of small cell units (cells), combining the small cell units into 1 block unit, then collecting gradients of all pixel points in each cell, and finally combining the characteristics to form the HOG descriptor.
Although the V-I locus diagrams have good discrimination, the V-I locus diagrams of some electric appliances have similar conditions, so that the electric appliances with similar V-I loci cannot be accurately identified, and the electric appliances with similar loci are classified into one type and then subjected to subsequent fine identification. The HOG features can maintain the characteristics of geometric deformation and therefore can well describe the V-I trajectory's outline.
S203, carrying out primary classification on the HOG characteristics of the switching load according to a clustering algorithm to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; and establishing a secondary classification model according to a machine learning algorithm.
In this embodiment, the Clustering algorithm may be a Density-Based noisy Spatial Clustering algorithm (Density-Based Clustering of Applications with Noise, DBSCAN), a K-means Clustering algorithm (K-means Clustering), and the like, which are not limited herein.
In this embodiment, the loads with similar tracks may be classified into a large class (i.e., a track type), the primary classification determines the large class to which the switching load belongs, and then the corresponding secondary classification model is used to perform further classification in the large class, so as to obtain the appliance type of the switching load.
And S204, extracting the multi-dimensional waveform characteristics of the switching load from the current data.
In this embodiment, the multi-dimensional waveform features may include, but are not limited to, at least one of: current average, current root mean square value, peak coefficient, form coefficient, skewness factor, impulse factor, margin factor, kurtosis factor.
And S205, inputting the multi-dimensional waveform characteristics into a secondary classification model corresponding to the switching load track type to obtain the switching load electric appliance type, and taking the switching load electric appliance type as a switching load identification result.
In this embodiment, the secondary classification model may be a K-nearest neighbor classification model, a neural network model, and the like, which is not limited herein.
In the embodiment, after a load switching event occurs, current data and voltage data of a switching load are acquired to construct a V-I track diagram; carrying out feature extraction on the V-I track graph to obtain HOG features of switching loads; carrying out primary classification on the HOG characteristics of the switching load according to a clustering algorithm to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; establishing a secondary classification model according to a machine learning algorithm; extracting multi-dimensional waveform characteristics of switching loads from the current data; and inputting the multi-dimensional waveform characteristics into a secondary classification model corresponding to the track type of the switching load to obtain the type of the electric appliance of the switching load and taking the type of the electric appliance as the identification result of the switching load. By dividing the loads with similar tracks into one track type, determining the track type of the switching load when identifying the load, and then further classifying the load in the determined track type, the technical problem that the loads with similar tracks cannot be effectively identified in the prior art can be solved, and the accuracy of load identification is effectively improved.
In some embodiments, before S201, the method may further include:
acquiring the current sampling frequency of a switching load;
selecting pixels of the V-I locus diagram from the first relation table according to the current sampling frequency of the switching load; the first relation table is a corresponding relation table of sampling frequencies and pixels, and each sampling frequency corresponds to a pixel of a V-I track graph; the sampling frequency and the pixel in the corresponding relation table are in positive correlation;
the construction of the V-I track map comprises the following steps:
and constructing a V-I track graph according to the current data, the voltage data and the pixels.
In this embodiment, the corresponding relationship between the sampling frequency and the pixel in the first relationship table may be determined through multiple experiments, or may be determined through expert experience, which is not limited herein.
Fig. 3 is a V-I trace diagram of various types of loads with 32 × 32 pixels at 6.4kHz according to an embodiment of the present invention. As shown in fig. 3, in some embodiments, the current sampling frequency of the switching load is 6.4kHz, and the current sampling frequency corresponds to 32 × 32 pixels.
In this embodiment, current data and voltage data of 9 kinds of electric appliances such as an air conditioner (refrigeration), a microwave oven, a water heater, a hot water bottle, an electric heating furnace, an electric cooker, an electromagnetic oven, a washing machine, and a range hood during operation are collected. It can be seen that the continuity of each V-I trace diagram is better, wherein the air conditioner (refrigeration) and the microwave oven belong to the same trace type, the water heater, the hot water kettle, the electric heating furnace and the electric rice cooker belong to the same trace type, and the induction cooker, the washing machine and the range hood belong to the same trace type.
In some embodiments, the current sampling frequency of the switching load is 3.2kHz, and the current sampling frequency corresponds to 20 × 20 pixels.
The higher the sampling frequency, the higher the cost of the practical application device, and in this embodiment, the device cost can be reduced by lowering the sampling frequency. Fig. 4 is a graph of the V-I trace for each type of load with 32 x 32 pixels at 3.2kHz, provided by an embodiment of the present invention. As shown in fig. 4, as the sampling frequency decreases, the continuity of the V-I trace map decreases, resulting in inaccurate load recognition.
Fig. 5 is a V-I trace diagram of various types of loads with 20 × 20 pixels at 3.2kHz according to an embodiment of the present invention. As shown in fig. 5, when the sampling frequency is reduced, the pixels of the V-I trace map can be correspondingly reduced, so as to ensure the continuity of the V-I trace map, and thus accurately perform load identification.
In some embodiments, before S201, the method may further include:
acquiring the current sampling frequency of a switching load;
adjusting pixels of the V-I track map according to the current sampling frequency so as to enable the track of the constructed V-I track map to be continuous; wherein, the sampling frequency and the pixel in the corresponding relation table are in positive correlation.
In the present embodiment, when the sampling frequency is decreased, the pixels are not limited to the above values as long as the pixels are decreased to the trajectory continuation of the V-I trajectory diagram.
And the mode that the pixels are correspondingly reduced along with the sampling frequency so as to reduce the equipment cost is set for two-stage identification of the scheme, the conventional load identification has poor identification on similar loads, the similar loads are more difficult to identify after the pixels are reduced, and the two-stage load identification can still keep better identification effect after the pixels are reduced.
In some embodiments, S203 may include:
and carrying out primary classification on the HOG characteristics of the switching load according to a K-means clustering algorithm to obtain the track type of the switching load.
Fig. 6 is a graph of a sum of squared cluster centers-error of the K-means clustering algorithm provided by the embodiment of the present invention. As shown in fig. 6, the horizontal axis represents the number of clusters and the vertical axis represents the sum of squares of errors. In this embodiment, 100 training sets may be used for K-means clustering to find the clustering center, and 200 test sets may be used for verifying the validity of the clustering center.
In order to determine the value of the K-means clustering center number K, a clustering test is carried out on the test set, the sum of the squares (SSE) of the clustering results under different K values is calculated, and the value of the clustering center number K is determined according to the SSE value. It can be seen that k =3 is the optimal number of clusters. Selecting k =3, respectively marking labels A, B and C on three clustering centers, and performing primary classification on each electric appliance by using a test set. For example, the 9 electric appliances shown in fig. 3 can be classified into three categories, i.e., a category a electric appliance including an air conditioner (cooling) and a microwave oven, a category B electric appliance including a water heater, a hot water kettle, an electric heating furnace and an electric cooker, and a category C electric appliance including an induction cooker, a washing machine and a range hood.
In some embodiments, S205 may include:
and inputting the multi-dimensional waveform characteristics into a K nearest neighbor classification model corresponding to the track type of the switching load to obtain the type of the switching load and taking the type of the switching load as an identification result of the switching load.
In this embodiment, the K nearest neighbor algorithm is a classic machine learning algorithm, and is simple and easy to implement. And the KNN classifies the samples to be detected by comparing the distance or the similarity between the training samples and the samples to be detected. The method comprises the following specific steps:
(1) Calculating the Euclidean distance between a sample to be detected and a training sample;
(2) Finding out r points which are most adjacent to the sample to be detected in the training sample;
(3) Counting the number of each category in the r points;
(4) And taking the class with the highest occurrence frequency in the r points as the prediction class of the sample point.
Because the KNN algorithm has a large dependence on data, and the extracted HOG features may have excessive dimensionality, redundant features, even noise features and the like, which all result in the reduction of the accuracy rate of KNN classification, the PSO can be adopted for feature optimization to search for an optimal feature subset.
In some embodiments, the method further comprises:
according to a particle swarm algorithm, selecting a feature subset with the lowest error identification rate from training samples of the K nearest neighbor classification model as an optimal feature subset of the K nearest neighbor classification model;
establishing a K nearest neighbor classification model according to the optimal feature subset;
the error identification rate is obtained according to the test set, and each track type corresponds to one test set.
The PSO algorithm is a heuristic algorithm based on group intelligence and is commonly used for solving multiple targets, nonlinearity and multivariable. The particle swarm optimization process starts from randomly generated initial particles, continuously updates the positions and the speeds of the particles, and uses a fitness value to represent the advantages and the disadvantages of the particles. The optimal solution is found by tracking the optimal particles.
In this embodiment, the optimal feature subset is selected through the PSO, so that the classification accuracy of the KNN algorithm can be effectively improved, the feature dimension is also reduced, the spatial complexity of the KNN classification is reduced, and the algorithm implementation process is as follows:
(1) Initializing parameters of a particle swarm algorithm: for the population number m, the position and the speed of each particle, the maximum iteration number H, the inertia factor w and the learning factor c 1 、c 2 Etc. for random initialization.
(2) The particle position and velocity are updated using the following equation:
wherein Pbestid represents a locally optimal position of the individual; gbestid represents the global optimal position of the whole population; h =1,2, \8230, H, which is the number of iterations; w is an inertia factor; r is a radical of hydrogen 1 、r 2 Is [0,1 ]]A random number in between; c. C 1 、c 2 Is a learning factor.
(3) Taking the error identification rate of the KNN classification as a fitness function, and updating Pbestid and Gestid according to the fitness function;
(4) Judging whether an ending condition is met, namely: the maximum number of iterations H is reached or the minimum error requirement is reached. If the end condition is met, ending the program and outputting an optimal result, and performing (5); otherwise, returning to the step (2) to continue execution;
(5) The optimal feature subset found by the optimization is used for KNN modeling.
To further illustrate the improvement effect of the PSO on the KNN algorithm by selecting the optimal feature subset, the optimal KNN load identification based on the PSO is given below (not the two-stage load identification of the present application, but only KNN is used):
fig. 7 shows a load recognition result of KNN load recognition. Fig. 8 is a KNN load recognition result based on PSO optimization. As shown in fig. 7 and 8, steady-state current and steady-state voltage data are extracted from the collected data, and 100 sets of steady-state current data are selected as a training set and 200 sets of steady-state current data are selected as a test set for each electrical appliance. And (3) respectively extracting seven waveform characteristics of a peak coefficient, a waveform coefficient, a skewness factor, a pulse factor, a margin factor, a kurtosis factor and a kurtosis factor from 100 training sets and 200 testing sets, and using the seven waveform characteristics to train the KNN model and verify the model. As can be seen from fig. 7 and 8, the accuracy of KNN recognition is significantly improved after PSO optimization.
In some embodiments, the multi-dimensional waveform features may include, but are not limited to, at least one of: current average, current root mean square value, peak coefficient, form coefficient, skewness factor, impulse factor, margin factor, kurtosis factor.
In this embodiment, time domain characteristics of current signals have certain differences when different loads operate in a steady state, and through analyzing time domain waveforms of the current signals, dimensionless amount data of the current is extracted to be used for representing waveform characteristics. Let the current signal be x (N), N =1, 2.
Wherein, P 1 、P 2 Is the mean and root mean square (i.e. effective) value of the current, P 3 ～P 9 The method is characterized by sequentially comprising seven waveform characteristics of a peak coefficient, a waveform coefficient, a skewness factor, a pulse factor, a margin factor, a kurtosis factor and a kurtosis factor, and mainly used for representing the distortion degree, the existence of impact and the smoothness degree of a current waveform.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
Fig. 9 is a schematic structural diagram of a two-stage non-intrusive load identification device according to an embodiment of the present invention. As shown in fig. 9, in some embodiments, the two-stage non-intrusive load identification device 9 includes:
the data acquisition module 910 is configured to acquire current data and voltage data of a switching load to construct a V-I trajectory diagram after a load switching event occurs;
the first extraction module 920 is configured to perform feature extraction on the V-I trajectory graph to obtain an HOG feature of a switching load;
the first classification module 930 is configured to perform first-level classification on the HOG features of the switching loads according to a clustering algorithm to obtain a trajectory type of the switching loads; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; establishing a secondary classification model according to a machine learning algorithm;
a second extraction module 940, configured to extract a multi-dimensional waveform feature of the switching load from the current data;
and the second classification module 950 is configured to input the multi-dimensional waveform characteristics into the secondary classification model corresponding to the trajectory type of the switching load, so as to obtain an electrical appliance type of the switching load, and use the electrical appliance type as an identification result of the switching load.
Optionally, the two-stage non-intrusive load identification apparatus 9 further includes: a pixel adjustment module 960.
Optionally, the pixel adjusting module 960 is configured to:
acquiring the current sampling frequency of a switching load;
selecting pixels of the V-I locus diagram from the first relation table according to the current sampling frequency of the switching load; the first relation table is a corresponding relation table of sampling frequencies and pixels, and each sampling frequency corresponds to a pixel of a V-I locus diagram; the sampling frequency and the pixel in the corresponding relation table are in positive correlation;
the construction of the V-I track map comprises the following steps:
and constructing a V-I track map according to the current data, the voltage data and the pixels.
Optionally, the current sampling frequency of the switching load is 6.4kHz, and the pixel corresponding to the current sampling frequency is 32 × 32;
or, the current sampling frequency of the switching load is 3.2kHz, and the pixel corresponding to the current sampling frequency is 20 × 20.
Optionally, the pixel adjusting module 960 is configured to:
acquiring the current sampling frequency of a switching load;
adjusting pixels of the V-I track map according to the current sampling frequency so as to enable the track of the constructed V-I track map to be continuous; wherein, the sampling frequency and the pixel in the corresponding relation table are in positive correlation.
Optionally, the first classification module 930 is configured to perform first-level classification on the HOG features of the switching load according to a K-means clustering algorithm to obtain a trajectory type of the switching load.
Optionally, the second classification module 950 is configured to:
and inputting the multi-dimensional waveform characteristics into the K nearest neighbor classification model corresponding to the track type of the switching load to obtain the type of the electric appliance of the switching load, and taking the type of the electric appliance as the identification result of the switching load.
Optionally, the two-stage non-intrusive load identification device 9 further includes: model building module 970.
According to a particle swarm algorithm, selecting a feature subset with the lowest error identification rate from training samples of the K nearest neighbor classification model as an optimal feature subset of the K nearest neighbor classification model;
establishing a K nearest neighbor classification model according to the optimal feature subset;
the error identification rate is obtained according to the test set, and each track type corresponds to one test set.
Optionally, the multi-dimensional waveform features include at least one of: current average value, current root mean square value, peak coefficient, waveform coefficient, skewness factor, pulse factor, margin factor, kurtosis and kurtosis factor.
The two-stage non-intrusive load identification device provided in this embodiment may be used to implement the method embodiments, and the implementation principle and technical effects are similar, which are not described herein again.
Fig. 10 is a schematic diagram of a detection apparatus provided in an embodiment of the present invention. As shown in fig. 10, an embodiment of the present invention provides a non-intrusive load identification terminal 10, where the non-intrusive load identification terminal 10 of the embodiment includes: a processor 1000, a memory 1010, and a computer program 1020 stored in the memory 1010 and operable on the processor 1000. The processor 1000, when executing the computer program 1020, implements the steps of the two-stage non-intrusive load identification method embodiments described above, such as steps 201 through 205 shown in fig. 2. Alternatively, the processor 1000, when executing the computer program 1020, implements the functions of each module/unit in each system embodiment described above, for example, the functions of the modules 910 to 950 shown in fig. 9.
Illustratively, the computer program 1020 may be partitioned into one or more modules/units, which are stored in the memory 1010 and executed by the processor 1000 to implement the present invention. One or more of the modules/units may be a series of computer program instruction segments capable of performing certain functions that describe the execution of the computer program 1020 in the non-intrusive load identification terminal 10.
The non-intrusive load identification terminal 10 may be a single chip microcomputer, a MCU, a desktop computer, a notebook, a palm computer, or other computing devices. The terminal may include, but is not limited to, a processor 1000, a memory 1010. Those skilled in the art will appreciate that fig. 10 is merely an example of a non-intrusive load identification terminal 10 and does not constitute a limitation on a non-intrusive load identification terminal 10 and may include more or fewer components than those shown, or some components may be combined, or different components, e.g., the terminal may also include input output devices, network access devices, buses, etc.
The Processor 1000 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 1010 may be an internal storage unit of the non-intrusive load identification terminal 10, such as a hard disk or a memory of the non-intrusive load identification terminal 10. The memory 1010 may also be an external storage device of the non-intrusive load identification terminal 10, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, provided on the non-intrusive load identification terminal 10. Further, the memory 1010 may also include both internal and external memory units of the non-intrusive load identification terminal 10. The memory 1010 is used for storing computer programs and other programs and data required by the terminal. The memory 1010 may also be used to temporarily store data that has been output or is to be output.
An embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the steps in the two-stage non-intrusive load identification system embodiment are implemented.
The computer-readable storage medium stores a computer program 1020, the computer program 1020 includes program instructions, and when the program is executed by the processor 1000, all or part of the processes in the method according to the embodiments are implemented, or the program 1020 is implemented by hardware related to the instructions, and the computer program 1020 may be stored in a computer-readable storage medium, and when the computer program 1020 is executed by the processor 1000, the steps of the method embodiments may be implemented. Computer program 1020 includes, among other things, computer program code, which may be in the form of source code, object code, an executable file or some intermediate form. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like. It should be noted that the computer readable medium may include any suitable increase or decrease as required by legislation and patent practice in the jurisdiction, for example, in some jurisdictions, computer readable media may not include electrical carrier signals and telecommunications signals in accordance with legislation and patent practice.
The computer readable storage medium may be an internal storage unit of the terminal of any of the foregoing embodiments, such as a hard disk or a memory of the terminal. The computer readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk provided on the terminal, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of the terminal. The computer-readable storage medium is used for storing a computer program and other programs and data required by the terminal. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
It should be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional units and modules is only used for illustration, and in practical applications, the above function distribution may be performed by different functional units and modules as needed, that is, the internal structure of the device is divided into different functional units or modules, so as to perform all or part of the above described functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only used for distinguishing one functional unit from another, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal and method may be implemented in other ways. For example, the above-described apparatus/terminal embodiments are merely illustrative, and for example, a module or a unit may be divided into only one logical function, and may be implemented in other ways, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow in the method according to the embodiments of the present invention may also be implemented by a computer program instructing related hardware, and the computer program may be stored in a computer readable storage medium, and when executed by a processor, the computer program may implement the steps of the above-described embodiments of the method. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like. It should be noted that the computer readable medium may include any suitable increase or decrease as required by legislation and patent practice in the jurisdiction, for example, in some jurisdictions, computer readable media may not include electrical carrier signals and telecommunications signals in accordance with legislation and patent practice.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (10)
1. A two-stage non-intrusive load identification method, comprising:
after a load switching event occurs, acquiring current data and voltage data of a switching load to construct a V-I track graph;
carrying out feature extraction on the V-I track graph to obtain the HOG feature of the switching load;
carrying out primary classification on the HOG characteristics of the switching load according to a clustering algorithm to obtain the track type of the switching load; switching loads with similar tracks of the V-I track graph belong to the same track type; each track type corresponds to a secondary classification model; the secondary classification model is established according to a machine learning algorithm;
extracting multi-dimensional waveform characteristics of the switching load from the current data;
and inputting the multi-dimensional waveform characteristics into a secondary classification model corresponding to the track type of the switching load to obtain the electric appliance type of the switching load, and taking the electric appliance type of the switching load as the identification result of the switching load.
2. The two-stage non-intrusive load identification method according to claim 1, before acquiring the current data and the voltage data of the switching load to construct a V-I trajectory diagram, further comprising:
acquiring the current sampling frequency of the switching load;
selecting pixels of the V-I locus diagram from a first relation table according to the current sampling frequency of the switching load; the first relation table is a corresponding relation table of sampling frequencies and pixels, and each sampling frequency corresponds to a pixel of a V-I track graph; the sampling frequency and the pixel in the corresponding relation table are in positive correlation;
the constructing of the V-I trajectory graph comprises the following steps:
and constructing a V-I track graph according to the current data, the voltage data and the pixels.
3. The two-stage non-intrusive load identification method according to claim 2, wherein the switching load has a current sampling frequency of 6.4kHz, and the current sampling frequency corresponds to 32 × 32 pixels;
or, the current sampling frequency of the switching load is 3.2kHz, and the pixel corresponding to the current sampling frequency is 20 × 20.
4. The two-stage non-intrusive load identification method as defined in claim 1, further comprising, prior to obtaining current data and voltage data for the switched load to construct a V-I trace plot:
acquiring the current sampling frequency of the switching load;
adjusting pixels of the V-I track map according to the current sampling frequency so as to enable the track of the constructed V-I track map to be continuous; wherein, the sampling frequency and the pixel in the corresponding relation table are in positive correlation.
5. The two-stage non-intrusive load identification method according to claim 1, wherein the step of classifying the HOG features of the switching loads at one stage according to a clustering algorithm to obtain the track types of the switching loads comprises:
and carrying out primary classification on the HOG characteristics of the switching load according to a K-means clustering algorithm to obtain the track type of the switching load.
6. The two-stage non-intrusive load identification method according to claim 1, wherein the step of inputting the multidimensional waveform characteristics into a two-stage classification model corresponding to the switching load trajectory type to obtain the switching load appliance type as the switching load identification result comprises the steps of:
and inputting the multi-dimensional waveform characteristics into a K nearest neighbor classification model corresponding to the track type of the switching load to obtain the electric appliance type of the switching load and taking the electric appliance type of the switching load as the identification result of the switching load.
7. The two-stage non-intrusive load identification method of claim 6, further comprising:
according to a particle swarm algorithm, selecting a feature subset with the lowest error identification rate from training samples of the K nearest neighbor classification model as an optimal feature subset of the K nearest neighbor classification model;
establishing the K nearest neighbor classification model according to the optimal feature subset;
and the error identification rate is obtained according to the test of the test sets, and each track type corresponds to one test set.
8. A two-stage non-intrusive load identification method as defined in any of claims 1-7, wherein the multi-dimensional waveform characteristics include at least one of: current average, current root mean square value, peak coefficient, form coefficient, skewness factor, impulse factor, margin factor, kurtosis factor.
9. A non-intrusive load recognition terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the two-stage non-intrusive load recognition method of any of claims 1 to 8 above.
10. A computer-readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the steps of the two-stage non-intrusive load identification method as defined in any one of claims 1 to 8 above.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210692685.5A CN115375921A (en) | 2022-06-17 | 2022-06-17 | Two-stage non-intrusive load identification method and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210692685.5A CN115375921A (en) | 2022-06-17 | 2022-06-17 | Two-stage non-intrusive load identification method and terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115375921A true CN115375921A (en) | 2022-11-22 |
Family
ID=84062309
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210692685.5A Pending CN115375921A (en) | 2022-06-17 | 2022-06-17 | Two-stage non-intrusive load identification method and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115375921A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115982576A (en) * | 2023-03-17 | 2023-04-18 | 石家庄科林电气股份有限公司 | Malignant load identification method and device and electric energy meter |
| CN118114137A (en) * | 2024-04-30 | 2024-05-31 | 浙江大华技术股份有限公司 | Non-invasive load identification method, device and computer storage medium |
-
2022
- 2022-06-17 CN CN202210692685.5A patent/CN115375921A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115982576A (en) * | 2023-03-17 | 2023-04-18 | 石家庄科林电气股份有限公司 | Malignant load identification method and device and electric energy meter |
| CN118114137A (en) * | 2024-04-30 | 2024-05-31 | 浙江大华技术股份有限公司 | Non-invasive load identification method, device and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115375921A (en) | Two-stage non-intrusive load identification method and terminal | |
| CN107272655B (en) | Batch process fault monitoring method based on multistage ICA-SVDD | |
| CN105809125B (en) | Face recognition system based on multi-core ARM platform | |
| CN110533089B (en) | Self-adaptive non-invasive load identification method based on random forest | |
| EP3364157A1 (en) | Method and system of outlier detection in energy metering data | |
| CN104063719A (en) | Method and device for pedestrian detection based on depth convolutional network | |
| CN113542241B (en) | Intrusion detection method and device based on CNN-BiGRU hybrid model | |
| CN109145516B (en) | Analog circuit fault identification method based on improved extreme learning machine | |
| CN107579846B (en) | Cloud computing fault data detection method and system | |
| CN112821559B (en) | Non-invasive household appliance load depth re-identification method | |
| CN112732748B (en) | Non-invasive household appliance load identification method based on self-adaptive feature selection | |
| CN113255236B (en) | Non-invasive load self-adaptive identification method based on twin network | |
| CN111626360B (en) | Method, apparatus, device and storage medium for detecting boiler fault type | |
| CN112381248A (en) | Power distribution network fault diagnosis method based on deep feature clustering and LSTM | |
| CN111144522A (en) | Power grid NFC equipment fingerprint authentication method based on hardware intrinsic difference | |
| CN114359674A (en) | Non-invasive load identification method based on metric learning | |
| CN109726770A (en) | A kind of analog circuit fault testing and diagnosing method | |
| CN110502989A (en) | A kind of small sample EO-1 hyperion face identification method and system | |
| CN115100466A (en) | Non-invasive load monitoring method, device and medium | |
| CN113627289A (en) | Deep learning-based power equipment identification method and device | |
| CN117170979B (en) | Energy consumption data processing method, system, equipment and medium for large-scale equipment | |
| CN117115197A (en) | Intelligent processing method and system for design data of LED lamp bead circuit board | |
| CN116628620A (en) | Non-invasive load identification calculation method | |
| CN109031057A (en) | A kind of high-tension electricity apparatus local discharge mode identification method and device | |
| CN112256766A (en) | Power consumption behavior analysis method for energy collection terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |